Query Parsing and Exploration (QPE) and Short Text Parsing

Tasks and procedures of QPE

  • Split a query into multiple segments
  • Map segments to entities, types, properties, or property values in the knowledge base (depending on the type of each segment). Please note that disambiguation is needed here
  • Identify the type of each segment
  • Identify the type of the whole query
  • Identify the relations between segments
  • Perform query exploration based on each segment or the whole query
  • Identify the potential user action(s) related to the query

 

Requirements from the query/text parsing module:

Knowledge Type  Usage/Purpose  Priority 
A clean, high-coverage term list  Query segmentation (SEG)  P1 
Basic statistics (e.g., frequency) of terms  Query segmentation  P1 
IsA relation (term-level)  Query segment labeling (LAB)  P1 
Peer similarity; semantic classes  Enhance SEG & LAB  P2 
Term co-occurrence information  Enhance SEG & LAB  P1 
Term transformation information (e.g., singular-plural, verb tense)  Segment normalization  P2 
List of entities with types  Segment relations (REL); enhance LAB  P2 
Mapping of terms to entities: coverage and statistics  Assign entities to segments (S2E)  P1 
Entity properties and facts  Implement REL; enhance LAB & SEG  P1 
Term relations & entity relations  Implement REL; enhance LAB & SEG  P1 
Synonyms of entities, attributes and relations  Improve matching  P2