Encoding knowledge graph entries with searchable geotemporal values for evaluating transitive geotemporal proximity of entity mentions

Patent metadata

Field	Value
Publication number	US-11526769-B2
Application number	US-201916370968-A
Country	US
Kind code	B2
Filing date	Mar 30, 2019
Priority date	Mar 30, 2019
Publication date	Dec 13, 2022
Grant date	Dec 13, 2022

Abstract

A controller, responsive to detecting geospatial and temporal information associated with a mention of an entity in a document, for converting the geospatial information into a specified geospatial format and the temporal information into a specified temporal format. The controller for computing prefix-based geospatial values for the converted geospatial information and prefix-based temporal values for the converted temporal information. The controller for encoding an entry in a knowledge graph for the mention of the entity from the document with the prefix-based geospatial values and the prefix-based temporal values, wherein each digit of the prefix-based geospatial values and the prefix-based temporal values in the knowledge graph that matches another one or more prefix-based geospatial values and another one or more prefix-based temporal values in another entry for another entity mention in the knowledge graph reflects a degree of granularity of geotemporal proximity of the entity and the another entity.

First claim

What is claimed is:

1. A method, comprising:
responsive to detecting geospatial information and temporal information associated with a mention of an entity in a document, converting, by a computer system, the geospatial information into a specified geospatial format and the temporal information into a specified temporal format;
computing, by the computer system, one or more prefix-based geospatial values for the converted geospatial information and one or more prefix-based temporal values for the converted temporal information; and
encoding, by the computer system, an entry in a knowledge graph for the mention of the entity from the document with the one or more prefix-based geospatial values and the one or more prefix-based temporal values, wherein each digit of the one or more prefix-based geospatial values and the one or more prefix-based temporal values in the knowledge graph that matches a digit of another one or more prefix-based geospatial values and another one or more prefix-based temporal values in another entry for another entity mention in the knowledge graph reflects a degree of granularity of geotemporal proximity of the entity and the another entity.

2. The method according to claim 1 , wherein computing, by the computer system, one or more prefix-based geospatial values for the converted geospatial information and one or more prefix-based temporal values for the converted temporal information further comprises:
computing, by the computer system, a prefix-based geospatial hash value by applying a hash function to the converted geospatial information; and
computing, by the computer system, a prefix-based geospatial offset hash value by rotating a latitude and longitude of the prefix-based geospatial hash value by a rotation value, wherein each digit of the prefix-based geospatial hash value and prefix-based geospatial offset hash value that matches another prefix-based geospatial hash value and another prefix-based geospatial offset hash value in the another entry for the another entity mention in the knowledge graph reflects the degree of granularity of geospatial proximity of the entity and the another entity.

3. The method according to claim 1 , wherein computing, by the computer system, one or more prefix-based geospatial values for the converted geospatial information and one or more prefix-based temporal values for the converted temporal information further comprises:
computing, by the computer system, a prefix-based temporal hash value by applying a hash function to the converted temporal information; and
computing, by the computer system, a prefix-based temporal offset hash value by applying an offset mask to the prefix-based temporal hash value to increase each digit of the prefix-based temporal hash value by half a potential value, wherein each digit of the prefix-based temporal hash value and prefix-based temporal offset hash value that matches another prefix-based temporal hash value and another prefix-based temporal offset hash value in the another entry for the another entity mention in the knowledge graph reflects the degree of granularity of temporal proximity of the entity and the another entity.

4. The method according to claim 1 , wherein encoding, by the computer system, an entry in the knowledge graph for the mention of the entity from the document with the one or more prefix-based geospatial values and the one or more prefix-based temporal values further comprises:
encoding, by the computer system, the entry in the knowledge graph for the mention of the entity from the document with an identifier of a location relationship to the one or more prefix-based geospatial values and the one or more prefix-based temporal values.

5. The method according to claim 1 , further comprising:
storing, by the computer system, the entry in the knowledge graph for the mention of the entity from the document with an identifier of a location relationship, the geospatial information, and the temporal information.

6. The method according to claim 1 , further comprising:
searching, by the computer system, a corpus of documents for one or more mentions of the entity; and
responsive to identifying the document of the corpus of documents with the mention of the entity, searching, by the computer system, the particular document for the geospatial information and the temporal information associated with the mention.

7. The method according to claim 1 , further comprising:
searching, by the computer system, the document for the geospatial information by applying one or more geospatial classifiers each trained to detect one or more geospatial reference types from among a plurality of geospatial reference types;
in response to the one or more geospatial classifiers returning a classification of a first selection of text in the document as the geospatial information, converting, by the computer system, the geospatial information into the specified geospatial format;
searching, by the computer system, the document for the temporal information by applying one or more temporal classifiers each trained to detect one or more temporal reference types from among a plurality of temporal reference types; and
in response to the one or more temporal classifiers returning a classification of a second selection of text in the document as the temporal information, converting, by the computer system, the temporal information into the specified temporal format.

8. A computer system comprising one or more processors, one or more computer-readable memories, one or more computer-readable storage devices, and program instructions, stored on at least one of the one or more storage devices for execution by at least one of the one or more processors via at least one of the one or more memories, the stored program instructions comprising:
program instructions, responsive to detecting geospatial information and temporal information associated with a mention of an entity in a document, to convert the geospatial information into a specified geospatial format and the temporal information into a specified temporal format;
program instructions to compute one or more prefix-based geospatial values for the converted geospatial information and one or more prefix-based temporal values for the converted temporal information; and
program instructions encode an entry in a knowledge graph for the mention of the entity from the document with the one or more prefix-based geospatial values and the one or more prefix-based temporal values, wherein each digit of the one or more prefix-based geospatial values and the one or more prefix-based temporal values in the knowledge graph that matches a digit of another one or more prefix-based geospatial values and another one or more prefix-based temporal values in another entry for another entity mention in the knowledge graph reflects a degree of granularity of geotemporal proximity of the entity and the another entity.

9. The computer system according to claim 8 , the program instructions to compute one or more prefix-based geospatial values for the converted geospatial information and one or more prefix-based temporal values for the converted temporal information further comprising:
program instructions to compute a prefix-based geospatial hash value by applying a hash function to the converted geospatial information; and
program instructions to compute a prefix-based geospatial offset hash value by rotating a latitude and longitude of the prefix-based geospatial hash value by a rotation value, wherein each digit of the prefix-based geospatial hash value and prefix-based geospatial offset hash value that matches another prefix-based geospatial hash value and another prefix-based geospatial offset hash value in the ano

Assignees

• IBM

Inventors

• Beller Charles E
• Katz Edward G
• Purdy Michael
• Behrens Jr Richard

Classifications

• G06F16/909

  using geographical or spatial information, e.g. location (spatiotemporally dependent retrieval from the web G06F16/9537) · CPC title

• G06N5/022

  Knowledge engineering; Knowledge acquisition · CPC title

• G06N20/00

  Machine learning · CPC title

• G06F16/9024 Primary

  Graphs; Linked lists (G06F16/9027 takes precedence) · CPC title

• G06F16/387

  using geographical or spatial information, e.g. location · CPC title