Browse Prior Art Database

Method and System for Automatically Explaining Entity Relationships in a Knowledge Graph

IP.com Disclosure Number: IPCOM000242025D
Publication Date: 2015-Jun-15
Document File: 2 page(s) / 35K

Publishing Venue

The IP.com Prior Art Database

Related People

Edgar Meij: INVENTOR [+2]

Abstract

A method and system is disclosed for automatically explaining entity relationships in a knowledge graph. An entity relationship is analyzed by extracting sentences from documents in a corpus which is related to the knowledge graph. The sentences are enriched and ranked in a best way to define the entity relationship. Here, ranking of sentences is done by leveraging a rich set of features and utilizing supervised machine learning.

This text was extracted from a Microsoft Word document.
This is the abbreviated version, containing approximately 52% of the total text.

Method and System for Automatically Explaining Entity Relationships in a Knowledge Graph

Abstract

A method and system is disclosed for automatically explaining entity relationships in a knowledge graph.  An entity relationship is analyzed by extracting sentences from documents in a corpus which is related to the knowledge graph.  The sentences are enriched and ranked in a best way to define the entity relationship.  Here, ranking of sentences is done by leveraging a rich set of features and utilizing supervised machine learning.

Description

Conventional methods are used to automatically construct knowledge graphs from unstructured web data and are inferred by humans.  However, the method of providing explicit evidence of constructing the knowledge graph between entities and relating the entities remains unaddressed.  However, such approaches are not able to provide an automatic method of inferring human-readable explanations for entity relationships.

Disclosed is a method and system for automatically explaining entity relationships in a knowledge graph.  The knowledge graph is used to aggregate information between entities through multiple content sources and to link the respective entities. The method and system for analyzing a relationship between entities is performed by the below two steps.

The first step is to extract and enrich sentences that refer to an entity pair in a corpus for constructing a set of candidate sentences.  The method starts by splitting documents into sentences and extracting sentences that refer to an entity pair.  Here, the method of extracting sentences is performed based on “surface forms” of the given entities.  The “surface forms” may include, but need not be limited to, title of the entity, synonyms of the title, and all anchor texts associated with hyperlinks to the entity.  Subsequently, if the sentence contains surface forms of both entities in the entity pair, each sentence is considered as a candidate.

Next, the method of enriching the sentence is performed by applying two enrichment steps to the sentences.  The first enrichment step is to make sentences self-contained and readable outside the context of...