System and Method for Identifying Text in Legal documents for Preparation of Headnotes

• US 20130238316A1
• Filed: 03/07/2013
• Published: 09/12/2013
• Est. Priority Date: 03/07/2012
• Status: Active Grant

• Current Assignee: Infosys Limited
• Original Assignee: Infosys Limited
• Inventors: SHASTRI, Lokendra, MAHARATHY, Sudesna, KUMAR, Abhishek, MISHRA, Gaurav, RANGI, Sankalp Singh, SIVAKUMAR, Sivasubramaniam, CHALAVADI, Suman Kumar
• Granted Patent: US 9,058,308 B2
• Time in Patent Office: Days
• US Class Current: 704/9
• CPC Class Codes: G06F 40/10 Text processing natural language analysis G06F40/20; semantic analysis G06F40/30; processing or translation of natural language G06F40/40
• CPC Class Codes: G06F 40/279 Recognition of textual entities

Abstract

A method for generating feature graphs employed for creation of a head note in a legal document is provided. The method enables identifying one or more predetermined features in a plurality of legal documents. The one or more predetermined features are based on grammatical constituents of text in the legal document. The plurality of legal documents is manually identified as headnote and non headnote. The method further enables obtaining data related to the availability of the one or more identified predetermined features in the sentences manually identified as headnote and non headnote in the plurality of legal documents. Furthermore, the method enables computing likelihood of a sentence being a headnote based on the obtained data. The method further enables generating feature graphs corresponding to each predetermined feature based on the computed likelihood and obtained data and storing the generated feature graphs in a repository.

Claims

1. A method for generating feature graphs employed for creation of a headnote in a legal document, the method comprising:

identifying one or more predetermined features in a plurality of legal documents, wherein the one or more predetermined features are based on grammatical constituents of text in the legal document, the plurality of legal documents being manually identified as headnote and non headnote;

obtaining data related to the availability of the one or more identified predetermined features in the sentences manually identified as headnote and non-headnote in the plurality of legal documents;

computing likelihood of a sentence being a headnote based on the obtained data;

generating feature graphs corresponding to each predetermined feature based on the computed likelihood and obtained data; and

storing the generated feature graphs in a repository.

2. The method of claim 1, wherein the one or more predetermined features comprise at least one of:

a proper noun feature, a date feature, a provenance feature, a sentence complexity feature, a sentence structure feature, a syntactic structure feature, an abstract-entity extraction feature, a comma feature, an ellipses feature, a colon feature, a see or cite or refer feature, a versus feature, a statute and precedence feature, a sentence position feature and a document structure feature.

3. The method of claim 1, wherein identifying the one or more predetermined features in the plurality of legal documents comprises identifying the one or more predetermined features by referring to at least one of:

lexical, syntactic and semantic information stored in a knowledge repository.

4. The method of claim 1, wherein obtaining data related to the availability of the one or more identified predetermined features in the sentences manually identified as headnote and non headnote comprises:

determining feature values for each of the identified predetermined features, wherein the feature values represent the number of occurrences of the identified predetermined features;

obtaining number of headnote sentences for a specific feature value of each of the identified predetermined features; and

obtaining number of non headnote sentences for the specific feature value of each of the identified predetermined features.

5. The method of claim 4, wherein computing likelihood of a sentence being a headnote based on the obtained data comprises calculating probability of a sentence being a headnote using the equation:

6. The method of claim 5, wherein generating feature graphs corresponding to each predetermined feature based on the computed likelihood and obtained data comprises:

generating a feature graph where the X-Axis represents the feature values of the predetermined feature and the Y-Axis represents the calculated probability values corresponding to each feature value.

7. The method of claim 1, wherein computing likelihood of a sentence being a headnote based on the obtained data comprises computing evidence for a sentence being a headnote or a non-headnote using Dempster-Shafer theory or any other evidence combination techniques.

8. The method of claim 1, wherein computing likelihood of a sentence being a headnote based on the obtained data comprises computing evidence for a sentence being a headnote or a non-headnote using one or more Fuzzy Logic techniques.

9. A method for extracting text in a legal document for preparation of headnotes, the method comprising:

detecting one or more predetermined features in each sentence of the legal document, wherein the one or more predetermined features are based on grammatical constituents of text in the legal document;

computing occurrence of the detected one or more predetermined features in each sentence of the legal document;

retrieving feature graphs corresponding to the one or more detected predetermined features from a repository, wherein the feature graphs represent likelihood of a sentence being a headnote for a specific number of occurrences of the predetermined features in the sentences;

analyzing the detected one or more predetermined features based on the computed occurrence of the one or more detected predetermined features and the retrieved feature graphs; and

rendering the legal document with text tagged as headnote on a user interface based on the analysis.

