Identification of legal concepts in legal documents

1. A method for identifying a legal concept associated with a legal document, comprising:

training a machine learning model for predicting an identification of statements and associated citations to cited documents in legal documents using training legal documents annotated to identify statements and associated citations;

identifying a statement and an associated citation to a cited document in a legal document by determining that the statement corresponds with the associated citation to the cited document based on a type of the cited document using the trained machine learning model;

determining a correspondence between a statement in the cited document and the statement identified in the legal document by comparing possible statements identified in the cited document with the statement identified from the legal document; and

identifying a legal concept associated with the legal document as the statement identified in the legal document based on the correspondence.

2. The method as recited in claim 1 , wherein determining a correspondence between a statement in the cited document and the statement identified in the legal document further comprises:

determining a machine-readable representation of the citation to the cited document; and retrieving the cited document based on the machine-readable representation.

3. The method as recited in claim 1 , wherein determining a correspondence between a statement in the cited document and the statement identified in the legal document further comprises:

filtering the possible statements prior to the comparing.

4. The method as recited in claim 1 , wherein identifying a statement and an associated citation to a cited document in a legal document comprises:

identifying the statement and the associated citation to the cited document in the legal document based on candidate citations identified in the legal document, candidate statements identified in the legal document, and sentence boundaries identified in the legal document.

5. The method as recited in claim 4 , wherein identifying the statement and the associated citation to the cited document in the legal document based on candidate citations identified in the legal document, candidate statements identified in the legal document, and sentence boundaries identified in the legal document comprises:

parsing the legal document based on the candidate citations identified in the legal document, the candidate statements identified in the legal document, and the sentence boundaries identified in the legal document; and

applying the trained machine learning model to identify the statement and the associated citation to the cited document in the legal document based on the parsed legal document.

6. The method as recited in claim 1 , further comprising: identifying other legal documents associated with the legal concept; and generating a map showing a relationship between the legal document and the other legal documents.

7. The method as recited in claim 6 , wherein generating a map showing a relationship between the legal document and the other legal documents comprises:

generating a directed graph having nodes representing the legal document and the other legal documents, and edges connecting the nodes, the edges directed from a node representing a citing document to a node representing a cited document.

8. An apparatus, comprising:

a processor; and

a memory to store computer program instructions for identifying a legal concept associated with a legal document, the computer program instructions when executed on the processor cause the processor to perform operations comprising:

training a machine learning model for predicting an identification of statements and associated citations to cited documents in legal documents using training legal documents annotated to identify statements and associated citations;

identifying a statement and an associated citation to a cited document in a legal document by determining that the statement corresponds with the associated citation to the cited document based on a type of the cited document using the trained machine learning model;

determining a correspondence between a statement in the cited document and the statement identified in the legal document by comparing possible statements identified in the cited document with the statement identified from the legal document; and

identifying a legal concept associated with the legal document as the statement identified in the legal document based on the correspondence.

9. The apparatus as recited in claim 8 , wherein determining a correspondence between a statement in the cited document and the statement identified in the legal document further comprises:

determining a machine-readable representation of the citation to the cited document; and

retrieving the cited document based on the machine-readable representation.

10. The apparatus as recited in claim 8 , wherein determining a correspondence between a statement in the cited document and the statement identified in the legal document further comprises:

filtering the possible statements prior to the comparing.

11. A non-transitory computer readable medium storing computer program instructions for identifying a legal concept associated with a legal document, which, when executed on a processor, cause the processor to perform operations comprising:

training a machine learning model for predicting an identification of statements and associated citations to cited documents in legal documents using training legal documents annotated to identify statements and associated citations;

identifying a statement and an associated citation to a cited document in a legal document by determining that the statement corresponds with the associated citation to the cited document based on a type of the cited document using the trained machine learning model;

determining a correspondence between a statement in the cited document and the statement identified in the legal document by comparing possible statements identified in the cited document with the statement identified from the legal document; and

identifying a legal concept associated with the legal document as the statement identified in the legal document based on the correspondence.

12. The non-transitory computer readable medium as recited in claim 11 , wherein identifying a statement and an associated citation to a cited document in a legal document comprises:

identifying the statement and the associated citation to the cited document in the legal document based on candidate citations identified in the legal document, candidate statements identified in the legal document, and sentence boundaries identified in the legal document.

13. The non-transitory computer readable medium as recited in claim 12 , wherein identifying the statement and the associated citation to the cited document in the legal document based on candidate citations identified in the legal document, candidate statements identified in the legal document, and sentence boundaries identified in the legal document comprises:

parsing the legal document based on the candidate citations identified in the legal document, the candidate statements identified in the legal document, and the sentence boundaries identified in the legal document; and

applying the trained machine learning model to identify the statement and the associated citation to the cited document in the legal document based on the parsed legal document.

14. The non-transitory computer readable medium as recited in claim 11 , the operations further comprising:

identifying other legal documents associated with the legal concept; and

generating a map showing a relationship between the legal document and the other legal documents.

15. The non-transitory computer readable medium as recited in claim 14 , wherein generating a map showing a relationship between the legal document and the other legal documents comprises:

generating a directed graph having nodes representing the legal document and the other legal documents, and edges connecting the nodes, the edges directed from a node representing a citing document to a node representing a cited document.