US20070300295A1 - Systems and methods to extract data automatically from a composite electronic document - Google Patents

Systems and methods to extract data automatically from a composite electronic document Download PDF

Info

Publication number
US20070300295A1
US20070300295A1 US11/472,868 US47286806A US2007300295A1 US 20070300295 A1 US20070300295 A1 US 20070300295A1 US 47286806 A US47286806 A US 47286806A US 2007300295 A1 US2007300295 A1 US 2007300295A1
Authority
US
United States
Prior art keywords
contract
document
list
pattern
recited
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Abandoned
Application number
US11/472,868
Inventor
Thomas Yu-Kiu Kwok
Thao Ngoc Nguyen
Kakan Roy
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
International Business Machines Corp
Original Assignee
International Business Machines Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by International Business Machines Corp filed Critical International Business Machines Corp
Priority to US11/472,868 priority Critical patent/US20070300295A1/en
Assigned to INTERNATIONAL BUSINESS MACHINES CORPORATION reassignment INTERNATIONAL BUSINESS MACHINES CORPORATION ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: NGUYEN, THAO NGOC, ROY, KAKAN, KWOK, THOMAS YU-KIU
Publication of US20070300295A1 publication Critical patent/US20070300295A1/en
Priority to US12/132,845 priority patent/US8140468B2/en
Abandoned legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00Administration; Management
    • G06Q10/10Office automation; Time management

