Acquisition of anecdotes from the web for free conversation systems

Abstract

A free conversation system that can talk freely with users gives computers human-like qualities that can enrich our daily lives. This paper proposes a method to automatically acquire anecdotes from various people on the Internet as a source of knowledge to be accessed by free conversation systems. Firstly, passages that are related to a given person are retrieved by searching the relevant web pages. Secondly, each page is segmented and then, thirdly, each passage is judged as to whether it is an anecdote by applying several rules based on the linguistic features of the anecdotes. The experimental results showed that the precision of our method was increased by 25% when compared to the baseline.