<?xml version="1.0" standalone="yes"?>
<Paper uid="H05-1121">
  <Title>Proceedings of Human Language Technology Conference and Conference on Empirical Methods in Natural Language Processing (HLT/EMNLP), pages 963-970, Vancouver, October 2005. c(c)2005 Association for Computational Linguistics Query Expansion with the Minimum User Feedback by Transductive Learning</Title>
  <Section position="4" start_page="0" end_page="963" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Query expansion is a simple but very useful technique to improve search performance by adding some terms to an initial query. While many query expansion techniques have been proposed so far, a standard method of performing is to use relevance information from a user (Ruthven, 2003). If we can use more relevant documents in query expansion, the likelihood of selecting query terms achieving high search improvement increases. However it is impractical to expect enough relevance information. Some researchers said that a user usually notifies few relevance feedback or nothing (Dumais and et al., 2003).</Paragraph>
    <Paragraph position="1"> In this paper we investigate the potential performance of query expansion under the condition that we can utilize little relevance information, especially we only know a relevant document and a non-relevant document. To overcome the lack of relevance information, we tentatively increase the number of relevant documents by a machine learning technique called Transductive Learning. Compared with ordinal inductive learning approach, this learning technique works even if there is few training examples. In our case, we can use many documents in a hit-list, however we know the relevancy of few documents. When applying query expansion, we use those increased documents as if they were true relevant ones. When applying the learning, there occurs some difficult problems of parameter settings.</Paragraph>
    <Paragraph position="2"> We also try to provide a reasonable resolution for the problems and show the effectiveness of our proposed method in experiments.</Paragraph>
    <Paragraph position="3"> The point of our query expansion method is that we focus on the availability of relevance information in practical situations. There are several researches which deal with this problem. Pseudo relevance feedback which assumes top n documents as relevant ones is one example. This method is simple and relatively effective if a search engine returns a hit- null list which contains a certain number of relative documents in the upper part. However, unless this assumption holds, it usually gives a worse ranking than the initial search. Thus several researchers propose some specific procedure to make pseudo feedback be effective (Yu and et al, 2003; Lam-Adesina and Jones, 2001). In another way, Onoda (Onoda et al., 2004) tried to apply one-class SVM (Support Vector Machine) to relevance feedback. Their purpose is to improve search performance by using only non-relevant documents. Though their motivation is similar to ours in terms of applying a machine learning method to complement the lack of relevance information, the assumption is somewhat different. Our assumption is to utilizes manual but the minimum relevance judgment.</Paragraph>
    <Paragraph position="4"> Transductive leaning has already been applied in the field of image retrieval (He and et al., 2004). In this research, they proposed a transductive method called the manifold-ranking algorithm and showed its effectiveness by comparing with active learning based Support Vector Machine. However, their setting of relevance judgment is not different from many other traditional researches. They fix the total number of images that are marked by a user to 20.</Paragraph>
    <Paragraph position="5"> As we have already claimed, this setting is not practical because most users feel that 20 is too much for judgment. We think none of research has not yet answered the question. For relevance judgment, most of the researches have adopted either of the following settings. One is the setting of &amp;quot;Enough relevant documents are available&amp;quot;, and the other is &amp;quot;No relevant document is available&amp;quot;. In contrast to them, we adopt the setting of &amp;quot;Only one relevant document is available&amp;quot;. Our aim is to achieve performance improvement with the minimum effort of judging relevancy of documents.</Paragraph>
    <Paragraph position="6"> The reminder of this paper is structured as follows. Section 2 describes two fundamental techniques for our query expansion method. Section 3 explains a technique to complement the smallness of manual relevance judgment. Section 4 introduces a whole procedure of our query expansion method step by step. Section 5 shows empirical evidence of the effectiveness of our method compared with two traditional query expansion methods. Section 6 investigates the experimental results more in detail.</Paragraph>
    <Paragraph position="7"> Finally, Section 7 summarizes our findings.</Paragraph>
  </Section>
class="xml-element"></Paper>