Definitions

  • the present invention relates to data extraction from documents and more particularly to systems and methods which extract contract data automatically and efficiently from an electronic contract composed of a number of documents in a given format.
  • a single electronic contract can encompass a large number of collateral documents including master and customer agreements, supplements, addenda and the like. These various documents are of different contract document types. There can be over a hundred different basic types of contract documents in a large company. A few examples of these contract document types are as follows: “Master Agreement”, “Customer Agreement”, “Term Lease Supplement”, “Addendum to Term Lease Supplement”, “Statement of Work for Services”, “Change Authorization for Services”, etc. Moreover, they can also be in different file formats, such as PDF, XML, Microsoft WordTM, Lotus WordProTM.
  • An electronic contract management system can be used to automatically convert all these contract documents of different types into PDF format and then merge them together to form a single electronic contract PDF document.
  • data extraction and mining on this kind of electronic contract is still very difficult if not impossible.
  • a user should find out how many contract documents are in an electronic contract composed of a number of contract documents, and then determine their contract document types.
  • what contract data to extract should be decided and from which contract document. The user would further need to find out where on the contract document the contract data is located, such as page and line numbers.
  • An electronic contract can encompass a large number of collateral contract documents in, e.g., PDF format. These contract documents are of different contract document types and converted from different original formats. Data extraction and thus data mining for this kind of electronic contracts is very difficult.
  • a novel system and method are presented to automatically extract contract data from these kinds of electronic contracts.
  • the automatic electronic contract data extraction system comprises an administrator module, a PDF parser, a pattern recognition engine and a contract data extraction engine.
  • the administrator module provides templates for inputting document patterns and a list of contract data tags for each contract document type.
  • the administrator module also constructs pattern matrices and stores them in a database.
  • the PDF parser converts the contract PDF document into the contract text document with the insertion of formatting bookmarks, such as a new page, paragraph or line.
  • the pattern recognition engine determines a list of contract document types in the electronic contract by comparing and matching the patterns of all known contract document types with the pattern of the contract text document.
  • the contract data extraction engine retrieves the corresponding list of contract data tags and then extracts contract data accordingly for each contract document type on the list.
  • the automatic electronic contract data extraction system has been found to be very accurate, efficient and useful in extracting contract data for data mining.
  • a system and method for automatically extracting contract data from electronic contracts includes an administrator module configured to provide templates for inputting document patterns and a list of contract data tags for each of a plurality of contract document types.
  • a parser is configured to convert an electronic contract document into a contract text document and reformat the contract text document to provide a pattern for the text contract document.
  • a pattern recognition engine is configured to determine a list of contract document types in the electronic contract by comparing and matching patterns of all known contract document types with the pattern of the contract text document.
  • a contract data extraction engine is configured to extract contract data for each contract document type on the list.
  • Another method for automatically extracting contract data from electronic contracts includes providing templates for inputting document patterns and a list of contract data tags for each of a plurality of contract document types; parsing an electronic contract document to convert the electronic contract document into a contract text document; determining a list of contract document types in the electronic contract by comparing and matching patterns of all known contract document types with the pattern of the contract text document; and extracting contract data for each contract document type on the list.
  • FIG. 1 is a block/flow diagram showing an architecture framework of an automatic data extraction system for electronic contract documents in accordance with one illustrative embodiment
  • FIG. 2 is a block/flow diagram showing an administrator module in accordance with an illustrative embodiment
  • FIG. 3 is a block/flow diagram depicting a contract pattern recognition engine in accordance with an illustrative embodiment
  • FIG. 4 is a block/flow diagram depicting a contract data extraction engine in accordance with an illustrative embodiment.
  • an electronic contract can encompass a large number of collateral contract documents in, e.g., a portable document format (PDF). These contract documents may be of different contract document types and converted from different original formats. Data extraction and thus data mining for these kinds of electronic contracts is very difficult. Novel systems and methods will be presented herein to automatically extract contract data from electronic contracts.
  • an automatic electronic contract data extraction system includes an administrator module, a PDF parser, a pattern recognition engine and a contract data extraction engine.
  • the administrator module provides templates for inputting document patterns and a list of contract data tags for each contract document type.
  • the administrator module also constructs pattern matrices and stores them in a database.
  • the PDF parser converts the contract PDF document into the contract text document with the insertion of formatting bookmarks, such as a new page, paragraph or line.
  • the pattern recognition engine determines a list of contract document types in the electronic contract by comparing and matching the patterns of all known contract document types with the pattern of the contract text document.
  • the contract data extraction engine retrieves a corresponding list of contract data tags and then extracts contract data accordingly for each contract document type on the list.
  • Embodiments of the present invention can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment including both hardware and software elements.
  • the present invention is implemented in software, which includes but is not limited to firmware, resident software, microcode, etc.
  • a computer-usable or computer-readable medium can be any apparatus that may include, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.
  • the medium can be an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system (or apparatus or device) or a propagation medium.
  • Examples of a computer-readable medium include a semiconductor or solid state memory, magnetic tape, a removable computer diskette, a random access memory (RAM), a read-only memory (ROM), a rigid magnetic disk and an optical disk. Current examples of optical disks include compact disk-read only memory (CD-ROM), compact disk-read/write (CD-R/W) and DVD.
  • a data processing system suitable for storing and/or executing program code may include at least one processor coupled directly or indirectly to memory elements through a system bus.
  • the memory elements can include local memory employed during actual execution of the program code, bulk storage, and cache memories which provide temporary storage of at least some program code to reduce the number of times code is retrieved from bulk storage during execution.
  • I/O devices including but not limited to keyboards, displays, pointing devices, etc. may be coupled to the system either directly or through intervening I/O controllers.
  • Network adapters may also be coupled to the system to enable the data processing system to become coupled to other data processing systems or remote printers or storage devices through intervening private or public networks.
  • Modems, cable modem and Ethernet cards are just a few of the currently available types of network adapters.
  • An electronic contract 12 may be composed of a number of documents in, e.g., PDF format.
  • the system 10 in accordance with one embodiment includes an administrator module 14 , a (PDF) parser 16 , a contract pattern recognition engine 18 and a contract data extraction engine 20 .
  • the framework 10 also illustrates the interaction between the PDF parser 16 and an electronic contract management system 20 .
  • a Web-based electronic contract management system 20 may be employed as known in the art.
  • This particular management system 20 converts electronic contracts (eContracts) in a format (e.g., PDF) in block 44 .
  • Management system 20 then merges all related documents of an electronic contract into one PDF document in block 46 .
  • Documents may be related based on the parties, dates, contract types or any other criteria.
  • the administrator module 14 provides templates for inputting document patterns in block 30 and contract data tags for each new contract document type in block 36 . Note that the steps in block 14 may be performed for each contract type. Patterns include common terms or entered data fields that can be used to identify a contract by its type. Module 14 also constructs pattern matrices in block 32 and generates a list of contract data tags in block 38 . The pattern matrices and list of contract data tags are stored in a local or remote database in blocks 34 and 40 . Pattern matrices are used to compare the pattern characteristics of all the existing or known contract document types with the contract text document of an electronic contract. Contract data tags include tags for data, metadata or fields of data in a contract.
  • a PDF parser 16 When a user uploads an electronic contract 12 into a Web-based electronic contract management and process system 20 , a PDF parser 16 is used to convert the PDF document into a text document. There are a number of special functions or features in this PDF parser 16 . First, the PDF formatting characters or words originally in the PDF document are removed and replaced with either a blank space or line feed in the text document depending on their locations in the contract document. Second, consecutive blank spaces or line feeds are reduced into a single blank space or line feed, respectively. Third, an extra text line with special coding characters or words indicating a new page or a new paragraph is inserted into the text document whenever the parser finds a new page or a new paragraph. Fourth, other special coding characters or formatting bookmarks are also used to indicate a new line, a change of fonts, an image or a picture, as well as to replace other PDF formats. These steps are employed to clean up the text for the recognition steps later on in the process.
  • the pattern recognition engine 18 uses the pattern matrices of all known contract document types from block 34 to determine, in block 48 , a list of contract document types in the electronic contract by comparing and matching the pattern matrices with the pattern of the contract text document.
  • a contract data extraction engine 50 retrieves the corresponding list of contract data tags from block 40 and then extracts contract data accordingly for each contract document type on the list.
  • the retrieved contract data are stored on either the local or remote database for data mining in block 52 .
  • intrinsic metadata from the retrieved contract data are propagated back to the Web-based electronic contract management and process system 20 to assist the user in filling out the contract input template.
  • contract terms may be stored in the database such that when a use is filling out a template on the system a plurality of previously used or available terms are presented for entry into the contract form.
  • An administrator uses the administrator module 14 to construct document type pattern matrices in block 114 and to generate a list of contract data tags in block 122 for each new contract document type entering into the electronic contract management and process system in block 102 .
  • the administrator first has to determine characteristics of beginning and ending patterns of this new contract document type in block 104 . These characteristics should be special or unique to the pattern of this new contract document type and at the same time different from the pattern characteristics of other contract document types. Usually, one to two characteristics for each of the beginning or ending patterns are enough to distinguish this new contract document type from other existing or known contract document types stored in the system. Each pattern characteristic usually involves two to five lines and two to ten words or terms per line. These pattern characteristics can be a specific contract document type number written either on the first or the last page of the contract document.
  • Characteristics of the beginning pattern are usually the first few lines of descriptions on parties involved in a contract document in the first page.
  • Characteristics of the ending pattern are usually the last few lines of the agreement or signing information of parties involved in a contract document in the last page.
  • a template for the contract may be selected from a database which corresponds to the pattern detected in block 108 . If there is no pattern characteristic which can distinguish a new contract document type from other known contract document types on either the first page or the last page of this new contract document type, pattern characteristics at other pages besides the first and last pages can also be used. Other examples of pattern characteristics are special headers or footers of a contract document, dates, length of the documents, etc.
