Graph-based event schema induction for information retrieval

Patent metadata

Field	Value
Publication number	US-11615152-B2
Application number	US-202117223774-A
Country	US
Kind code	B2
Filing date	Apr 6, 2021
Priority date	Apr 6, 2021
Publication date	Mar 28, 2023
Grant date	Mar 28, 2023

Abstract

Systems, devices, computer-implemented methods, and/or computer program products that facilitate event schema induction from unstructured or semi-structured data. In one example, a system can comprise a processor that executes computer executable components stored in memory. The computer executable components can comprise a schema component and a retrieval component. The schema component can derive an event schema for a document corpus using parsing results obtained from the document corpus. The retrieval component can populate a response to a query with a document of the document corpus using events extracted from the query and the document using the event schema.

First claim

What is claimed is:

1. A system, comprising:
a memory that stores computer executable components; and
a processor that executes the computer-executable components stored in memory, wherein the computer executable components comprise:
a schema component that derives an event schema for a document corpus using parsing results obtained from the document corpus; and
a retrieval component that populates a response to a query with a document of the document corpus using events extracted from the query and the document using the event schema, wherein the events extracted from the query and the document include an extracted event that comprises a list of tuple representations, and wherein a tuple representation in the list of tuple representations is a vector formed by concatenating respective vector representations of an event type, an event trigger, an event argument, and an argument role.

2. The system of claim 1 , further comprising:
an extraction component that extracts the events from the query and the document of the document corpus using the event schema.

3. The system of claim 2 , wherein the extraction component assigns a weight to an extracted event based on a usage frequency of the extracted event by the retrieval component, a context in which the extracted event appears, or a combination thereof.

4. The system of claim 1 , wherein the schema component derives the event schema for the document corpus by identifying candidate event triggers and event arguments from the parsing results to form proto-events.

5. The system of claim 1 , wherein the schema component derives the event schema for the document corpus by generating vector representations of events using a graph neural network.

6. The system of claim 1 , wherein the schema component derives the event schema for the document corpus by clustering vector representations of events into a plurality of clusters to identify event types.

7. The system of claim 1 , further comprising:
a feedback component that adjusts the event schema based on feedback data obtained from usage logs.

8. The system of claim 1 , wherein the parsing results are obtained using a parser.

9. A computer-implemented method, comprising:
deriving, by a system operatively coupled to a processor, an event schema for a document corpus using parsing results obtained from the document corpus; and
populating, by the system, a response to a query with a document of the document corpus using events extracted from the query and the document using the event schema, wherein the events extracted from the query and the document include an extracted event that comprises a list of tuple representations, and wherein a tuple representation in the list of tuple representations is a vector formed by concatenating respective vector representations of an event type, an event trigger, an event argument, and an argument role.

10. The computer-implemented method of claim 9 , further comprising:
extracting, by the system, the events from the query and the document of the document corpus using the event schema.

11. The computer-implemented method of claim 9 , wherein the system derives the event schema for the document corpus by identifying candidate event triggers and event arguments from the parsing results to form proto-events.

12. The computer-implemented method of claim 9 , wherein the system derives the event schema for the document corpus by generating vector representations of events using a graph neural network.

13. The computer-implemented method of claim 9 , wherein the system derives the event schema for the document corpus by clustering vector representations of events into a plurality of clusters to identify event types.

14. The computer-implemented method of claim 9 , further comprising:
adjusting, by the system, the event schema based on feedback data obtained from usage logs.

15. A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:
derive, by the processor, an event schema for a document corpus using parsing results obtained from the document corpus; and
populate, by the processor, a response to a query with a document of the document corpus using events extracted from the query and the document using the event schema, wherein the events extracted from the query and the document include an extracted event that comprises a list of tuple representations, and wherein a tuple representation in the list of tuple representations is a vector formed by concatenating respective vector representations of an event type, an event trigger, an event argument, and an argument role.

16. The computer program product of claim 15 , wherein the processor derives the event schema for the document corpus by identifying candidate event triggers and event arguments from the parsing results to form proto-events.

17. The computer program product of claim 15 , wherein the processor derives the event schema for the document corpus by generating vector representations of events using a graph neural network.

18. The computer program product of claim 15 , wherein the processor derives the event schema for the document corpus by clustering vector representations of events into a plurality of clusters to identify event types.

19. The computer program product of claim 15 , wherein the processor adjusts the event schema based on feedback data obtained from usage logs.

20. The computer program product of claim 15 , wherein the parsing results are obtained using a parser.

Assignees

• IBM
• Univ Illinois

Inventors

• Haldar Rajarshi
• Deng Yu
• Wu Lingfei
• Mahindru Ruchi
• Hockenmaier Julia Constanze
• Kaya Sinem Guven

Classifications

• G06N3/0895

  Weakly supervised learning, e.g. semi-supervised or self-supervised learning · CPC title

• G06N3/0475

  Generative networks · CPC title

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06N3/091

  Active learning · CPC title