A Natural Language Dialog System Based on Active Ontologies

Begutachtete Veröffentlichung in Tagungsband

Autor(en):Alexander Wachtel, Jonas Klamroth und Walter F. Tichy
In:The Ninth International Conference on Advances in Computer-Human Interactions (ACHI 2016)


  author    = {Alexander Wachtel and Jonas Klamroth and {Walter F.} Tichy},
  title     = {A Natural Language Dialog System Based on Active Ontologies},
  year      = {2016},
  month     = apr,
  booktitle = {The Ninth International Conference on Advances in
               Computer-Human Interactions ({ACHI} 2016)},
  eventdate = {2016-04-24/2016-04-28},
  venue     = {Venive, Italy},
  abstract  = {Programming today requires years of training.
               With natural language, programming would become
               available to everyone and enable end users to
               program their devices or extend their functionality
               without any knowledge of programming languages.
               We present an assistant usable in technical domains
               that uses natural language understanding, programming
               stepby-step and an active dialog management system.
               It allows users to manipulate spreadsheet data by
               using natural language. We extend our previous system
               with active ontologies. By adding additional information 
               to an ontology, such as a rule evaluation system and a
               fact store, it becomes an execution environment instead
               of just being a representation of knowledge. Sensor nodes
               register certain events and store them in the fact store.
               An evaluation mechanism tests the new facts against the
               existing rules and performs the associated action if one
               or more rules apply to the stored facts. The system also
               handles references to previous results and expressions,
               allowing the construction of complex expressions step-by-step.
               It also creates new formulas by using End-User Programming
               concepts and supports the use of repetitive tasks that
               involve use of conditions and negations.
               An evaluation shows that the active ontology-based approach
               resolves 90\% of the input tasks which is an increase of
               10\% over the pattern matching approach.}