Semantic parsing for complex knowledge extraction

Patent metadata

Field	Value
Publication number	US-2016275073-A1
Application number	US-201514664760-A
Country	US
Kind code	A1
Filing date	Mar 20, 2015
Priority date	Mar 20, 2015
Publication date	Sep 22, 2016
Grant date	—

Abstract

The system that performs semantic parsing may automatically extract complex information from databases. Complex information may comprise nested event structures. In one example process, a processor may receive unannotated text and may access a natural-language database that includes nested events. The processor, in performing semantic parsing, may automatically generate syntactic trees that include annotations that represent the semantic information. In particular, the natural-language sentences and the database include nested event structures.

First claim

What is claimed is:

1 . A system comprising:
one or more processors; and memory storing instructions that, when executed by the one or more processors, configure the one or more processors to perform operations comprising: receiving one or more syntactic dependency trees; accessing a database that includes nested events; and generating an event table based, at least in part, on (i) the one or more syntactic dependency trees and (ii) the nested events, wherein the generated event table includes one or more nested relationships.

2 . The system of claim 1 , wherein the syntactic dependency trees are derived from unannotated text, and wherein generating the event table is based, at least in part, on the unannotated text.

3 . The system of claim 1 , wherein generating the event table further comprises:
generating a probability distribution as a function of (i) one or more of the nested events of the database and (ii) latent variables; and determining a set of values of the latent variables with maximum probability.

4 . The system of claim 3 , wherein generating the event table further comprises:
generating additional probability distributions that are respectively functions of different sets of (i) one or more of the nested events of the database and (ii) the latent variables; and biasing each of the additional probability distributions with respective weighting factors.

5 . The system of claim 4 , wherein the respective weighting factors comprise virtual evidence.

6 . The system of claim 1 , wherein the event table comprises two or more nested events.

7 . The system of claim 1 , wherein the event table is generated using distant supervision.

8 . A computing device comprising:
an input port to receive a collection of sentences with syntactic dependency trees; and a processor to:
access a database that includes nested events; and
based, at least in part, on the sentences and the nested events, assign probabilities or scores to syntactic trees annotated with semantic annotations.

9 . The computing device of claim 8 , wherein the processor is configured to:
train a statistical model that assigns the probabilities or the scores to the syntactic trees annotated with the semantic annotations.

10 . The computing device of claim 9 , wherein the processor is configured to:
generate an event table based, at least in part, on the statistical model.

11 . The computing device of claim 10 , wherein the event table comprises two or more nested events.

12 . The computing device of claim 8 , wherein the processor is configured to:
determine emission parameters and/or transition parameters for a joint model of (i) the syntactic dependency trees and (ii) semantic trees by maximizing likelihoods of the syntactic dependency trees.

13 . The computing device of claim 10 , wherein the likelihood of the syntactic dependency trees with the semantic annotations is augmented by the database that includes the nested events.

14 . Computer-readable storage media of a client device storing computer-executable instructions that, when executed by one or more processors of the client device, configure the one or more processors to perform operations comprising:
receiving one or more syntactic dependency trees; accessing a database comprising a plurality of events that include nested relationships; and based, at least in part, on the nested relationships, generating one or more syntactic trees annotated with semantic annotations.

15 . The computer-readable storage medium of claim 14 , the operations further comprising:
generating an event table representing the one or more syntactic trees annotated with the semantic annotations, wherein the generated event table includes one or more nested relationships.

16 . The computer-readable storage medium of claim 14 , the operations further comprising:
counting the number of times that a particular syntactic tree annotated with particular semantic annotations is generated.

17 . The computer-readable storage medium of claim 14 , wherein the one or more syntactic dependency trees are derived from unannotated text.

18 . The computer-readable storage medium of claim 14 , the operations further comprising:
receiving a search query relating to a knowledge domain of the database in a natural or a formal language; and retrieving answers for the search query based, at least in part, on (i) information in the database, and (ii) a derived semantic representation of the query.

19 . The computer-readable storage medium of claim 14 , the operations further comprising:
receiving a search query comprising a nested search term; and comparing the search query with the one or more syntactic trees annotated with the semantic annotations.

20 . The computer-readable storage medium of claim 14 , wherein the one or more syntactic trees annotated with the semantic annotations is generated using distant supervision.

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Poon Hoifung
• Toutanova Kristina
• Parikh Ankur P

Classifications

• G06F40/30 Primary

  Semantic analysis · CPC title

• G06F40/211 Primary

  Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars · CPC title

• G06F17/271

  Physics · mapped topic

• G06F17/2785 Primary

  Physics · mapped topic