Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Method And System For Managing Cases When An Answer Changes In A Question Answering System

IP.com Disclosure Number: IPCOM000237471D
Publication Date: 2014-Jun-18
Document File: 2 page(s) / 63K

Publishing Venue

The IP.com Prior Art Database

Abstract

A method and system for managing cases when an answer changes in a question answering (QA) system is disclosed.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 01 of 2

Method And System For Managing Cases When An Answer Changes In A Question Answering System

Disclosed is a method and system for managing cases when an answer changes in a question answering (QA) system.

QA systems provides a different answer over time, as more data becomes available and the QA system is better trained. In this case, users need to know when they will get new answers to existing questions. Polling the system on a simple interval (say daily) is too resource intensive for the QA system. It's important to know when to expect a new answer from the QA system. The rapid growth in information and knowledge requires improvements in predicting when an answer changes. The disclosed method and embodiments herein identify mechanisms for dynamically managing cases where the correct answers in QA systems change. Methods and systems are provided that dynamically manage past answers and the users that queried the systems.

The users of a QA system are informed when an answer to one of their questions has changed because of an update to the QA system (new corpus, better model, etc). Rather than naively poll the system with the same questions, the system analyzes the questions and the QA system training results (between version n and n+1) to determine when the QA system has improved at answering questions in a given domain. The system only "re-ask" questions with a reasonable likelihood of having a new answer, thus saving considerable system load.

In a preferred embodiment, users register a list of questions with the QA system, to be notified when the answer has changed. When the corpus/training of the system is updated, the system dynamically determines which questions may now have new answers based on changes in the corpus/training of the system. For questions with updated answers, the system notifies users of the new answer.

IMPLEMENTATION DETAILS
Users register one or more questions with the system


1.

(called Watson in this article) oncology question (health

  The user may select a domain/topic from a list made available by the system
The system may dynamically determine the domain/topic by one of several met...