Systems and methods for natural language processing of structured documents

Patent metadata

Field	Value
Publication number	US-2019005029-A1
Application number	US-201816021112-A
Country	US
Kind code	A1
Filing date	Jun 28, 2018
Priority date	Jun 30, 2017
Publication date	Jan 3, 2019
Grant date	—

Abstract

Systems and methods for natural language processing of structured documents. In another embodiment, in an information processing apparatus comprising at least one computer processor, a method for processing a structured document may include: (1) receiving a document; (2) parsing the document into a plurality of components using a statistical parser; (3) extracting a plurality of entities from each component; (4) identifying a potential relationship between two of the plurality of entities; (5) generating a numeric representation for the potential relationship; (6) confirming the potential relationship with a logical regression model; and (7) generating and storing a unified structured file for the document.

First claim

What is claimed is:

1 . A method for processing a structured document, comprising:
in an information processing apparatus comprising at least one computer processor:
receiving a document;
parsing the document into a plurality of components using a statistical parser;
extracting a plurality of entities from each component;
identifying a potential relationship between two of the plurality of entities;
generating a numeric representation for the potential relationship;
confirming the potential relationship with a logical regression model; and
generating and storing a unified structured file for the document.

2 . The method of claim 1 , wherein the statistical parser comprises a neural network.

3 . The method of claim 1 , wherein the plurality of components comprise at least one of a participating party, an article, a section, a subsection, and a subsubsection.

4 . The method of claim 1 , wherein the statistical parser parses the document based on a first vector of word embeddings and a second vector of orthographic properties of words in the document.

5 . The method of claim 1 , wherein the step of parsing the document into a plurality of components comprises identifying a relationship among the plurality of components.

6 . The method of claim 1 , further comprising filtering the document into a plurality of sections using a statistical section filter.

7 . The method of claim 1 , further comprising:
generating a score for each sentence or paragraph of the document

8 . The method of claim 7 , wherein the score is generated using a latent semantic indexing model.

9 . The method of claim 7 , wherein the score is generated using a continuous bag-of-words model.

10 . The method of claim 1 , wherein the plurality of entities are extracted using a Conditional Random Field model.

11 . The method of claim 1 , wherein the potential relationship is based on an ontology.

12 . The method of claim 1 , wherein the potential relationship is based on a hierarchical correspondence rule.

13 . The method of claim 1 , wherein the numeric representation for the potential relationship is based on functional features, tail features, and head features of the potential relationship.

14 . The method of claim 1 , further comprising:
identifying a plurality of defined terms in the document.

15 . The method of claim 1 , wherein the logical regression model confirms each potential relationship as being true or false.

16 . The method of claim 1 , further comprising:
generating a graphical representation of the document.

17 . The method of claim 1 , wherein the document comprises a structured document.

18 . The method of claim 1 , further comprising:
receiving feedback on an accuracy of at least one of the plurality of components identified using the statistical parser; and updating the statistical parser based on the feedback.

19 . The method of claim 1 , further comprising:
receiving feedback on an accuracy of at least one of the entities extracted using the Conditional Random Field model; and updating the Conditional Random Field model based on the feedback.

20 . The method of claim 1 , further comprising:
receiving feedback on an accuracy of at least one of the potential relationships confirmed using the logical regression model; and updating the logical regression model based on the feedback.

Assignees

• Jpmorgan Chase Bank Na

Inventors

• Mills Jason D
• Wu Zheng
• Rabowsky Jennifer
• Song Kenny
• Bugrara Khaled

Classifications

• G06F40/30

  Semantic analysis · CPC title

• G06F40/40 Primary

  Processing or translation of natural language (natural language analysis G06F40/20; semantic analysis G06F40/30) · CPC title

• G06F40/106

  Display of layout of documents; Previewing · CPC title

• G06F40/205 Primary

  Parsing · CPC title

• G06F17/2785

  Physics · mapped topic