10. The method of claim 9, wherein analyzing the detected one or more predetermined features based on the computed occurrence of the one or more detected predetermined features and the retrieved corresponding feature graphs comprises:

assigning a score to the one or more predetermined features detected in each sentence by referring to the retrieved feature graphs corresponding to the one or more detected predetermined features;

combining the assigned score in each sentence to obtain a final headnote score for each sentence;

comparing the final headnote score with a predetermined threshold; and

tagging text in each sentence as headnote and non headnote based on the comparison.

11. A method for extracting text in a legal document for preparation of headnotes, the method comprising:

detecting one or more predetermined features in each sentence of the legal document, wherein the one or more predetermined features are based on grammatical constituents of text in the legal document;

computing occurrence of the detected one or more predetermined features in each sentence of the legal document;

assigning a score to each predetermined feature detected in each sentence by referring to feature graphs corresponding to each predetermined feature, wherein the feature graphs are retrieved from a repository;

combining the assigned score in each sentence to obtain a final headnote score for each sentence;

comparing the final headnote score with a predetermined threshold;

tagging text in each sentence as headnote and non headnote based on the comparison; and

rendering the legal document with text tagged as headnote on a user interface.

12. The method of claim 11 further comprising:

removing unparseable text from the legal document;

chunking the received legal document; and

splitting the legal document into sentences by using the chunks.

13. The method of claim 11, wherein the one or more predetermined features comprise at least one of:

a proper noun feature, a date feature, a provenance feature, a sentence complexity feature, a sentence structure feature, a syntactic structure feature, an abstract-entity extraction feature, a comma feature, an ellipses feature, a colon feature, a see or cite or refer feature, a versus feature, a statute and precedence feature, a sentence position feature and a document structure feature.

14. The method of claim 11, wherein detecting the one or more predetermined features in each sentence of the legal document comprises detecting the one or more predetermined features in each sentence of the legal document by referring to lexical, syntactic and semantic information stored in a knowledge repository.

15. The method of claim 14, wherein detecting the one or more predetermined features in each sentence of the legal document comprises detecting the one or more predetermined features based on a predetermined pattern stored in a repository.

16. The method of claim 14, wherein detecting the one or more predetermined features in each sentence of the legal document comprises detecting the one or more predetermined features using a list of words stored in a repository.

17. The method of claim 11, wherein computing occurrence of the detected one or more predetermined features in each sentence of the legal document comprises at least one of:

computing the number of occurrences of the detected one or more predetermined features in each sentence and computing the presence or absence of the detected one or more predetermined features in each sentence.

18. The method of claim 17, wherein assigning a score to each predetermined feature detected in each sentence by referring to feature graphs corresponding to each predetermined feature comprises:

determining a probability value in the feature graph that corresponds to the computed occurrence of the one or more predetermined features; and

assigning a score to the one or more predetermined features based on the probability value.

19. The method of claim 11, wherein combining the assigned score in each sentence to obtain a final headnote score for each sentence comprises:

multiplying the score assigned to each predetermined feature in each sentence to obtain a final headnote score for each sentence.

20. The method of claim 19, wherein the final headnote score obtained is normalized.

21. The method of claim 11, wherein combining the assigned score in each sentence to obtain a final headnote score for each sentence comprises combining the scores assigned to each predetermined feature in the sentence using evidence combination based integration techniques.

22. The method of claim 11, wherein combining the assigned score in each sentence to obtain a final headnote score for each sentence comprises combining the scores assigned to each predetermined feature in each sentence using Fuzzy Logic based integration techniques.

23. The method of claim 11 further comprising:

referring to a scatter diagram representing distribution of number of sentences and corresponding final headnote scores;

selecting a final headnote score that provides a balance between recall and precision; and

selecting the final headnote score as the predetermined threshold.

24. The method of claim 23, wherein comparing the final headnote score with a predetermined threshold comprises selecting the final headnote score of each sentence that are equal to or more than the predetermined threshold.

25. The method of claim 24, wherein tagging text in each sentence as headnote and non headnote based on the comparison comprises tagging the sentences with the selected final headnote score as headnote and tagging the sentences with final headnote score less than the predetermined threshold as non headnote.

26. A system for generating feature graphs employed for creation of a headnote in a legal document, the system comprising:

a feature identification module configured to identify one or more predetermined features in a plurality of legal documents, wherein the one or more predetermined features are based on grammatical constituents of text in the legal document and further wherein the feature identification module receives the plurality of legal documents manually identified as headnote and non headnote from a user interface;

a feature availability computing module configured to;

obtain data related to the availability of the one or more identified predetermined features in the sentences manually identified as headnote and non headnote in the plurality of legal documents; and

store the obtained data in a repository; and

a feature graph generating module configured to;

compute likelihood of a sentence being a headnote based on the data stored in the repository;

generate feature graphs corresponding to each predetermined feature based on the computed likelihood and stored data; and

store the generated feature graphs in the repository.

