Systems and methods for analysis and comparison of electronic documents
AssigneeQ2 Software, Inc.
InventorsAli Zonoozi, Cameron Christopher Farr Plouffe, Daniel Wagner, Frederico Tommasi Caroli, Kollol Das
Life of this patent
Estimating the term…
Abstract
Systems and methods for document comparison are disclosed. Embodiments may determine matching entities within documents utilizing a model such as a large language model or the like to produce graphs of documents being compared according to a graph schema defining entities and relationships. The graphs generated for the documents can be compared to determine matching entities in the documents despite the fact that these entities may be referred to differently in each of those documents.
Claims22
A system for comparing documents, comprising:
- a processor;
- and
- a non-transitory computer readable medium, comprising instructions for:
- receiving a plurality of documents to be compared;
- obtaining a graph schema defining entity types and relationships, the graph schema being subdivided into one or more subgraph schemas;
- for each document:
- generating a subgraph associated with each of the one or more subgraph schemas by:
- providing a representation of that subgraph schema to a large language model (LLM) in association with text of the document,
- receiving, from the LLM, entity data corresponding to the subgraph schema, and
- generating the subgraph based on the entity data received from the LLM;
- merging the plurality of subgraphs generated for the document to produce a unified graph representation of the document;
- and
- comparing the unified graph representations of the plurality of documents to identify matching entities associated with the plurality of documents.
A method for comparing documents, comprising:
- receiving a plurality of documents to be compared;
- obtaining a graph schema defining entity types and relationships, the graph schema being subdivided into one or more subgraph schemas;
- for each document:
- generating a subgraph associated with each of the one or more subgraph schemas by:
- providing a representation of that subgraph schema to a large language model (LLM) in association with text of the document,
- receiving, from the LLM, entity data corresponding to the subgraph schema, and
- generating the subgraph based on the entity data received from the LLM;
- merging the plurality of subgraphs generated for the document to produce a unified graph representation of the document;
- and
- comparing the unified graph representations of the plurality of documents to identify matching entities associated with the plurality of documents.
A non-transitory computer readable medium, comprising instructions for:
- receiving a plurality of documents to be compared;
- obtaining a graph schema defining entity types and relationships, the graph schema being subdivided into one or more subgraph schemas;
- for each document:
- generating a subgraph associated with each of the one or more subgraph schemas by:
- providing a representation of that subgraph schema to a large language model (LLM) in association with text of the document,
- receiving, from the LLM, entity data corresponding to the subgraph schema, and
- generating the subgraph based on the entity data received from the LLM;
- merging the plurality of subgraphs generated for the document to produce a unified graph representation of the document;
- and
- comparing the unified graph representations of the plurality of documents to identify matching entities associated with the plurality of documents.
Dependencies are inferred from the claim text.