  • the administrator enters the location for each pattern (block 110 ) on the template for document type patterns.
  • the location may include a specific page number, paragraph number, line number, word number, etc. Instead of an exact number, a range of numbers or within certain numbers can also be used. This is particularly true for the line location where the exact line number can vary and depend on the length of contract document content for different contract documents of the same contract document type.
  • the number of lines and number of words or terms for each line involved in the pattern are also entered.
  • the pattern matrices are constructed in block 114 .
  • the administrator For the contract data, the administrator first has to decide what contract data is needed for extraction for this new contract document type in block 106 . Then, the administrator has to determine the corresponding data tags of these contract data in block 118 . For each contract data and its tag, the module 14 enters locations on the template for contract data tags in block 120 and 112 . Again, the location may include a specific page number, paragraph number, line number, word number, etc. Usually, two to five words or terms are enough to identify a contract data tag. A number of word tokens to be extracted for contract data corresponding to a particular contract data tag are also entered in block 112 . Word tokens extracted do not necessary locate in the same line. The word tokens can be located on several consecutive lines.
  • the administrator module 14 will construct the document type pattern matrices in block 114 and generate a list of contract data tags in block 122 for this new contract document type entering the system.
  • the pattern matrices and contract data tags may be stored in a local database in block 116 or may be transferred to a remote database 126 .
  • An application server 124 may include the remote database 126 which can stored pattern matrices and contract data tags in block 128 .
  • the application server 124 may be accessed via a network 130 such as the Internet or the like.
  • a block/flow diagram illustrates steps performed by a contract pattern recognition engine 18 for a contract text document.
  • the engine 18 has to read and decode those extra text lines with special coding characters or words of bookmarks from the contract text document generated by the PDF parser in block 202 . These extra text lines are originally inserted into the contract text document by the PDF parser 16 ( FIG. 1 ) to indicate a new page or a new paragraph.
  • the engine 18 retrieves the pattern matrices including the beginning and ending patterns of each known or existing contract document type in the system in block 206 .
  • the engine 18 compares the beginning and ending patterns of each known or existing contract document type to each page pattern of this contract text document.
  • the engine 18 records the contract document type and the page number of a particular page in the contract text document if all the characteristics of its beginning pattern matching a particular page in the contract text document.
  • the engine also records the contract document type and the page number of a particular page in the contract text document if all the characteristics of its ending pattern matching a particular page in the contract text document in block 216 .
  • the matching page numbers along with the matching contract document types are sorted for both the beginning and ending patterns in blocks 224 and 222 .
  • the engine can then determine a list of contract document types in the contract text document.
  • the page numbers on the contract text document corresponding to the beginning and ending pages of those contract document types in the list are also determined. This information is stored on a local database for the contract data extraction engine to use in block 218 .
  • Pattern matrices are used to compare the pattern characteristics of all the existing or known contract document types with the contract text document of an electronic contract.
  • a pattern matrix of a known contract document type be q
  • the entries in q are simply the occurrence of a list of terms or words in different lines.
  • the occurrence of terms has to be in a specific order according to a term list, and the same term can appear in the term list more than one time. Occurrence is set to 1 while non-occurrence is set to 0.
  • each pattern characteristic may involve two to five lines and two to ten terms per line. If there are 5 lines and each line includes a list of 10 terms in a specific order, the maximum dimensions of a pattern matrix is 50 ⁇ 5 assuming that no more than one line has the same terms arranged in the same order in the term list.
  • a contract text document is generated by the PDF parser 16 ( FIG. 1 ) in block 318 when an electronic contract including a number of documents in PDF format enters the system.
  • the contract pattern recognition engine 18 determines and constructs a list of contract document types in this contract text document in block 302 . Then, the contract data extraction engine 50 will retrieve this list of contract document types from the local database in block 304 . The extraction engine 50 will also retrieve a list of contract data tags corresponding to each document type on the type list in block 306 .
  • engine 50 For each contract data tag on the tag list (block 308 ), engine 50 will parse its location in terms of page, paragraph, table, line and word token number in block 310 . A range of numbers or within certain numbers can also be used instead of an exact number. The number of terms involved in this contract data tag is also parsed. Moreover, the number of word tokens to be extracted for the contract data corresponding to this particular contract data tag is also parsed.
  • the contract data extraction engine 50 compares the pattern of a contract data tag to the line pattern of each page of this contract text document in block 312 and 314 . Instead of using a pattern matrix for the comparison as described above, engine 50 uses a pattern vector in a similar way for comparison. If a particular line pattern matches the pattern of a contract data tag, contract data corresponding to this particular contract data tag will be extracted according to the parsed number of word tokens to be extracted in block 316 . Word tokens extracted do not necessary have to be located along the same line. The word tokens can be located on several lines.
  • the contract data and intrinsic metadata is sorted according to their tags.
  • the intrinsic metadata is propagated back to the management system 20 ( FIG. 1 ) to assist filling out contract input plates by a user in block 322 .
  • a contract database is built for data mining.
  • a matrix r of the same dimension as matrix q is formed from the contract text document.
  • the entries in r are the occurrence of the same list of terms in different lines.
  • a different matrix r can be formed by shifting the first starting line down a particular page of the contract text document. The number of these different matrices r in this particular page is the number of lines in the page minus the number of lines in the pattern, for example, 5 in this case.
  • the particular contract text document includes this known contract document type. If the pattern matrix q is from the beginning pattern in the first page, then this particular page is the first page of a known contract document type. Similarly, if the pattern matrix q is from the ending pattern in the last page, then this particular page is the last page of a known contract document type. Other pages in the known contract document type can also be identifies in a similar way.
  • An electronic contract that includes a plurality of document types is loaded in the management system 20 ( FIG. 1 ).
  • the contract e.g., Lease Agreement Document
  • the contract is in PDF format and includes two document types, “Term Lease Supplement” and Addendum to Term Lease”, which are of course in PDF format as well.
  • the format of the contract will trickle different format type library's APIs which will be used to extract text from the contract depending on format type.
  • the format type is PDF and the library will include APIs from ADOBETM.
  • a text file will be created. The extraction results in some formatting words being added during the text extraction. With the format type library's APIs, these formatting words are removed and replaced with either a blank space or line feed, etc. as described above. Then, this text file is stored in a text content database ready for a content search engine to carry out any content search.
  • common words such as “by”, “the”, “and”; since these common words do not pay a role in the pattern recognition engine, they are also removed from the text file.
  • a list of common words can be supplied by the administrator by entering them into a template or obtaining them from a dictionary book company, etc.
  • Every contract document of the lease type in this illustration would include the following 11 lines of heading as its pattern characteristics.
  • a pattern matrix may be constructed by setting to one every term entry with a pattern characteristic. Otherwise, set it to 0. For the list of terms above, we get:
  • pattern characteristics are located in the heading of the first page and the ending of the last page. However, they can be located in the body of the first and the last pages. Pattern characteristics have to be entered preferably for the first and last page of any new contract document type. If no pattern characteristics can be found on either the first or last page, then pattern characteristics of other pages can be used provided that the number of pages after the first page or the number of pages before the last page is provided for this page with pattern characteristics.
  • This information is used to identify the first and the last page of such a contract document type if it is inside or presented in the extracted text file.
  • the number of documents and their contract document types can be determined in the extracted text file by searching the known pattern characteristics of all the contract document types in the system by the following steps.
  • pattern characteristics of the first page of each known contract document type are searched for the first page of the extracted text file.
  • a matrix r is formed from the search result. For every term search, set a value in the matrix to 1 if a match is found for the extracted words in the text when compared to a list. Otherwise, set it to 0.
  • the search has to be carried out in the order of pattern characteristics and is preferably a forward search. As a result, whether the first page of this extracted text file is the first page of a particular contract document type or not, will depend on value of P from Equation (1).
  • a template for the administrator to enter a list of metadata tags and their page location that the administrator wants to add as extra to the list of metadata from a particular contract document type is also provided.
  • a list of contract document types on the extracted text file has been determined, a list of metadata tags corresponding to each contract document type is used to build the metadata database for this particular electronic contract.
  • the extracted text file still includes a large number of common terms, such as the standard legal terms, which do not pay any role in semantic search, these common terms are eliminated from the file.
  • a number of templates are provided for the administrator to enter these legal terms.
  • the file is used to build a semantic indexing database for semantic search using Laten Semantic Indexing (LSI) techniques.
  • LSI Laten Semantic Indexing
  • An administrator can enter a list of key terms on templates provided by the system. These key terms can be used to build a correlation database using Correlated Filamentous Propagation (CFP) methods for electronic contract relationship search as well as other data mining techniques.
  • CFP Correlated Filamentous Propagation
  • the contract data extraction system described herein provides an automated and efficient way to extract contract data for electronic contracts composed of a number of documents in, e.g., PDF format.
  • the system has solved a number of difficult problems involving the comparison of patterns or pattern recognition.
  • the pattern recognition technique described is based on the calculation of the dot product of pattern matrices.
  • the system needs an administrator to input a distinct pattern of a new contract document type. This is a rather time consuming processing step for the user. However, this step is a one time setup for any new contract document type coming into the system.
  • the number of different contract document types can vary. New contract document types can be input into the system at any time. 2) A list of contract data to be extracted can also be modified by modifying the corresponding list of data tags. 3) Intrinsic contract metadata extracted from the electronic contract document can be propagated back in the electronic contract management system right after the user uploads the electronic contract document into the system. The extracted intrinsic contract metadata can be displayed in the graphical user interface of the user module to assist the user in entering contract information of the electronic contract that the user just submitted into the system. It has been found to save users a lot of time from manually entering contract data from scratch. The system also avoids wrong contract metadata mistakenly entered by the user. 4) The present system also provides an efficient way to generate contract text content database for key word or term search. Other advantages and benefits are also provided in accordance with the present principles.

