Generating natural language dialog using a questions corpus

Patent metadata

Field	Value
Publication number	US-10049152-B2
Application number	US-201514864057-A
Country	US
Kind code	B2
Filing date	Sep 24, 2015
Priority date	Sep 24, 2015
Publication date	Aug 14, 2018
Grant date	Aug 14, 2018

Abstract

Generating a natural language dialog by finding missing semantic information in a user question by comparing it to the closest question available in a question corpus. Incrementally improved question precision is targeted during each round of the natural language dialog by generating follow-up questions that clarify semantic and syntactic characteristics of the user question. The follow-up questions are derived from analysis of the user question to identify areas of improvement on the user question.

First claim

What is claimed is:

1. A method comprising:
receiving a target question in a natural language dialog from a human user;
computing a set of target syntax features based on the target question;
identifying a first set of reference syntax features based on a first reference question;
identifying a second set of reference syntax features based on a second reference question;
selecting the first set of reference syntax features based, at least in part, on the set of target syntax features being more similar to the first set of reference syntax features when compared to the second set of reference syntax features;
determining a set of missing syntax features based at least in part on a comparison of the set of target syntax features and the first set of reference syntax features;
computing a first weight for a first missing syntax feature of the set of missing syntax features and a second weight for a second missing syntax feature of the set of missing syntax features;
responsive to the first weight being higher than the second weight, determining a follow-up question based at least in part on a semantic analysis of the target question and the first missing syntax feature; and
displaying, by a computer, the follow-up question to the human user.

2. The method of claim 1 , wherein the set of target syntax features, the first set of reference syntax features, and the second set of reference syntax features are each in the form of a parse tree.

3. The method of claim 1 , wherein selecting the first set of reference syntax features includes:
determining a first count of reference syntax features for the first set of reference syntax features that match at least one target syntax feature of the set of target syntax features; and
determining a second count of reference syntax features for the second set of reference syntax features that match at least one target syntax feature of the set of target syntax features;
wherein:
the set of target syntax features are more similar to the first set of reference syntax features than the second set of reference syntax features at least because the first count of reference syntax features is greater than the second count of reference syntax features.

4. The method of claim 1 , wherein selecting the first set of reference syntax features includes:
determining a first count of missing reference syntax features for the first set of reference syntax features that are not present in the set of target syntax features; and
determining a second count of missing reference syntax features for the second set of reference syntax features that are not present in the set of target syntax features;
wherein:
the set of target syntax features are more similar to the first set of reference syntax features than the second set of reference syntax features at least because the first count of missing reference syntax features is smaller than the second count of missing reference syntax features.

5. The method of claim 1 , wherein:
the first weight is determined according to a pre-determined weight of a first word type corresponding to the first missing syntax feature; and
and the second weight is determined according to a pre-determined weight of a second word type corresponding to the second missing syntax feature.

6. The method of claim 5 , wherein the word type is selected from the group consisting of a noun, a determiner, a pronoun, a verb, an adjective, an adverb, a preposition, and a conjunction.

7. The method of claim 1 , wherein:
the first reference question and the second reference question are identified in a historical user query store as well-formulated questions; and
the historical user query store includes questions previously received from users including the first reference question and the second reference question.

8. A computer program product comprising a non-transitory computer-readable storage medium having a set of instructions stored therein which, when executed by a processor, causes the processor to perform the following steps:
receiving a target question in a natural language dialog from a human user;
computing a set of target syntax features based on the target question;
identifying a first set of reference syntax features based on a first reference question;
identifying a second set of reference syntax features based on a second reference question;
selecting the first set of reference syntax features based, at least in part, on the set of target syntax features being more similar to the first set of reference syntax features when compared to the second set of reference syntax features;
determining a set of missing syntax features based at least in part on a comparison of the set of target syntax features and the first set of reference syntax features;
computing a first weight for a first missing syntax feature of the set of missing syntax features and a second weight for a second missing syntax feature of the set of missing syntax features;
responsive to the first weight being higher than the second weight, determining a follow-up question based at least in part on a semantic analysis of the target question and the first missing syntax feature; and
displaying the follow-up question to the human user.

9. The computer program product of claim 8 , wherein the set of target syntax features, the first set of reference syntax features, and the second set of reference syntax features are each in the form of a parse tree.

10. The computer program product of claim 8 , wherein selecting the first set of reference syntax features includes:
determining a first count of reference syntax features for the first set of reference syntax features that match at least one target syntax feature of the set of target syntax features; and
determining a second count of reference syntax features for the second set of reference syntax features that match at least one target syntax feature of the set of target syntax features;
wherein:
the set of target syntax features are more similar to the first set of reference syntax features than the second set of reference syntax features at least because the first count of reference syntax features is greater than the second count of reference syntax features.

11. The computer program product of claim 8 , wherein selecting the first set of reference syntax features includes:
determining a first count of missing reference syntax features for the first set of reference syntax features that are not present in the set of target syntax features; and
determining a second count of missing reference syntax features for the second set of reference syntax features that are not present in the set of target syntax features;
wherein:
the set of target syntax features are more similar to the first set of reference syntax features than the second set of reference syntax features at least because the first count of missing reference syntax features is smaller than the second count of missing reference syntax features.

12. The computer program product of claim 8 , wherein:
the first weight is determined according to a pre-determined weight of a first word type corresponding to the first missing syntax feature; and
and the second weight is determined according to a pre-determined weight of a second word type corresponding to the second missing syntax feature.

13. The computer program product of claim 12 , wherein the word type is selected from the group consisting of a noun, a determiner, a pronoun, a verb, an adjective, an adverb, a preposition, and a conjunction.

14. A computer system comprising:
a processor set; and
a computer readable storage med

Assignees

• IBM

Inventors

• Ajmera Jitendra
• Gupta Ajay K
• Joshi Sachindra
• Toshniwal Shubham

Classifications

• G06F40/30

  Semantic analysis · CPC title

• G06F40/211

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

• G06F16/3329 Primary

  Natural language query formulation · CPC title

• G06F17/30654 Primary

  Physics · mapped topic

• G06F17/2785

  Physics · mapped topic