Query-relevant summarization using FAQs

Abstract

This paper introduces a statistical model for query-relevant summarization: succinctly characterizing the relevance of a document to a query. Learning parameter values for the proposed model requires a large collection of summarized documents, which we do not have, but as a proxy, we use a collection of FAQ (frequently-asked question) documents. Taking a learning approach enables a principled, quantitative evaluation of the proposed system, and the results of some initial experiments-on a collection of Usenet FAQs and on a FAQ-like set of customer-submitted questions to several large retail companies-suggest the plausibility of learning for summarization.

Evaluating Web Search Result Summaries

“…Table 4 shows the summary quality results of the QTO and Google systems, where R represents the Representativeness score, J represents the Judgeability score and SQ represents the Summary Quality score. We used the data from Table 2 to convert the values in columns J by applying Formula (2). The values in R are also converted by using data in Table 1 and divided by 10 subjects.…”

“…This is also the reason for setting both R J and to between 0 and 1. Therefore, the summary's quality-SQ is averaged as formula (3) by the sum of formula (1) and formula (2).…”

“…The metrics we employ are representativeness and judgeability, and we combine these measures to arrive at a third metric: summary quality. The related work of Berger and Mittal [2] states that query relevant summaries should include fidelity and relevance. It is important to note, however, that while our notion of representativeness equates to fidelity, we do not measure relevance but rather the user's ability to judge relevance or irrelevance.…”

A Comparative Study of Information-Gathering Approaches for Answering Help-Desk Email Inquiries

“…Our work differs from these two examples in that it applies a purely data-driven approach that uses a corpus of request-response pairs to generate a single response. Similar corpus-based approaches have been implemented in FAQ domains [10,11]. However, this kind of corpus is significantly different to ours in that it lacks repetition and redundancy and the responses are not personalized.…”

A noisy-channel model for document compression

“…Headline generators are noisy-channel probabilistic systems that are trained on large corpora of ¢ H eadline, Text£ pairs (Banko et al, 2000;Berger and Mittal, 2000). These systems produce short sequences of words that are indicative of the content of the text given as input.…”