Using question answering (qa) systems to identify answers and evidence of different medium types

Abstract

First, a computer may receive an input query of a first medium type. The input query may then be analyzed. Based on the analysis, the input query may be categorized as being associated with at least a second medium type. A first-medium-type search of a set of corpora may then be performed. Based on the results of the first-medium-type search, a candidate answer of the first medium type may be generated. In response to the categorizing, a second-medium-type search of the set of corpora may also be performed.

What is claimed is: 
     
         1 . A method comprising:
 receiving an input query of a first medium type;   analyzing, by a processor, the input query;   categorizing, based on the analyzing, the input query as being associated with at least a second medium type;   performing a first-medium-type search of a set of corpora; and   performing, in response to the categorizing, a second-medium-type search of the set of corpora.   
     
     
         2 . The method of  claim 1  further comprising:
 generating, as a result of the first-medium-type search, a candidate answer of the first medium type; 
 calculating a confidence score for the candidate answer; and 
 determining that the confidence score is below a threshold confidence score, wherein the performing the second-medium-type search occurs in response to both the determining and the categorizing. 
 
     
     
         3 . The method of  claim 1 , wherein the input query is created by a user, the method further comprising:
 generating, as a result of the second-medium-type search, a candidate answer; and   presenting the candidate answer to the user.   
     
     
         4 . The method of  claim 3 , wherein the candidate answer is of the first medium type. 
     
     
         5 . The method of  claim 3 , wherein the candidate answer is of the second medium type. 
     
     
         6 . The method of  claim 3 , wherein the first medium type is a textual type, and wherein the second medium type is one of the group consisting of an image type, an audio type, and a video type. 
     
     
         7 . The method of  claim 6 , wherein the set of corpora includes a set of source items of the second medium type, and wherein the performing the second-medium-type search comprises:
 extracting data from the set of source items; and   processing the extracted data.   
     
     
         8 . The method of  claim 7 , wherein the extracted data includes non-textual extracted data. 
     
     
         9 . The method of  claim 8 , wherein the extracted data includes textual extracted data. 
     
     
         10 . The method of  claim 1 , wherein the set of corpora includes a first corpus and a second corpus, wherein the performing the first-medium-type search is a search of the first corpus, and wherein the performing the second-medium-type search is a search of the second corpus. 
     
     
         11 . The method of  claim 1  further comprising:
 generating, as a result of the first-medium-type search, a candidate answer of the first medium type; 
 analyzing the candidate answer based a plurality of factors; and 
 determining, based on the analyzing the candidate answer, that the candidate answer is unacceptable, wherein the performing the second-medium-type search occurs in response to both the determining and the categorizing. 
 
     
     
         12 . The method of  claim 11 , wherein the plurality of factors include at least one of the group consisting of a level of confidence in the accuracy of the candidate answer and a cost of the performing the second-medium-type search. 
     
     
         13 . The method of  claim 1 , wherein the at least the second medium type includes the second medium type and a third medium type, the method further comprising:
 performing a third-medium-type search of the set of corpora.   
     
     
         14 . The method of  claim 1 , wherein the categorizing the input query as being associated with at least a second medium type occurs in response to determining that at least a portion of the input query represents a concept encompassed within an ontology of the second medium type. 
     
     
         15 . The method of  claim 1 , wherein the first medium type is a textual type, and wherein the analyzing the input query includes:
 parsing, by a natural language processing technique configured to analyze syntactic and semantic content, the input query.   
     
     
         16 . The method of  claim 1 , wherein the first medium type is a textual type, wherein the second medium type is an image type, wherein the set of corpora include a textual source item and further include an image-type source item, wherein the performing the first-medium-type search includes extracting a first textual data from the textual source item and further includes processing the extracted first textual data, wherein the performing the second-medium-type search includes extracting second textual data from the image-type source item and further includes processing the extracted second textual data, the method further comprising:
 generating, based on the processing the extracted first textual data, a textual candidate answer;   calculating a confidence score for the textual candidate answer;   determining that the confidence score is below a threshold confidence score, wherein the performing the second-medium-type search occurs in response to both the determining and the categorizing;   generating, based on the processing the extracted second textual data, an image-type candidate answer including the image-type source item; and   presenting the image-type candidate answer to the user.