Abstract

A system and method for automatically extracting contract data from electronic contracts includes an administrator module configured to provide templates for inputting document patterns and a list of contract data tags for each of a plurality of contract document types. A parser is configured to convert an electronic contract document into a contract text document and reformat the contract text document to provide a pattern for the text contract document. A pattern recognition engine is configured to determine a list of contract document types in the electronic contract by comparing and matching patterns of all known contract document types with the pattern of the contract text document. A contract data extraction engine is configured to extract contract data for each contract document type on the list.

Description

    BACKGROUND
  • 1. Technical Field
  • The present invention relates to data extraction from documents and more particularly to systems and methods which extract contract data automatically and efficiently from an electronic contract composed of a number of documents in a given format.
  • 2. Description of the Related Art
  • Much business between enterprises is conducted under contract. Contracts constitute the binding relationship between a company and its customers or suppliers. Everyday, many contracts are created, executed and managed via paper-based manual processes in large enterprises. Automation of the contract lifecycle presents a substantial value creation opportunity for enterprises. This value stems from improved productivity and security, effectively aggregated contract information, accelerated contract lifecycle processes, reduced contractual errors and risk, enabled revenue forecast and profit optimization, as well as better compliance enforcement.
  • With the advent of Internet technology and electronic commerce, there are growing research activities and implementation efforts on electronic contracts. Currently, the International Association of Contract and Commercial Managers have listed twenty commercially available software products for electronic contract management. Most of the research activities reported is focused on electronic contract creation or representation language, negotiation, management, collaboration, execution, fulfillment and enforcement, performance, digital signatures and data mining. However, none of these aspects has provided an automatic electronic data extraction solution to enable data mining for revenue forecast and profit optimization.
  • A single electronic contract can encompass a large number of collateral documents including master and customer agreements, supplements, addenda and the like. These various documents are of different contract document types. There can be over a hundred different basic types of contract documents in a large company. A few examples of these contract document types are as follows: “Master Agreement”, “Customer Agreement”, “Term Lease Supplement”, “Addendum to Term Lease Supplement”, “Statement of Work for Services”, “Change Authorization for Services”, etc. Moreover, they can also be in different file formats, such as PDF, XML, Microsoft Word™, Lotus WordPro™.
  • An electronic contract management system can be used to automatically convert all these contract documents of different types into PDF format and then merge them together to form a single electronic contract PDF document. However, data extraction and mining on this kind of electronic contract is still very difficult if not impossible. To do this, a user should find out how many contract documents are in an electronic contract composed of a number of contract documents, and then determine their contract document types. Next, what contract data to extract should be decided and from which contract document. The user would further need to find out where on the contract document the contract data is located, such as page and line numbers. There are many more tasks to be overcome before one can implement a data extraction and mining on this kind for electronic contracts.
  • SUMMARY
  • An electronic contract can encompass a large number of collateral contract documents in, e.g., PDF format. These contract documents are of different contract document types and converted from different original formats. Data extraction and thus data mining for this kind of electronic contracts is very difficult. A novel system and method are presented to automatically extract contract data from these kinds of electronic contracts. The automatic electronic contract data extraction system comprises an administrator module, a PDF parser, a pattern recognition engine and a contract data extraction engine. The administrator module provides templates for inputting document patterns and a list of contract data tags for each contract document type. The administrator module also constructs pattern matrices and stores them in a database.
  • The PDF parser converts the contract PDF document into the contract text document with the insertion of formatting bookmarks, such as a new page, paragraph or line. The pattern recognition engine determines a list of contract document types in the electronic contract by comparing and matching the patterns of all known contract document types with the pattern of the contract text document. The contract data extraction engine retrieves the corresponding list of contract data tags and then extracts contract data accordingly for each contract document type on the list. The automatic electronic contract data extraction system has been found to be very accurate, efficient and useful in extracting contract data for data mining.
  • A system and method for automatically extracting contract data from electronic contracts includes an administrator module configured to provide templates for inputting document patterns and a list of contract data tags for each of a plurality of contract document types. A parser is configured to convert an electronic contract document into a contract text document and reformat the contract text document to provide a pattern for the text contract document. A pattern recognition engine is configured to determine a list of contract document types in the electronic contract by comparing and matching patterns of all known contract document types with the pattern of the contract text document. A contract data extraction engine is configured to extract contract data for each contract document type on the list.
  • Another method for automatically extracting contract data from electronic contracts includes providing templates for inputting document patterns and a list of contract data tags for each of a plurality of contract document types; parsing an electronic contract document to convert the electronic contract document into a contract text document; determining a list of contract document types in the electronic contract by comparing and matching patterns of all known contract document types with the pattern of the contract text document; and extracting contract data for each contract document type on the list.
  • These and other objects, features and advantages will become apparent from the following detailed description of illustrative embodiments thereof, which is to be read in connection with the accompanying drawings.
  • BRIEF DESCRIPTION OF DRAWINGS
  • The disclosure will provide details in the following description of preferred embodiments with reference to the following figures wherein:
  • FIG. 1 is a block/flow diagram showing an architecture framework of an automatic data extraction system for electronic contract documents in accordance with one illustrative embodiment;
  • FIG. 2 is a block/flow diagram showing an administrator module in accordance with an illustrative embodiment;
  • FIG. 3 is a block/flow diagram depicting a contract pattern recognition engine in accordance with an illustrative embodiment; and
  • FIG. 4 is a block/flow diagram depicting a contract data extraction engine in accordance with an illustrative embodiment.
  • DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS
  • An electronic contract can encompass a large number of collateral contract documents in, e.g., a portable document format (PDF). These contract documents may be of different contract document types and converted from different original formats. Data extraction and thus data mining for these kinds of electronic contracts is very difficult. Novel systems and methods will be presented herein to automatically extract contract data from electronic contracts. In one embodiment, an automatic electronic contract data extraction system includes an administrator module, a PDF parser, a pattern recognition engine and a contract data extraction engine.
  • The administrator module provides templates for inputting document patterns and a list of contract data tags for each contract document type. The administrator module also constructs pattern matrices and stores them in a database. The PDF parser converts the contract PDF document into the contract text document with the insertion of formatting bookmarks, such as a new page, paragraph or line.
  • The pattern recognition engine determines a list of contract document types in the electronic contract by comparing and matching the patterns of all known contract document types with the pattern of the contract text document. The contract data extraction engine retrieves a corresponding list of contract data tags and then extracts contract data accordingly for each contract document type on the list.
  • Embodiments of the present invention can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment including both hardware and software elements. In a preferred embodiment, the present invention is implemented in software, which includes but is not limited to firmware, resident software, microcode, etc.
  • Furthermore, the invention can take the form of a computer program product accessible from a computer-usable or computer-readable medium providing program code for use by or in connection with a computer or any instruction execution system. For the purposes of this description, a computer-usable or computer readable medium can be any apparatus that may include, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. The medium can be an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system (or apparatus or device) or a propagation medium. Examples of a computer-readable medium include a semiconductor or solid state memory, magnetic tape, a removable computer diskette, a random access memory (RAM), a read-only memory (ROM), a rigid magnetic disk and an optical disk. Current examples of optical disks include compact disk-read only memory (CD-ROM), compact disk-read/write (CD-R/W) and DVD.
  • A data processing system suitable for storing and/or executing program code may include at least one processor coupled directly or indirectly to memory elements through a system bus. The memory elements can include local memory employed during actual execution of the program code, bulk storage, and cache memories which provide temporary storage of at least some program code to reduce the number of times code is retrieved from bulk storage during execution. Input/output or I/O devices (including but not limited to keyboards, displays, pointing devices, etc.) may be coupled to the system either directly or through intervening I/O controllers.
  • Network adapters may also be coupled to the system to enable the data processing system to become coupled to other data processing systems or remote printers or storage devices through intervening private or public networks. Modems, cable modem and Ethernet cards are just a few of the currently available types of network adapters.
  • Referring now to the drawings in which like numerals represent the same or similar elements and initially to FIG. 1, an architectural framework for an automatic contract data extraction system 10 is illustratively shown in accordance with one embodiment. An electronic contract 12 may be composed of a number of documents in, e.g., PDF format. The system 10 in accordance with one embodiment includes an administrator module 14, a (PDF) parser 16, a contract pattern recognition engine 18 and a contract data extraction engine 20. The framework 10 also illustrates the interaction between the PDF parser 16 and an electronic contract management system 20. In one example, a Web-based electronic contract management system 20 may be employed as known in the art. This particular management system 20 converts electronic contracts (eContracts) in a format (e.g., PDF) in block 44. Management system 20 then merges all related documents of an electronic contract into one PDF document in block 46. Documents may be related based on the parties, dates, contract types or any other criteria.
  • The administrator module 14 provides templates for inputting document patterns in block 30 and contract data tags for each new contract document type in block 36. Note that the steps in block 14 may be performed for each contract type. Patterns include common terms or entered data fields that can be used to identify a contract by its type. Module 14 also constructs pattern matrices in block 32 and generates a list of contract data tags in block 38. The pattern matrices and list of contract data tags are stored in a local or remote database in blocks 34 and 40. Pattern matrices are used to compare the pattern characteristics of all the existing or known contract document types with the contract text document of an electronic contract. Contract data tags include tags for data, metadata or fields of data in a contract.
  • When a user uploads an electronic contract 12 into a Web-based electronic contract management and process system 20, a PDF parser 16 is used to convert the PDF document into a text document. There are a number of special functions or features in this PDF parser 16. First, the PDF formatting characters or words originally in the PDF document are removed and replaced with either a blank space or line feed in the text document depending on their locations in the contract document. Second, consecutive blank spaces or line feeds are reduced into a single blank space or line feed, respectively. Third, an extra text line with special coding characters or words indicating a new page or a new paragraph is inserted into the text document whenever the parser finds a new page or a new paragraph. Fourth, other special coding characters or formatting bookmarks are also used to indicate a new line, a change of fonts, an image or a picture, as well as to replace other PDF formats. These steps are employed to clean up the text for the recognition steps later on in the process.
  • These special coding characters or words are readable and decoded by both the contract pattern recognition engine 18 and a contract data extraction engine 50. Then, the pattern recognition engine 18 uses the pattern matrices of all known contract document types from block 34 to determine, in block 48, a list of contract document types in the electronic contract by comparing and matching the pattern matrices with the pattern of the contract text document.
  • A contract data extraction engine 50 retrieves the corresponding list of contract data tags from block 40 and then extracts contract data accordingly for each contract document type on the list. The retrieved contract data are stored on either the local or remote database for data mining in block 52. In block 54, intrinsic metadata from the retrieved contract data are propagated back to the Web-based electronic contract management and process system 20 to assist the user in filling out the contract input template.
  • In other words, contract terms may be stored in the database such that when a use is filling out a template on the system a plurality of previously used or available terms are presented for entry into the contract form.
  • Referring to FIG. 2, illustrative steps are depicted for the administration module 14. An administrator uses the administrator module 14 to construct document type pattern matrices in block 114 and to generate a list of contract data tags in block 122 for each new contract document type entering into the electronic contract management and process system in block 102.
  • For document type patterns, the administrator first has to determine characteristics of beginning and ending patterns of this new contract document type in block 104. These characteristics should be special or unique to the pattern of this new contract document type and at the same time different from the pattern characteristics of other contract document types. Usually, one to two characteristics for each of the beginning or ending patterns are enough to distinguish this new contract document type from other existing or known contract document types stored in the system. Each pattern characteristic usually involves two to five lines and two to ten words or terms per line. These pattern characteristics can be a specific contract document type number written either on the first or the last page of the contract document.
  • Characteristics of the beginning pattern are usually the first few lines of descriptions on parties involved in a contract document in the first page. Characteristics of the ending pattern are usually the last few lines of the agreement or signing information of parties involved in a contract document in the last page. Based on these patterns, a template for the contract may be selected from a database which corresponds to the pattern detected in block 108. If there is no pattern characteristic which can distinguish a new contract document type from other known contract document types on either the first page or the last page of this new contract document type, pattern characteristics at other pages besides the first and last pages can also be used. Other examples of pattern characteristics are special headers or footers of a contract document, dates, length of the documents, etc.
  • Then, in block 112, the administrator enters the location for each pattern (block 110) on the template for document type patterns. The location may include a specific page number, paragraph number, line number, word number, etc. Instead of an exact number, a range of numbers or within certain numbers can also be used. This is particularly true for the line location where the exact line number can vary and depend on the length of contract document content for different contract documents of the same contract document type. The number of lines and number of words or terms for each line involved in the pattern are also entered. The pattern matrices are constructed in block 114.
  • For the contract data, the administrator first has to decide what contract data is needed for extraction for this new contract document type in block 106. Then, the administrator has to determine the corresponding data tags of these contract data in block 118. For each contract data and its tag, the module 14 enters locations on the template for contract data tags in block 120 and 112. Again, the location may include a specific page number, paragraph number, line number, word number, etc. Usually, two to five words or terms are enough to identify a contract data tag. A number of word tokens to be extracted for contract data corresponding to a particular contract data tag are also entered in block 112. Word tokens extracted do not necessary locate in the same line. The word tokens can be located on several consecutive lines.
  • The administrator module 14 will construct the document type pattern matrices in block 114 and generate a list of contract data tags in block 122 for this new contract document type entering the system. The pattern matrices and contract data tags may be stored in a local database in block 116 or may be transferred to a remote database 126.
  • An application server 124 may include the remote database 126 which can stored pattern matrices and contract data tags in block 128. The application server 124 may be accessed via a network 130 such as the Internet or the like.
  • Referring to FIG. 3, a block/flow diagram illustrates steps performed by a contract pattern recognition engine 18 for a contract text document. The engine 18 has to read and decode those extra text lines with special coding characters or words of bookmarks from the contract text document generated by the PDF parser in block 202. These extra text lines are originally inserted into the contract text document by the PDF parser 16 (FIG. 1) to indicate a new page or a new paragraph. For each contract document type (204), the engine 18 retrieves the pattern matrices including the beginning and ending patterns of each known or existing contract document type in the system in block 206. In blocks 208, 210 and 214, the engine 18 compares the beginning and ending patterns of each known or existing contract document type to each page pattern of this contract text document. In block 212, the engine 18 records the contract document type and the page number of a particular page in the contract text document if all the characteristics of its beginning pattern matching a particular page in the contract text document. The engine also records the contract document type and the page number of a particular page in the contract text document if all the characteristics of its ending pattern matching a particular page in the contract text document in block 216.
  • The matching page numbers along with the matching contract document types are sorted for both the beginning and ending patterns in blocks 224 and 222. In block 220, based on this information, the engine can then determine a list of contract document types in the contract text document. The page numbers on the contract text document corresponding to the beginning and ending pages of those contract document types in the list are also determined. This information is stored on a local database for the contract data extraction engine to use in block 218.
  • Pattern matrices are used to compare the pattern characteristics of all the existing or known contract document types with the contract text document of an electronic contract. Let a pattern matrix of a known contract document type be q, the entries in q are simply the occurrence of a list of terms or words in different lines. The occurrence of terms has to be in a specific order according to a term list, and the same term can appear in the term list more than one time. Occurrence is set to 1 while non-occurrence is set to 0. As mentioned above, each pattern characteristic may involve two to five lines and two to ten terms per line. If there are 5 lines and each line includes a list of 10 terms in a specific order, the maximum dimensions of a pattern matrix is 50×5 assuming that no more than one line has the same terms arranged in the same order in the term list.
  • Referring to FIG. 4, a block/flow diagram is shown for a contract data extraction engine 50 in accordance with an illustrative embodiment. A contract text document is generated by the PDF parser 16 (FIG. 1) in block 318 when an electronic contract including a number of documents in PDF format enters the system. The contract pattern recognition engine 18 (FIG. 1) determines and constructs a list of contract document types in this contract text document in block 302. Then, the contract data extraction engine 50 will retrieve this list of contract document types from the local database in block 304. The extraction engine 50 will also retrieve a list of contract data tags corresponding to each document type on the type list in block 306. For each contract data tag on the tag list (block 308), engine 50 will parse its location in terms of page, paragraph, table, line and word token number in block 310. A range of numbers or within certain numbers can also be used instead of an exact number. The number of terms involved in this contract data tag is also parsed. Moreover, the number of word tokens to be extracted for the contract data corresponding to this particular contract data tag is also parsed.
  • Similar to the contract pattern recognition engine, the contract data extraction engine 50 compares the pattern of a contract data tag to the line pattern of each page of this contract text document in block 312 and 314. Instead of using a pattern matrix for the comparison as described above, engine 50 uses a pattern vector in a similar way for comparison. If a particular line pattern matches the pattern of a contract data tag, contract data corresponding to this particular contract data tag will be extracted according to the parsed number of word tokens to be extracted in block 316. Word tokens extracted do not necessary have to be located along the same line. The word tokens can be located on several lines.
  • In block 324 the contract data and intrinsic metadata is sorted according to their tags. The intrinsic metadata is propagated back to the management system 20 (FIG. 1) to assist filling out contract input plates by a user in block 322. In block 320, a contract database is built for data mining.
  • Turning now to an illustrative example, which will illustratively describe some of the present principles, a matrix r of the same dimension as matrix q is formed from the contract text document. Similarly, the entries in r are the occurrence of the same list of terms in different lines. A different matrix r can be formed by shifting the first starting line down a particular page of the contract text document. The number of these different matrices r in this particular page is the number of lines in the page minus the number of lines in the pattern, for example, 5 in this case.
  • Let P be the dot product of matrix q and r divided by the dot product of matrix q and q according to Equation (1)

