Artificial intelligence based corpus enrichment for knowledge population and query response

Patent metadata

Field	Value
Publication number	US-2020073882-A1
Application number	US-201816195471-A
Country	US
Kind code	A1
Filing date	Nov 19, 2018
Priority date	Aug 31, 2018
Publication date	Mar 5, 2020
Grant date	—

Abstract

In some examples, artificial intelligence based corpus enrichment for knowledge population and query response may include generating, based on annotated training documents, an entity and relation annotation model, identifying, based on application of the entity and relation annotation model to a document set that is to be annotated, entities and relations between the entities for each document of the document set to generate an annotated document set, and categorizing each annotated document into a plurality of categories. Artificial intelligence based corpus enrichment may include determining whether an identified category includes a specified number of annotated documents, and if not, additional annotated documents may be generated for the identified category that may represent a corpus. Further, artificial intelligence based corpus enrichment may include training, using the corpus, an artificial intelligence based decision support model, and utilizing the artificial intelligence based decision support model to respond to an inquiry.

First claim

What is claimed is:

1 . An artificial intelligence based corpus enrichment for knowledge population and query response apparatus comprising:
an entity and relation annotator, executed by at least one hardware processor, to
generate, based on annotated training documents, an entity and relation annotation model,
ascertain, a document set that is to be annotated, wherein documents of the document set include unstructured documents and semi-structured documents, and
identify, based on application of the entity and relation annotation model to the document set, entities and relations between the entities for each document of the document set to generate an annotated document set;
a document categorizer, executed by the at least one hardware processor, to
categorize each annotated document of the annotated document set into a respective category of a plurality of categories;
a corpus generator and enricher, executed by the at least one hardware processor, to
identify a category of the plurality of categories,
determine whether the identified category includes a specified number of annotated documents, and
based on a determination that the identified category does not include the specified number of annotated documents, generate, for the identified category, additional annotated documents, wherein the annotated documents and the additional annotated documents of the identified category together represent a corpus;
an artificial intelligence model generator, executed by the at least one hardware processor, to
train, using the corpus, an artificial intelligence based decision support model; and
an inquiry response generator, executed by the at least one hardware processor, to
ascertain an inquiry related to an entity of the corpus, and
generate, by invoking the artificial intelligence based decision support model, a response to the inquiry.

2 . The artificial intelligence based corpus enrichment for knowledge population and query response apparatus according to claim 1 , wherein the entity and relation annotator is further executed by the at least one hardware processor to generate, based on the annotated training documents, the entity and relation annotation model by:
transforming each annotated training document of the annotated training documents into a vector representation; and generating, based on vector representations of the annotated training documents, the entity and relation annotation model.

3 . The artificial intelligence based corpus enrichment for knowledge population and query response apparatus according to claim 1 , wherein the entity and relation annotator is further executed by the at least one hardware processor to:
determine, for each identified entity of the identified entities, an entity confidence score; determine, for each identified relation of the identified relations between the entities, a relation confidence score; identify, based on the entity confidence score, an entity that includes an entity confidence score that is less than an entity confidence score threshold; identify, based on the relation confidence score, a relation that includes a relation confidence score that is less than a relation confidence score threshold; generate another inquiry for verification of the entity and the relation that respectively include the entity confidence score and the relation confidence score that are respectively less than the entity confidence score threshold and the relation confidence score threshold; and train, based on a response to the other inquiry for verification of the entity and the relation that respectively include the entity confidence score and the relation confidence score that are respectively less than the entity confidence score threshold and the relation confidence score threshold, the entity and relation annotation model.

4 . The artificial intelligence based corpus enrichment for knowledge population and query response apparatus according to claim 1 , wherein the entity and relation annotator is further executed by the at least one hardware processor to:
determine, for each identified entity of the identified entities, an entity confidence score; determine, for each identified relation of the identified relations between the entities, a relation confidence score; identify entities for which a difference between entity confidence scores is less than a specified numerical value; identify relations for which a difference between relation confidence scores is less than the specified numerical value; generate another inquiry for verification of the entities for which the difference between the entity confidence scores is less than the specified numerical value, and the relations for which the difference between the relation confidence scores is less than the specified numerical value; and train, based on a response to the other inquiry for verification of the entities for which the difference between the entity confidence scores is less than the specified numerical value, and the relations for which the difference between the relation confidence scores is less than the specified numerical value, the entity and relation annotation model.

5 . The artificial intelligence based corpus enrichment for knowledge population and query response apparatus according to claim 1 , wherein the document categorizer is further executed by the at least one hardware processor to:
categorize each document of the document set that is to be annotated into the respective category of the plurality of categories, wherein the entity and relation annotator is further executed by the at least one hardware processor to identify, based on application of the entity and relation annotation model to the document set, entities and relations between the entities for each document of the document set to generate the annotated document set by: identifying, based on application of the entity and relation annotation model to documents of the identified category, entities and relations between the entities for each document of the identified category to generate annotated documents for the identified category.

6 . The artificial intelligence based corpus enrichment for knowledge population and query response apparatus according to claim 1 , wherein the document categorizer is further executed by the at least one hardware processor to categorize each annotated document of the annotated document set into the respective category of the plurality of categories by:
transforming each annotated document of the annotated document set into an entity vector; grouping, based on the entity vector for each annotated document, semantically similar entities; and categorizing, based on the grouping, each annotated document of the annotated document set into the respective category of the plurality of categories.

7 . The artificial intelligence based corpus enrichment for knowledge population and query response apparatus according to claim 1 , wherein the corpus generator and enricher is further executed by the at least one hardware processor to generate, for the identified category, additional annotated documents by:
segmenting the corpus into a plurality of sections to generate a preprocessed and segmented entity-annotated textual corpus.

8 . The artificial intelligence based corpus enrichment for knowledge population and query response apparatus according to claim 7 , wherein for an invoice document, the plurality of sections include header, body, payment, and reference.

9 . The artificial intelligence based corpus enrichment for knowledge population and query response apparatus according to claim 7 , wherein the corpus generator and enricher is further executed by the at least one hardware pr

Assignees

• Accenture Global Solutions Ltd

Inventors

• Guggilla Chinnappa
• Shishtla Praneeth
• Shivaram Madhura

Classifications

• G06F16/93 Primary

  Document management systems · CPC title

• G06F16/3347

  using vector based model · CPC title

• G06F16/353 Primary

  into predefined classes · CPC title

• G06N20/00

  Machine learning · CPC title

• G06N5/02

  Knowledge representation; Symbolic representation · CPC title