<?xml version="1.0" standalone="yes"?>
<Paper uid="C02-1151">
  <Title>Probabilistic Reasoning for Entity &amp; Relation Recognition/</Title>
  <Section position="2" start_page="0" end_page="13" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Recognizing and classifying entities and relations in text data is a key task in many NLP problems such as information extraction (IE) (Califf and Mooney, 1999; Freitag, 2000; Roth and Yih, 2001), question answering (QA) (Voorhees, 2000) and story comprehension (Hirschman et al., 1999). In a typical IE application of constructing a jobs database from unstructured text, the system has to extract meaningful entities like title and salary and, ideally, to determine whether the entities are associated with the same position. In a QA system, many questions ask for specific entities involved in some relations. For example, the question &amp;quot;Where was Poe born?&amp;quot; in TREC-9 asks for the location entity in which Poe was born. The question &amp;quot;Who killed Lee Harvey Oswald?&amp;quot; seeks a person entity that has the relation kill with the person Lee Harvey Oswald.</Paragraph>
    <Paragraph position="1"> In all earlier works we know of, the tasks of identifying entities and relations were treated as separate problems. The common procedure is to first identify and classify entities using a named entity recognizer and only / Research supported by NSF grants CAREER IIS-9984168 and ITR IIS-0085836 and an ONR MURI Award.</Paragraph>
    <Paragraph position="2"> then determine the relations between the entities. However, this approach has several problems. First, errors made by the named entity recognizer propagate to the relation classifier and may degrade its performance significantly. For example, if &amp;quot;Boston&amp;quot; is mislabeled as a person, it will never be classified as the location of Poe's birthplace. Second, relation information is sometimes crucial to resolving ambiguous named entity recognition.</Paragraph>
    <Paragraph position="3"> For instance, if the entity &amp;quot;JFK&amp;quot; is identified as the victim of the assassination, the named entity recognizer is unlikely to misclassify it as a location (e.g. JFK airport).</Paragraph>
    <Paragraph position="4"> This paper develops a novel approach for this problem - a probabilistic framework for recognizing entities and relations together. In this framework, separate classifiers are first trained for entities and relations. Their output is used to represent a conditional distribution for each entity and relation, given the observed data. This information, along with constraints induced among relations and entities (e.g. the first argument of kill is likely to be a person; the second argument of born in is a location) are used to make global inferences for the most probable assignment for all entities and relations of interest. Our global inference approach accepts as input conditional probabilities which are the outcomes of &amp;quot;local&amp;quot; classifiers. Note that each of the local classifiers could depend on a large number of features, but these are not viewed as relevant to the inference process and are abstracted away in this process of &amp;quot;inference with classifiers&amp;quot;. In this sense, this work extends previous works in this paradigm, such as (Punyakanok and Roth, 2001), in which inference with classifiers was studied when the outcomes of the classifiers were sequentially constrained; here the constraints are more general, which necessitates a different inference approach.</Paragraph>
    <Paragraph position="5"> The rest of the paper is organized as follows. Section 2 defines the problem in a formal way. Section 3 describes our approach to this problem. It first introduces how we learn the classifiers, and then introduces the belief network we use to reason for global predictions. Section 4 records preliminary experiments we ran and exhibits some promising results. Finally, section 5 discusses some of the open problems and future work in this framework.</Paragraph>
  </Section>
class="xml-element"></Paper>