Design considerations for agent systems that glean the Internet

“If software agents are such a promising technology, why can´t they do something simple for me, like find a movie?” This comment is proffered as a sort of litmus test for agent capability. Interestingly, the same question is also the one most often asked in information extraction (IE) circles. The assumption is that the suggested problem is relatively easy for a human to solve: if software agents can´t solve it, then they aren´t particularly useful. The authors address design considerations for Internet IE using a familiar domain: finding movies online. Although information about available movie showings is easily available by phone, browser, and daily and weekend newspapers, we discuss several reasons why choosing movies, theaters, and show times is a fundamentally tough IE and information fusion problem (even for a human)