27. The system of claim 26, wherein the one or more predetermined features comprise at least one of:

a proper noun feature, a date feature, a provenance feature, a sentence complexity feature, a sentence structure feature, a syntactic structure feature, an abstract-entity extraction feature, a comma feature, an ellipses feature, a colon feature, a see or cite or refer feature, a versus feature, a statute and precedence feature, a sentence position feature and a document structure feature.

28. The system of claim 26, wherein the repository is configured to store a set of predetermined features, a list of words corresponding to the predetermined features and a set of predetermined patterns corresponding to the one or more predetermined features.

29. The system of claim 26, wherein the feature identification module identifies one or more predetermined features by referring to a knowledge repository that stores at least one of:

lexical, syntactic and semantic information.

30. The system of claim 26, wherein the feature availability computing module obtains data related to the number of headnote sentences and non headnotes for a specific feature value of each of the predetermined features.

31. The system of claim 30, wherein the feature graph generating module computes probability value for a specific feature value of each of the predetermined features using the obtained data.

32. A headnote preparation system for extracting text in a legal document for preparation of headnotes, the system comprising:

a repository configured to store feature graphs corresponding to one or more predetermined features;

a feature detecting module configured to detect one or more predetermined features in each sentence of the legal document, wherein the one or more predetermined features are based on grammatical constituents of text in the legal document;

a feature score assigning module configured to;

compute occurrence of the detected one or more predetermined features in each sentence of the legal document;

assign a score to each predetermined feature detected in each sentence by referring to feature graphs corresponding to each predetermined feature stored in the repository based on the computed occurrence of the detected one or more features;

and

a feature integrating module configured to combine the assigned score in each sentence to obtain a final headnote score for each sentence;

a headnote generating module configured to;

compare the final headnote score with a predetermined threshold;

tag text in each sentence as headnote and non headnote based on the comparison; and

render the legal document with text tagged as headnote on a user interface, wherein the text tagged as headnote facilitates a user to prepare a headnote for the legal document.

33. The headnote preparation system of claim 32, wherein the feature score assigning module is configured to:

determine a probability value in the features graphs of each of the predetermined features in each sentence based on the computed occurrence of the predetermined features; and

assign a score to the one or more detected predetermined features in each sentence based on the probability value.

34. A method for extracting text in a legal document for preparation of headnotes, the method comprising:

retrieving feature graphs corresponding to one or more of predetermined features, wherein the feature graphs represent probability of a sentence being a headnote for a specific number of occurrence of the predetermined features in the sentence;

assigning a score to each predetermined feature in each sentence by referring to the retrieved feature graphs;

combining each assigned score in each sentence to obtain a final headnote score for each sentence;

tagging text in the legal document as headnote and non headnote by comparing the final headnote score in each sentence with a predetermined threshold; and

rendering text in the legal document that are tagged as headnotes on a user interface.

35. A headnote preparation system for extracting text in a legal document for preparation of headnotes, the system comprising:

a repository for storing feature graphs corresponding to each predetermined feature, wherein the feature graphs represent probability of a sentence being a headnote for a specific number of occurrence of the predetermined features in the sentence;

a feature score assigning module configured to assign a score to each predetermined feature in each sentence of the legal document by referring to the feature graphs retrieved from the repository;

a feature integration module configured to combine each assigned score in each sentence to obtain a final headnote score for each sentence;

a headnote generating module configured to tag text in the legal document as headnote and non headnote by comparing the final headnote score in each sentence with a predetermined threshold; and

a user interface configured to receive the legal document with text tagged as headnotes.

36. A computer program product comprising:

a non-transitory computer-readable medium having computer-readable program code stored thereon, the computer-readable program code comprising instructions that when executed by a processor, cause the processor to;

detect one or more predetermined features in each sentence of the legal document, wherein the one or more predetermined features are based on grammatical constituents of text in the legal document;

compute occurrence of the detected one or more predetermined features in each sentence of the legal document;

retrieve feature graphs corresponding to the one or more detected predetermined features from a repository, wherein the feature graphs represent likelihood of a sentence being a headnote for a specific number of occurrences of the predetermined features in the sentences;

analyze the detected one or more predetermined features based on the computed occurrence of the one or more detected predetermined features and the retrieved feature graphs; and

render the legal document with text tagged as headnote on a user interface based on the analysis.

37. The computer program product of claim 36, wherein the computer-readable program code further comprises instructions that when executed by the processor, cause the processor to:

assign a score to the one or more predetermined features detected in each sentence by referring to the retrieved feature graphs corresponding to the one or more detected predetermined features;

combine the assigned score in each sentence to obtain a final headnote score for each sentence;

compare the final headnote score with a predetermined threshold; and

tag text in each sentence as headnote and non headnote based on the comparison.