<?xml version="1.0" standalone="yes"?>
<Paper uid="W05-1207">
  <Title>Discovering entailment relations using &amp;quot;textual entailment patterns&amp;quot;</Title>
  <Section position="2" start_page="0" end_page="37" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Textual entailment has been recently defined as a common solution for modelling language variability in different NLP tasks (Glickman and Dagan, 2004).</Paragraph>
    <Paragraph position="1"> Roughly, the problem is to recognise if a given textual expression, the text (t), entails another expression, the hypothesis (h). An example is determining whether or not &amp;quot;Yahoo acquired Overture (t) entails Yahoo owns Overture (h)&amp;quot;. More formally, the problem of determining a textual entailment between t and h is to find a possibly graded truth value for the entailment relation t - h.</Paragraph>
    <Paragraph position="2"> Since the task involves natural language expressions, textual entailment has a more difficult nature with respect to logic entailment, as it hides two different problems: paraphrase detection and what can be called strict entailment detection. Generally, this task is faced under the simplifying assumption that the analysed text fragments represent facts (ft for the ones in the text and fh for those in the hypothesis) in an assertive or negative way. Paraphrase detection is then needed when the hypothesis h carries a fact f that is also in the target text t but is described with different words, e.g., Yahoo acquired Overture vs. Yahoo bought Overture. On the other hand, strict entailment emerges when target sentences carry different facts, fh negationslash= ft. The challenge here is to derive the truth value of the entailment ft - fh. For example, a strict entailment is &amp;quot;Yahoo acquired Overture - Yahoo owns Overture&amp;quot;. In fact, it does not depend on the possible paraphrasing between the two expressions but on an entailment of the two facts governed by acquire and own.</Paragraph>
    <Paragraph position="3"> Whatever the form of textual entailment is, the real research challenge consists in finding a relevant number of textual entailment prototype relations such as &amp;quot;X acquired Y entails X owns Y&amp;quot; or &amp;quot;X acquired Y entails X bought Y&amp;quot; that can be used to recognise entailment relations. Methods for acquiring such textual entailment prototype relations are based on the assumption that specific facts are often repeated in possibly different linguistic forms.</Paragraph>
    <Paragraph position="4"> These forms may be retrieved using their anchors, generally nouns or noun phrases completely characterising specific facts. The retrieved text fragments are thus considered alternative expressions for the same fact. This supposed equivalence is then exploited to derive textual entailment prototype relations. For example, the specific fact Yahoo bought Overture is characterised by the two anchors  {Yahoo, Overture}, that are used to retrieve in the corpus text fragments where they co-occur, e.g. &amp;quot;Yahoo purchased Overture (July 2003).&amp;quot;, &amp;quot;Now that Overture is completely owned by Yahoo!...&amp;quot;. These retrieved text fragments are then considered good candidate for paraphrasing X bought Y.</Paragraph>
    <Paragraph position="5"> Anchor-based learning methods have been used to investigate many semantic relations ranging from very general ones as the isa relation in (Morin, 1999) to very specific ones as in (Ravichandran and Hovy, 2002) where paraphrases of question-answer pairs are searched in the web or as in (Szpektor et al., 2004) where a method to scan the web for searching textual entailment prototype relations is presented.</Paragraph>
    <Paragraph position="6"> These methods are mainly devoted to induce entailment pairs related to the first kind of textual entailment, that is, paraphrasing as their target is mainly to look for the same &amp;quot;fact&amp;quot; in different textual forms. Incidentally, these methods can come across strict entailment relations whenever specific anchors are used for both a fact ft and a strictly entailed fact fh.</Paragraph>
    <Paragraph position="7"> In this work we will investigate specific methods to induce the second kind of textual entailment relations, that is, strict entailment. We will focus on entailment between verbs, due to the fact that verbs generally govern the meaning of sentences.</Paragraph>
    <Paragraph position="8"> The problem we are facing is to look for (or verify) entailment relations like vt - vh (where vt is the text verb and vh the hypothesis verb). Our approach is based on an intuition: strict entailment relations among verbs are often clearly expressed in texts. For instance the text fragment &amp;quot;Player wins $50K in Montana Cash&amp;quot; hides an entailment relation between two activities, namely play and win. If someone wins, he has first of all to play, thus, win play. The idea exploits the existence of what can be called textual entailment pattern, a prototypical sentence hiding an entailment relation among two activities. In the abovementioned example the pattern instance player win subsumes the entailment relation &amp;quot;win - play&amp;quot;.</Paragraph>
    <Paragraph position="9"> In the following we will firstly describe in Sec.</Paragraph>
    <Paragraph position="10"> 2 our method to recognise entailment relations between verbs that uses: (1) the prior linguistic knowledge of these textual entailment patterns and (2) statistical models to assess stability of the implied relations in a corpus. Then, we will experiment our method by using the WordNet entailment relations as test cases and the web as corpus where to estimate the probabilities (Sec. 3). Finally we will draw some conclusions (Sec. 4).</Paragraph>
  </Section>
class="xml-element"></Paper>