  • P=(q·r)/(q·q)  (1)
  • As a result, if P equals 1 for any matrix r in a particular page, then this particular page matches the pattern characteristics of a known contract document type. Thus, the particular contract text document includes this known contract document type. If the pattern matrix q is from the beginning pattern in the first page, then this particular page is the first page of a known contract document type. Similarly, if the pattern matrix q is from the ending pattern in the last page, then this particular page is the last page of a known contract document type. Other pages in the known contract document type can also be identifies in a similar way.
  • An electronic contract that includes a plurality of document types is loaded in the management system 20 (FIG. 1). The contract (e.g., Lease Agreement Document) is in PDF format and includes two document types, “Term Lease Supplement” and Addendum to Term Lease”, which are of course in PDF format as well. After loaded in to the system, the format of the contract will trickle different format type library's APIs which will be used to extract text from the contract depending on format type. In this case, the format type is PDF and the library will include APIs from ADOBE™. A text file will be created. The extraction results in some formatting words being added during the text extraction. With the format type library's APIs, these formatting words are removed and replaced with either a blank space or line feed, etc. as described above. Then, this text file is stored in a text content database ready for a content search engine to carry out any content search.
  • There are a number of common words, such as “by”, “the”, “and”; since these common words do not pay a role in the pattern recognition engine, they are also removed from the text file. A list of common words can be supplied by the administrator by entering them into a template or obtaining them from a dictionary book company, etc.
  • From this particular “Term Lease Supplement” contract document type, every contract document of the lease type in this illustration would include the following 11 lines of heading as its pattern characteristics.
  • Term Lease Supplement Date Prepared: Page Customer Supplement Customer Address Installed Location Location Address Term Lease Master Agreement Associated Supplement Summary Supplement Amendment Addendum Customer Reference: Quote Letter
  • There are usually no more than ten terms in a line. Four to five lines will be generally be enough to form the pattern characteristics. In this case, a matrix of 4×11 is formed as the pattern characteristics for the heading for this particular contract document type. Let this pattern recognition matrix be “q”. A pattern matrix may be constructed by setting to one every term entry with a pattern characteristic. Otherwise, set it to 0. For the list of terms above, we get: Thus,
  • q = 1 1 1 0 1 1 1 0 1 1 0 0 1 1 1 1 1 1 0 0 1 1 1 1 1 1 0 0 1 1 0 0 1 0 0 0 1 0 0 0 1 1 1 1
  • This is a one time setup requirement for any new contract document type coming into the system. A template for entering pattern characteristics is provided for the administrator. Usually, the pattern characteristics are located in the heading of the first page and the ending of the last page. However, they can be located in the body of the first and the last pages. Pattern characteristics have to be entered preferably for the first and last page of any new contract document type. If no pattern characteristics can be found on either the first or last page, then pattern characteristics of other pages can be used provided that the number of pages after the first page or the number of pages before the last page is provided for this page with pattern characteristics.
  • This information is used to identify the first and the last page of such a contract document type if it is inside or presented in the extracted text file. Thus, the number of documents and their contract document types can be determined in the extracted text file by searching the known pattern characteristics of all the contract document types in the system by the following steps.
  • First, pattern characteristics of the first page of each known contract document type are searched for the first page of the extracted text file. A matrix r is formed from the search result. For every term search, set a value in the matrix to 1 if a match is found for the extracted words in the text when compared to a list. Otherwise, set it to 0. The search has to be carried out in the order of pattern characteristics and is preferably a forward search. As a result, whether the first page of this extracted text file is the first page of a particular contract document type or not, will depend on value of P from Equation (1).
  • If P equals 1, this result will indicate that the first page of the file indeed is the first page of this particular contract document type. Otherwise, it is not. Similarly, the end page will be located. In addition, any other specific page of a particular contract document type can be located in this way if metadata is extracted on that specific page. A particular line item can also be located using this method if we need to find certain value such as total amount for a particular line item.
  • A template for the administrator to enter a list of metadata tags and their page location that the administrator wants to add as extra to the list of metadata from a particular contract document type is also provided. After a list of contract document types on the extracted text file has been determined, a list of metadata tags corresponding to each contract document type is used to build the metadata database for this particular electronic contract. Once the system has completed building the metadata database, it will trickle a metadata extraction engine to execute, extract metadata and propagate the metadata into the graphical user interfaces of a user module. This will save the user's time in entering these metadata and avoid any mistakes in entering the wrong metadata.
  • Furthermore, since the extracted text file still includes a large number of common terms, such as the standard legal terms, which do not pay any role in semantic search, these common terms are eliminated from the file. A number of templates are provided for the administrator to enter these legal terms. Then, the file is used to build a semantic indexing database for semantic search using Laten Semantic Indexing (LSI) techniques. An administrator can enter a list of key terms on templates provided by the system. These key terms can be used to build a correlation database using Correlated Filamentous Propagation (CFP) methods for electronic contract relationship search as well as other data mining techniques.
  • Implementation: The features and functions of the automatic contract data extraction system described herein have been implemented and integrated with an electronic management and process system at IBM™. The ADOBE™ PDF library application programming interfaces (APIs) were used to parse PDF documents into a text document in the PDF parser of this system. However, other PDF parsers can also be used.
  • The contract data extraction system described herein provides an automated and efficient way to extract contract data for electronic contracts composed of a number of documents in, e.g., PDF format. The system has solved a number of difficult problems involving the comparison of patterns or pattern recognition. The pattern recognition technique described is based on the calculation of the dot product of pattern matrices. Thus, the system needs an administrator to input a distinct pattern of a new contract document type. This is a rather time consuming processing step for the user. However, this step is a one time setup for any new contract document type coming into the system.
  • There are many advantages of this system. Some of these include: 1) The number of different contract document types can vary. New contract document types can be input into the system at any time. 2) A list of contract data to be extracted can also be modified by modifying the corresponding list of data tags. 3) Intrinsic contract metadata extracted from the electronic contract document can be propagated back in the electronic contract management system right after the user uploads the electronic contract document into the system. The extracted intrinsic contract metadata can be displayed in the graphical user interface of the user module to assist the user in entering contract information of the electronic contract that the user just submitted into the system. It has been found to save users a lot of time from manually entering contract data from scratch. The system also avoids wrong contract metadata mistakenly entered by the user. 4) The present system also provides an efficient way to generate contract text content database for key word or term search. Other advantages and benefits are also provided in accordance with the present principles.
  • Having described preferred embodiments for systems and methods to extract data automatically from a composite electronic document (which are intended to be illustrative and not limiting), it is noted that modifications and variations can be made by persons skilled in the art in light of the above teachings. It is therefore to be understood that changes may be made in the particular embodiments disclosed which are within the scope and spirit of the invention as outlined by the appended claims. Having thus described aspects of the invention, with the details and particularity required by the patent laws, what is claimed and desired protected by Letters Patent is set forth in the appended claims.

