Electrical Engineering
      and Computer Sciences

Electrical Engineering and Computer Sciences


UC Berkeley


2008 Research Summary

From Search Engines to Question-Answering Systems--A Challenge that Is Hard to Meet

View Current Project Information

Lotfi A. Zadeh

Existing search engines, with Google at the top, have many truly remarkable capabilities. Furthermore, constant progress is being made in improving their performance. But what is not widely recognized is that there is a basic capability which existing search engines do not have: deduction capability--the capability to synthesize an answer to a query by drawing on bodies of information which reside in various parts of the knowledge base. By definition, a question-answering system is a system which has deduction capability. Can a search engine be upgraded to a question-answering system through the use of existing tools--tools which are based on bivalent logic and probability theory? A view which is articulated in the following is that the answer is: no.

There are three major obstacles: (a) world knowledge; (b) relevance; and (c) deduction. The problem with world knowledge is that in large measure it is perception-based and hence is intrinsically imprecise. Example: usually it does not rain in San Francisco in midsummer. Perception-based information does not lend itself to manipulation through the use of bivalent logic and probability theory.

The problem with relevance is that existing approaches to assessment of relevance attempt to deal with relevance in a statistical framework, with no consideration of semantics. The results leave much to be desired.

The problem with deduction is that in realistic settings the premises are generally imprecise, uncertain, and partially true. In such settings, conventional methods of deduction do not work.

To deal with the problems of world knowledge, assessment of relevance, and deduction, new tools are needed. The new tools which are outlined in my work are Precisiated Natural Language (PNL), Protoform Theory (PFT), and Generalized Theory of Uncertainty (GTU). The centerpiece of these tools is the concept of a generalized constraint. The concept of a generalized constraint is what makes it possible to deal effectively with information which is imprecise, uncertain, incomplete, and partially true.