Determining User-Friendly Information to Solicit in a Question and Answer System

Patent metadata

Field	Value
Publication number	US-2016306791-A1
Application number	US-201514686860-A
Country	US
Kind code	A1
Filing date	Apr 15, 2015
Priority date	Apr 15, 2015
Publication date	Oct 20, 2016
Grant date	—

Abstract

Mechanisms are provided for processing natural language text. The natural language text is parsed to generate a logical parse which is then analyzed to identify a solicitation node of the logical parse for which a user is able to provide additional data to assist in processing the natural language text. A clarifying question is output to a computing device associated with a user based on content of the solicitation node and a response to the clarifying question is received from the user. The natural language text is then processed based on the response to the clarifying question received from the user to thereby generate a result. The result of the processing of the natural language text is output to a source of the natural language text.

First claim

What is claimed is:

1 . A method, in a data processing system comprising a processor and a memory, for processing an input natural language text, the method comprising:
parsing, by the data processing system, the input natural language text to generate a logical parse data structure representing the input natural language text; analyzing, by the data processing system, the logical parse data structure to identify at least one solicitation node of the logical parse data structure for which a user is able to provide additional data to assist in processing the input natural language text; outputting, by the data processing system, a clarifying question to a computing device associated with the user based on content of the at least one solicitation node; receiving, by the data processing system, a response to the clarifying question from the user via the computing device; processing, by the data processing system, the natural language text based on the response to the clarifying question received from the user via the computing device to thereby generate a result of the processing of the input natural language text; and outputting, by the data processing system, the result of the processing of the natural language text to a source of the input natural language text.

2 . The method of claim 1 , wherein analyzing the logical parse data structure to identify the solicitation node comprises:
associating a strength value with each edge, in a plurality of edges of the logical parse data structure, based on one or more characteristics of nodes associated with the edge; selecting a subset of nodes in the logical parse data structure based on the strength values associated with the edges in the plurality of edges; and selecting the at least one solicitation node from the subset of nodes based on an evaluation of characteristics of the subset of nodes.

3 . The method of claim 2 , wherein associating a strength value with each edge comprises, for each edge, calculating the strength value of the edge based on an evaluation of at least one of a depth of a corresponding node associated with the edge, a type of logical operator connecting the corresponding node to other nodes in the logical parse data structure, a relevance value associated with the corresponding node, or a supporting evidence value associated with the node.

4 . The method of claim 2 , wherein selecting the subset of nodes comprises selecting nodes in the logical parse data structure that are associated with edges having strength values that are equal to or greater than a threshold strength value.

5 . The method of claim 2 , wherein selecting the at least one solicitation node from the subset of nodes based on an evaluation of characteristics of the subset of nodes comprises analyzing at least one of content of nodes in the subset of nodes or a structure of a portion of the logical parse data structure associated with the nodes in the subset of nodes.

6 . The method of claim 2 , wherein selecting the at least one solicitation node from the subset of nodes based on an evaluation of characteristics of the subset of nodes comprises analyzing content of each node in the subset of nodes based on an indicator data structure that identifies indicators in annotations of the nodes or textual content of the nodes that is indicative of a node for selection, wherein in response to a node having an indicator matching an indicator in the indicator data structure in at least one of the annotations of the node or textual content of the node, the node is selected as a solicitation node of the at least one solicitation node.

7 . The method of claim 2 , wherein selecting the at least one solicitation node from the subset of nodes based on an evaluation of characteristics of the subset of nodes comprises analyzing a hierarchical structure of a portion of the logical parse data structure associated with the nodes in the subset of nodes, wherein in response to a node having a predetermined associated structure within the logical parse data structure, the node is selected as a solicitation node of the at least one solicitation node.

8 . The method of claim 7 , wherein the predetermined associated structure comprises at least one of the node being a low level leaf node in the logical parse data structure or the node having at least a predetermined number of child nodes.

9 . The method of claim 1 , wherein analyzing the logical parse data structure to identify the at least one solicitation node comprises:
filtering out first nodes in the logical parse data structure that have a zero relevance value from further processing for identifying the node; filtering out second nodes in the logical parse data structure that have a supporting evidence value equal to or higher than a threshold supporting evidence value, from further processing for identifying the node; and selecting the at least one solicitation node from nodes remaining in the logical parse data structure after filtering out the first and second nodes.

10 . The method of claim 1 , wherein processing the natural language text based on the response to the clarifying question received from the user via the computing device to thereby generate a result of the processing of the input natural language text comprises updating the logical parse data structure based on the response, wherein the logical parse data structure is updated by at least one of modifying a characteristic of at least one existing node in the logical parse data structure, adding an additional node to the logical parse data structure, or removing a node in the logical parse data structure from further processing by the data processing system when processing the natural language text.

11 . The method of claim 1 , wherein the data processing system is a Question and Answer (QA) system and the input natural language text is a natural language question input to the QA system for generation of an answer to the natural language question, and wherein the result of the processing of the input natural language text is an answer to the natural language question.

12 . A computer program product comprising a computer readable storage medium having a computer readable program stored therein, wherein the computer readable program, when executed on a data processing system, causes the data processing system to:
parse an input natural language text to generate a logical parse data structure representing the input natural language text; analyze the logical parse data structure to identify at least one solicitation node of the logical parse data structure for which a user is able to provide additional data to assist in processing the input natural language text; output a clarifying question to a computing device associated with the user based on content of the at least one solicitation node; receive a response to the clarifying question from the user via the computing device; process the natural language text based on the response to the clarifying question received from the user via the computing device to thereby generate a result of the processing of the input natural language text; and output the result of the processing of the natural language text to a source of the input natural language text.

13 . The computer program product of claim 12 , wherein the computer readable program further causes the data processing system to analyze the logical parse data structure to identify the solicitation node at least by:
associating a strength value with each edge, in a plurality of edges of the logical parse data structure, based on one or more characteristics of nodes associated with the edge; select

Assignees

• IBM

Inventors

• Allen Corville O
• Bruno Nicholas V
• Smythe Jared M D
• Will Scott A
• Woods Eric

Classifications

• G06F16/3329 Primary

  Natural language query formulation · CPC title

• G06F40/205 Primary

  Parsing · CPC title

• G06F17/2705

  Physics · mapped topic

• G06F17/28 Primary

  Physics · mapped topic