Claims (20)

1. A system for automatically extracting contract data from electronic contracts, comprising:
an administrator module configured to provide templates for inputting document patterns and a list of contract data tags for each of a plurality of contract document types;
a parser configured to convert an electronic contract document into a contract text document and reformat the contract text document to provide a pattern for the text contract document;
a pattern recognition engine configured to determine a list of contract document types in the electronic contract by comparing and matching patterns of all known contract document types with the pattern of the text contract document; and
a contract data extraction engine configured to extract contract data for each contract document type on the list.
2. The system as recited in claim 1, wherein the administrator module is configured to construct pattern matrices to provide the templates for inputting document patterns.
3. The system as recited in claim 1, wherein the administrator module stores pattern matrices and contract data tags for comparison with input electronic contracts.
4. The system as recited in claim 3, wherein the parser formats the contract text document to eliminate less relevant information and to permit comparison with the pattern matrices and contract data tags.
5. The system as recited in claim 4, wherein the parser formats the contract text document by inserting formatting bookmarks.
6. The system as recited in claim 1, wherein the administrator module is configured to retrieve a list of contract data tags and extracts contract data accordingly for each contract document type on the list.
7. The system as recited in claim 1, wherein the administrator module is configured to provide beginning and ending pattern matrices for each known document type.
8. The system as recited in claim 1, wherein the administrator module is configured to determine a list of document types in the contract text document by comparing and matching pattern matrices using a dot product calculation.
9. The system as recited in claim 1, wherein the administrator module is configured to determine to a list of data tags corresponding to a document type by comparing and matching line pattern vectors.
10. A method for automatically extracting contract data from electronic contracts, comprising:
providing templates for inputting document patterns and a list of contract data tags for each of a plurality of contract document types;
parsing an electronic contract document to convert the electronic contract document into a contract text document;
determining a list of contract document types in the electronic contract by comparing and matching patterns of all known contract document types with the pattern of the contract text document; and
extracting contract data for each contract document type on the list.
11. The method as recited in claim 10, wherein providing includes constructing pattern matrices to provide the templates for inputting document patterns.
12. The method as recited in claim 10, further comprising storing pattern matrices and contract data tags for comparison with input electronic contracts.
13. The method as recited in claim 12, wherein parsing includes and reformatting the contract text document to eliminate less relevant information to permit comparison with the pattern matrices and contract data tags.
14. The method as recited in claim 13, wherein reformatting includes reformatting the contract text document by inserting formatting bookmarks.
15. The method as recited in claim 10, further comprising retrieving a list of contract data tags and extracting contract data accordingly for each contract document type on the list.
16. The system as recited in claim 10, wherein providing includes providing beginning and ending pattern matrices for each known document type.
17. The method as recited in claim 10, wherein determining a list of document types includes comparing and matching pattern matrices using a dot product calculation.
18. The method as recited in claim 10, wherein determining includes determining a list of data tags corresponding to a document type by comparing and matching line pattern vectors.
19. The method as recited in claim 10, further comprising searching document patterns and contract data tags for each of a plurality of contract document types in a database.
20. A computer program product for automatically extracting contract data from electronic contracts comprising a computer useable medium including a computer readable program, wherein the computer readable program when executed on a computer causes the computer to perform:
providing templates for inputting document patterns and a list of contract data tags for each of a plurality of contract document types;
parsing an electronic contract document to convert the electronic contract document into a contract text document;
determining a list of contract document types in the electronic contract by comparing and matching patterns of all known contract document types with the pattern of the contract text document; and
extracting contract data for each contract document type on the list.
US11/472,868 2006-06-22 2006-06-22 Systems and methods to extract data automatically from a composite electronic document Abandoned US20070300295A1 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
US11/472,868 US20070300295A1 (en) 2006-06-22 2006-06-22 Systems and methods to extract data automatically from a composite electronic document
US12/132,845 US8140468B2 (en) 2006-06-22 2008-06-04 Systems and methods to extract data automatically from a composite electronic document

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
US11/472,868 US20070300295A1 (en) 2006-06-22 2006-06-22 Systems and methods to extract data automatically from a composite electronic document

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US12/132,845 Continuation US8140468B2 (en) 2006-06-22 2008-06-04 Systems and methods to extract data automatically from a composite electronic document

Publications (1)

Publication Number Publication Date
US20070300295A1 true US20070300295A1 (en) 2007-12-27

Family

ID=38874940

Family Applications (2)

Application Number Title Priority Date Filing Date
US11/472,868 Abandoned US20070300295A1 (en) 2006-06-22 2006-06-22 Systems and methods to extract data automatically from a composite electronic document
US12/132,845 Expired - Fee Related US8140468B2 (en) 2006-06-22 2008-06-04 Systems and methods to extract data automatically from a composite electronic document

Family Applications After (1)

Application Number Title Priority Date Filing Date
US12/132,845 Expired - Fee Related US8140468B2 (en) 2006-06-22 2008-06-04 Systems and methods to extract data automatically from a composite electronic document

Country Status (1)

Country Link
US (2) US20070300295A1 (en)

