Future of QA in the Business Environment Yael Ravin T. J. Watson Research Center, IBM
Current status QA markets itself –Natural interface –Succinct –Easy to assess success But this can backfire –difference b/w QA and full NLP understanding (ellipsis, dialog context) –continuum from single facts to judgements
Market for QA: Search For employees in the company –increase productivity "What is the meal limit for Toulouse?" –training "When should I open the emergency door?" –support innovation "What new sedan features were introduced in the last 6 months?"
Market for QA: Search (continued) For consumers –online shopping "Do you sell mice with no cords?" –online support "How do I attach a printer to my ThinkPad 570?" Between companies –match making "Who does PR for small theater productions?"
Integrate QA within other products Work within different business domains and user needs: –shopping, expertise locator, business intelligence,... Integrate with other functionalities –a generic search engine –a dialog system Operate with multiple sources of info: –databases, text, presentations, business rules
Issues for QA in business Develop case scenarios with partners Prototype QA for applications and learn what is useful Develop measures: –performance - simultaneous users –what is good enough for customers to use –cost of customization –one time and on-going –linguists vs.domain experts