Inter Thread Anaphora Resolution

Abstract

An approach is provided to resolve anaphors between posts, or threads, in a threaded discussion, for example an online forum. The approach analyzes a number of posts that are included in threads of an online forum. During the analysis, the approach identifies terms in parent posts, detects anaphors in child posts that reference the terms in the parent posts, and resolves the anaphor found in the child post with the term. The parent post with the identified term and the child post with the resolved anaphor are then stored in the memory for use by information handling systems, such as question answering (QA) systems.

1 . A method implemented by an information handling system that includes a memory and a processor, the method comprising:
 analyzing a plurality of posts included in one or more threads of a threaded discussion, wherein the analyzing further comprises:
 identifying a term in a parent post of the threaded discussion; 
 detecting that an anaphor in a child post of the threaded discussion references the identified term; and 
 resolving the anaphor found in the child post with the identified term; and 
   storing the parent post with the identified term and the child post with the resolved anaphor in the memory.   
     
     
         2 . The method of  claim 1  further comprising:
 ingesting the parent post with the identified term and the child post with the resolved anaphor into a corpus utilized by a question answering (QA) system. 
 
     
     
         3 . The method of  claim 2  further comprising:
 identifying one or more referential types corresponding to one or more words included in the parent post; 
 identifying one or more matching anaphora types corresponding to the identified referential types in one or more child posts, wherein the child posts include the child post; 
 detecting the anaphors in the one or more child posts that relate to one or more of the words included in the parent post based upon matching the referential types to the anaphora types; 
 resolving each of the anaphors detected in the child posts with the corresponding words found in the parent post; and 
 associating each of the child posts that include one or more anaphors relating to corresponding words in the parent post to the parent post, wherein the association is accessible by the QA system. 
 
     
     
         4 . The method of  claim 2  further comprising:
 identifying referential data in the child post and the parent post, wherein at least one of the referential data is selected from the group consisting of domain, question, focus, concept, statements, and a lexical answer type (LAT); and 
 storing the referential data in the corpus utilized by the QA system. 
 
     
     
         5 . The method of  claim 4  further comprising:
 receiving a question at the QA system; 
 identifying, by the QA system, one or more candidate answers, wherein at least one of the candidate answers is derived from the child post with one or more anaphors relating to the corresponding words in the parent post; and 
 responding with at least one of the candidate answers. 
 
     
     
         6 . The method of  claim 1  further comprising:
 identifying a referential type of the identified term; 
 identifying a matching anaphora type corresponding to the anaphor; 
 matching the referential term to an anaphora type to detect that the anaphor references the identified term. 
 
     
     
         7 . The method of  claim 1  further comprising:
 analyzing each of a plurality of posts in the threaded discussion, wherein the plurality of posts include the child post and the parent post; 
 identify any referential types corresponding to a plurality of words included in each of the posts; 
 identify any anaphora types corresponding to a plurality of words included in each of the posts; 
 associating each of a plurality of child posts with at least one parent post as a relationship; 
 resolving the anaphora types included in the child posts with at least one of the referential types included in the respective associated parent posts; and 
 building a thread tree corresponding to the threaded discussion, wherein the thread tree includes the plurality of posts, the relationships between posts, and the resolved anaphora types. 
 
     
     
         8 . A method implemented by an information handling system that includes a memory and a processor, the method comprising:
 initializing a forum tree to store data from a thread of a threaded discussion from an on-line forum;   storing, in the forum tree, post data associated with each post included in the thread, wherein the storing further comprises:
 storing any referential types included in each of the posts; 
 resolving any anaphors included in each of the posts with referential types included in a parent post; and 
 storing any resolved anaphor data in each of the posts. 
   
     
     
         9 . The method of  claim 8  further comprising:
 ingesting the forum tree into a corpus utilized by a question answering (QA) system. 
 
     
     
         10 . The method of  claim 8  further comprising:
 identifying a plurality of parent-child relationships between the posts; and 
 storing a plurality of relationship associations pertaining to the relationships in the forum tree. 
 
     
     
         11 . The method of  claim 8  further comprising:
 identifying an anaphora type corresponding to each of the anaphors, wherein the anaphora types are selected from the anaphora types group consisting of a pronoun type anaphor, a fragment type anaphor, and an agreement type anaphor; 
 wherein the referential types are selected from a referential types group consisting of a noun type, a lexical answer type (LAT) type, a statement type, a question type, and a candidate answer type; and 
 matching the anaphora type found in a child post to the referential type found in a parent post, wherein the parent post is a parent to the child post in the forum tree.