Cited By (20)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080033903A1 (en) * 2006-08-04 2008-02-07 Andrew Carol Methods and apparatuses for using location information
US20090300054A1 (en) * 2008-05-29 2009-12-03 Kathleen Fisher System for inferring data structures
US20100010981A1 (en) * 2008-07-08 2010-01-14 International Business Machines Corporation Technique for enhancing a set of website bookmarks by finding related bookmarks based on a latent similarity metric
US20130111326A1 (en) * 2011-10-26 2013-05-02 Kimber Lockhart Enhanced multimedia content preview rendering in a cloud content management system
WO2014146032A2 (en) * 2013-03-15 2014-09-18 Suarez Sergio David Jr System for method for data sweeping using keywords
US9501495B2 (en) 2010-04-22 2016-11-22 Apple Inc. Location metadata in a media file
US20180239959A1 (en) * 2017-02-22 2018-08-23 Anduin Transactions, Inc. Electronic data parsing and interactive user interfaces for data processing
WO2018156781A1 (en) * 2017-02-22 2018-08-30 Anduin Transactions, Inc. Compact presentation of automatically summarized information according to rule-based graphically represented information
US10311140B1 (en) * 2018-10-25 2019-06-04 BlackBoiler, LLC Systems, methods, and computer program products for a clause library
CN110096626A (en) * 2019-03-18 2019-08-06 平安普惠企业管理有限公司 Processing method, device, equipment and the storage medium of contract text data
CN110943828A (en) * 2019-11-05 2020-03-31 武汉理工大学 Secret number operation conversion method and system
US10713436B2 (en) 2018-03-30 2020-07-14 BlackBoiler, LLC Method and system for suggesting revisions to an electronic document
US10824797B2 (en) 2015-08-03 2020-11-03 Blackboiler, Inc. Method and system for suggesting revisions to an electronic document
CN112529743A (en) * 2020-12-18 2021-03-19 平安银行股份有限公司 Contract element extraction method, contract element extraction device, electronic equipment and medium
US20210295261A1 (en) * 2020-03-20 2021-09-23 Codexo Generating actionable information from documents
US11232481B2 (en) 2012-01-30 2022-01-25 Box, Inc. Extended applications of multimedia content previews in the cloud-based content management system
CN114973290A (en) * 2022-07-26 2022-08-30 简单汇信息科技(广州)有限公司 Intelligent order examination method and system based on OCR engine
CN115017550A (en) * 2022-06-02 2022-09-06 湖南长银五八消费金融股份有限公司 Electronic contract data processing method and device, computer equipment and medium
US11681864B2 (en) 2021-01-04 2023-06-20 Blackboiler, Inc. Editing parameters
US20230394221A1 (en) * 2022-06-06 2023-12-07 Microsoft Technology Licensing, Llc Converting a portable document format to a latex format

Families Citing this family (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7860887B2 (en) 2007-02-20 2010-12-28 The Invention Science Fund I, Llc Cross-media storage coordination
US9008116B2 (en) * 2007-02-20 2015-04-14 The Invention Science Fund I, Llc Cross-media communication coordination
US8650221B2 (en) * 2007-09-10 2014-02-11 International Business Machines Corporation Systems and methods to associate invoice data with a corresponding original invoice copy in a stack of invoices
US20170147577A9 (en) * 2009-09-30 2017-05-25 Gennady LAPIR Method and system for extraction
WO2012048234A2 (en) 2010-10-07 2012-04-12 Morgan Stanley System and method for risk monitoring of rated legal entities
CN104680276A (en) * 2013-11-29 2015-06-03 上海新世界信息产业有限公司 Method and system for managing negotiation and signature of electronic contract online
CN104680277A (en) * 2013-11-29 2015-06-03 上海新世界信息产业有限公司 Method and system for negotiating and signing electronic contract online
US10095986B2 (en) 2014-05-14 2018-10-09 Pegasus Transtech Llc System and method of electronically classifying transportation documents
US20160306883A1 (en) * 2015-04-20 2016-10-20 Tobias Weller Business Intelligence Computing System with Offline Usage
US10713431B2 (en) 2015-12-29 2020-07-14 Accenture Global Solutions Limited Digital document processing based on document source or document type
US10303938B2 (en) * 2016-12-29 2019-05-28 Factset Research Systems Inc Identifying a structure presented in portable document format (PDF)
CN107480288A (en) * 2017-08-24 2017-12-15 天津市深大天星科技发展有限公司 Measure of managing contract and system
EP3511834B1 (en) * 2018-01-10 2020-10-21 Tata Consultancy Services Limited System and method for tool chain data capture through parser for empirical data analysis
US11308083B2 (en) 2019-04-19 2022-04-19 International Business Machines Corporation Automatic transformation of complex tables in documents into computer understandable structured format and managing dependencies
US11194798B2 (en) 2019-04-19 2021-12-07 International Business Machines Corporation Automatic transformation of complex tables in documents into computer understandable structured format with mapped dependencies and providing schema-less query support for searching table data
US11194797B2 (en) 2019-04-19 2021-12-07 International Business Machines Corporation Automatic transformation of complex tables in documents into computer understandable structured format and providing schema-less query support data extraction
US20220374791A1 (en) * 2021-05-19 2022-11-24 Kpmg Llp System and method for implementing a commercial leakage platform
US11709813B2 (en) * 2021-06-10 2023-07-25 Jpmorgan Chase Bank, N.A. System and method for implementing a contract data management module

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5991709A (en) * 1994-07-08 1999-11-23 Schoen; Neil Charles Document automated classification/declassification system
US6202060B1 (en) * 1996-10-29 2001-03-13 Bao Q. Tran Data management system
US6263335B1 (en) * 1996-02-09 2001-07-17 Textwise Llc Information extraction system and method using concept-relation-concept (CRC) triples
US20020049705A1 (en) * 2000-04-19 2002-04-25 E-Base Ltd. Method for creating content oriented databases and content files
US6546133B1 (en) * 1999-09-08 2003-04-08 Ge Capital Commercial Finance, Inc. Methods and apparatus for print scraping
US6629097B1 (en) * 1999-04-28 2003-09-30 Douglas K. Keith Displaying implicit associations among items in loosely-structured data sets
US6775665B1 (en) * 1999-09-30 2004-08-10 Ricoh Co., Ltd. System for treating saved queries as searchable documents in a document management system
US20040261016A1 (en) * 2003-06-20 2004-12-23 Miavia, Inc. System and method for associating structured and manually selected annotations with electronic document contents
US20050289456A1 (en) * 2004-06-29 2005-12-29 Xerox Corporation Automatic extraction of human-readable lists from documents

Family Cites Families (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6339767B1 (en) 1997-06-02 2002-01-15 Aurigin Systems, Inc. Using hyperbolic trees to visualize data generated by patent-centric and group-oriented data processing
US5737442A (en) * 1995-10-20 1998-04-07 Bcl Computers Processor based method for extracting tables from printed documents
WO1998012616A2 (en) 1996-09-23 1998-03-26 Lowrie Mcintosh Defining a uniform subject classification system incorporating document management/records retention functions
US6038561A (en) 1996-10-15 2000-03-14 Manning & Napier Information Services Management and analysis of document information text
US5835905A (en) * 1997-04-09 1998-11-10 Xerox Corporation System for predicting documents relevant to focus documents by spreading activation through network representations of a linked collection of documents
US6298357B1 (en) * 1997-06-03 2001-10-02 Adobe Systems Incorporated Structure extraction on electronic documents
WO2000045297A1 (en) 1998-11-20 2000-08-03 Smart Online, Inc. Systems, methods and computer program products for mining data from host computers via the internet
US6601233B1 (en) * 1999-07-30 2003-07-29 Accenture Llp Business components framework
FR2806814B1 (en) * 2000-03-22 2006-02-03 Oce Ind Sa METHOD OF RECOGNIZING AND INDEXING DOCUMENTS
US7689906B2 (en) * 2000-04-06 2010-03-30 Avaya, Inc. Technique for extracting data from structured documents
JP4651876B2 (en) * 2001-07-19 2011-03-16 富士通株式会社 PATTERN IDENTIFICATION DEVICE, PATTERN IDENTIFICATION METHOD, AND PATTERN IDENTIFICATION PROGRAM
JP3829667B2 (en) * 2001-08-21 2006-10-04 コニカミノルタホールディングス株式会社 Image processing apparatus, image processing method, program for executing image processing method, and storage medium storing program
US7142728B2 (en) * 2002-05-17 2006-11-28 Science Applications International Corporation Method and system for extracting information from a document
WO2004095195A2 (en) * 2003-04-21 2004-11-04 Document Images, Llc System and method for managing imaged freight documents
US20060242180A1 (en) 2003-07-23 2006-10-26 Graf James A Extracting data from semi-structured text documents
US8156427B2 (en) * 2005-08-23 2012-04-10 Ricoh Co. Ltd. User interface for mixed media reality
JP4789516B2 (en) * 2005-06-14 2011-10-12 キヤノン株式会社 Document conversion apparatus, document conversion method, and storage medium
US20060294101A1 (en) * 2005-06-24 2006-12-28 Content Analyst Company, Llc Multi-strategy document classification system and method

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5991709A (en) * 1994-07-08 1999-11-23 Schoen; Neil Charles Document automated classification/declassification system
US6263335B1 (en) * 1996-02-09 2001-07-17 Textwise Llc Information extraction system and method using concept-relation-concept (CRC) triples
US6202060B1 (en) * 1996-10-29 2001-03-13 Bao Q. Tran Data management system
US6629097B1 (en) * 1999-04-28 2003-09-30 Douglas K. Keith Displaying implicit associations among items in loosely-structured data sets
US6546133B1 (en) * 1999-09-08 2003-04-08 Ge Capital Commercial Finance, Inc. Methods and apparatus for print scraping
US6775665B1 (en) * 1999-09-30 2004-08-10 Ricoh Co., Ltd. System for treating saved queries as searchable documents in a document management system
US20020049705A1 (en) * 2000-04-19 2002-04-25 E-Base Ltd. Method for creating content oriented databases and content files
US20040261016A1 (en) * 2003-06-20 2004-12-23 Miavia, Inc. System and method for associating structured and manually selected annotations with electronic document contents
US20050289456A1 (en) * 2004-06-29 2005-12-29 Xerox Corporation Automatic extraction of human-readable lists from documents

Cited By (31)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080033903A1 (en) * 2006-08-04 2008-02-07 Andrew Carol Methods and apparatuses for using location information
US20090300054A1 (en) * 2008-05-29 2009-12-03 Kathleen Fisher System for inferring data structures
US20100010981A1 (en) * 2008-07-08 2010-01-14 International Business Machines Corporation Technique for enhancing a set of website bookmarks by finding related bookmarks based on a latent similarity metric
US8117205B2 (en) 2008-07-08 2012-02-14 International Business Machines Corporation Technique for enhancing a set of website bookmarks by finding related bookmarks based on a latent similarity metric
US9501495B2 (en) 2010-04-22 2016-11-22 Apple Inc. Location metadata in a media file
US20130111326A1 (en) * 2011-10-26 2013-05-02 Kimber Lockhart Enhanced multimedia content preview rendering in a cloud content management system
US11210610B2 (en) * 2011-10-26 2021-12-28 Box, Inc. Enhanced multimedia content preview rendering in a cloud content management system
US11232481B2 (en) 2012-01-30 2022-01-25 Box, Inc. Extended applications of multimedia content previews in the cloud-based content management system
WO2014146032A3 (en) * 2013-03-15 2014-12-24 Mercury File, Llc System for method for data sweeping using keywords
WO2014146032A2 (en) * 2013-03-15 2014-09-18 Suarez Sergio David Jr System for method for data sweeping using keywords
US11630942B2 (en) 2015-08-03 2023-04-18 Blackboiler, Inc. Method and system for suggesting revisions to an electronic document
US10824797B2 (en) 2015-08-03 2020-11-03 Blackboiler, Inc. Method and system for suggesting revisions to an electronic document
US10970475B2 (en) 2015-08-03 2021-04-06 Blackboiler, Inc. Method and system for suggesting revisions to an electronic document
US11093697B2 (en) 2015-08-03 2021-08-17 Blackboiler, Inc. Method and system for suggesting revisions to an electronic document
WO2018156781A1 (en) * 2017-02-22 2018-08-30 Anduin Transactions, Inc. Compact presentation of automatically summarized information according to rule-based graphically represented information
US11755997B2 (en) 2017-02-22 2023-09-12 Anduin Transactions, Inc. Compact presentation of automatically summarized information according to rule-based graphically represented information
US20180239959A1 (en) * 2017-02-22 2018-08-23 Anduin Transactions, Inc. Electronic data parsing and interactive user interfaces for data processing
US11709995B2 (en) 2018-03-30 2023-07-25 Blackboiler, Inc. Method and system for suggesting revisions to an electronic document
US10713436B2 (en) 2018-03-30 2020-07-14 BlackBoiler, LLC Method and system for suggesting revisions to an electronic document
US11244110B2 (en) 2018-03-30 2022-02-08 Blackboiler, Inc. Method and system for suggesting revisions to an electronic document
US10614157B1 (en) 2018-10-25 2020-04-07 BlackBoiler, LLC Systems, methods, and computer program products for slot normalization of text data
US10311140B1 (en) * 2018-10-25 2019-06-04 BlackBoiler, LLC Systems, methods, and computer program products for a clause library
CN110096626A (en) * 2019-03-18 2019-08-06 平安普惠企业管理有限公司 Processing method, device, equipment and the storage medium of contract text data
CN110943828A (en) * 2019-11-05 2020-03-31 武汉理工大学 Secret number operation conversion method and system
US20210295261A1 (en) * 2020-03-20 2021-09-23 Codexo Generating actionable information from documents
US11688027B2 (en) * 2020-03-20 2023-06-27 Codexo Generating actionable information from documents
CN112529743A (en) * 2020-12-18 2021-03-19 平安银行股份有限公司 Contract element extraction method, contract element extraction device, electronic equipment and medium
US11681864B2 (en) 2021-01-04 2023-06-20 Blackboiler, Inc. Editing parameters
CN115017550A (en) * 2022-06-02 2022-09-06 湖南长银五八消费金融股份有限公司 Electronic contract data processing method and device, computer equipment and medium
US20230394221A1 (en) * 2022-06-06 2023-12-07 Microsoft Technology Licensing, Llc Converting a portable document format to a latex format
CN114973290A (en) * 2022-07-26 2022-08-30 简单汇信息科技(广州)有限公司 Intelligent order examination method and system based on OCR engine

Also Published As

Publication number Publication date
US8140468B2 (en) 2012-03-20
US20080235227A1 (en) 2008-09-25

Similar Documents

Publication Publication Date Title
US8140468B2 (en) Systems and methods to extract data automatically from a composite electronic document
CN109062874B (en) Financial data acquisition method, terminal device and medium
US9495347B2 (en) Systems and methods for extracting table information from documents
CN114616572A (en) Cross-document intelligent writing and processing assistant
US9304993B2 (en) Methods and data structures for multiple combined improved searchable formatted documents including citation and corpus generation
US8954839B2 (en) Contract authoring system and method
US8892579B2 (en) Method and system of data extraction from a portable document format file
US20060288268A1 (en) Method for extracting, interpreting and standardizing tabular data from unstructured documents
US20040221233A1 (en) Systems and methods for report design and generation
US20090148048A1 (en) Information classification device, information classification method, and information classification program
US10699112B1 (en) Identification of key segments in document images
US20130275451A1 (en) Systems And Methods For Contract Assurance
US10216733B2 (en) Smart commenting software
JP7208872B2 (en) Systems and methods for generating proposals based on request for proposals (RFPs)
CN114444465A (en) Information extraction method, device, equipment and storage medium
US20060248037A1 (en) Annotation of inverted list text indexes using search queries
CN116521621A (en) Data processing method and device, electronic equipment and storage medium
CN115934926A (en) Information extraction method and device, computer equipment and storage medium
US20140201193A1 (en) Intellectual property asset information retrieval system
Kwok et al. An automatic method to extract data from an electronic contract composed of a number of documents in PDF format
US20090259995A1 (en) Apparatus and Method for Standardizing Textual Elements of an Unstructured Text
WO2016060548A1 (en) Electronic document and electronic file
KR102518843B1 (en) Enterprise content management system using a latene dirichlet allocation
CN116257602B (en) Method and device for constructing universal word stock based on public words and electronic equipment
US20220358287A1 (en) Text mining based on document structure information extraction

Legal Events

Date Code Title Description
AS Assignment

Owner name: INTERNATIONAL BUSINESS MACHINES CORPORATION, NEW Y

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:KWOK, THOMAS YU-KIU;NGUYEN, THAO NGOC;ROY, KAKAN;SIGNING DATES FROM 20060622 TO 20060626;REEL/FRAME:017889/0275

Owner name: INTERNATIONAL BUSINESS MACHINES CORPORATION, NEW Y

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:KWOK, THOMAS YU-KIU;NGUYEN, THAO NGOC;ROY, KAKAN;REEL/FRAME:017889/0275;SIGNING DATES FROM 20060622 TO 20060626

STCB Information on status: application discontinuation

Free format text: ABANDONED -- FAILURE TO RESPOND TO AN OFFICE ACTION