<?xml version="1.0" standalone="yes"?>
<Paper uid="N06-4005">
  <Title>the Semantic Web</Title>
  <Section position="6" start_page="270" end_page="271" type="concl">
    <SectionTitle>
4 Related Work
</SectionTitle>
    <Paragraph position="0"> This scenario is similar to research in NL queries to databases (NLIDB). However, the SW provides a new and potentially important context in which results from this research area can be applied.</Paragraph>
    <Paragraph position="1"> There are linguistic problems common in most kinds of NL understanding systems, see (Androutsopoulos, 1995) for an overview of the state of the art. In contrast with the latest generation of NLIDB systems (see (Popescu, 2003) for recent work) AquaLog uses an intermediate representation from the representation of the user's query (NL front end) to the representation of an ontology compliant triple, from which an answer can be directly inferred. It takes advantage of the use of ontologies in a way that the entire process highly portable and it is easy to handle unknown vocabulary. For instance, in PRECISE (Popescu, 2003) the problem of finding a mapping from the tokenization to the database requires that all tokens must be distinct, questions with unknown words are not semantically tractable and cannot be handled. In contrast with PRECISE, AquaLog interpret the user query by means of the ontology vocabulary and structure in order to make sense of unknown vocabulary which appears not to have any match.</Paragraph>
    <Paragraph position="2"> Current work on QA is somewhat different in nature from AquaLog as they are open-domain systems. QA applications to text typically involve (Hirschman, 2001) identifying the semantic type of the entity sought by the question (a date, a person...); and determining key words or relations to be use in matching candidate answers. Moreover, as pointed by Srihari et al. (Srihari, 2004) Named Entity (NE) tagging is often necessary. The main differences between AquaLog and open-domains systems are: (1) it is not necessary to build hierarchies or heuristic to recognize NE, as all the semantic information needed is in the ontology. (2) AquaLog has mechanisms to exploit the relationships to understand a query. Nevertheless, the RSS goal is to map the relationships in the Query-Triple into an ontology-compliant-triple. Both AquaLog and open-domain systems attempt to find synonyms plus their morphological variants. AquaLog also automatically classifies the question before hand, based on the kind of triple needed, while most of the open-domain QA systems classify questions according to their answer target. The triple contains information not only about the an- null swer expected, but also about the relationships of the other terms in the query. To conclude, other QA systems also follow a relational data model (triple-based), e.g. the START &amp;quot;object-propertyvalue&amp;quot; approach (Katz, 2002). 5 AquaLog in action: illustrative example.</Paragraph>
    <Paragraph position="3"> For demonstration purposes AquaLog application is used with the AKT ontology in the context of the academic domain in our department (Lei, 2006), e.g., AquaLog translates the query &amp;quot;what is the homepage of Peter who has an interest on the semantic web?&amp;quot; into a conjunction of ontology-compliant non-ground triples: &lt;what is?, has-webaddress, peter-scott&gt; &amp; &lt;person?, has-researchinterest, Semantic Web area&gt;.</Paragraph>
    <Paragraph position="4"> Consider the query &amp;quot;what is the homepage of Peter?&amp;quot; on Fig. 2. Given that the system is unable to disambiguate between Peter-Scott, Peter-Sharpe, etc, user feedback is required. Also the user is call to disambiguate that &amp;quot;homepage&amp;quot; is the same that &amp;quot;has-web-address&amp;quot; as it is the first time the system came across this term, no synonyms have been identified, and the ontology does not provide further ways to disambiguate. The system will learn the mapping and context for future occasions.</Paragraph>
    <Paragraph position="5">  On Fig. 3 we are asking for the web address of Peter, who has an interest in SW. In this case AquaLog does not need any assistance from the user, given that only one of the Peters has an interest in SW. Also the similarity relation between &amp;quot;homepage&amp;quot; and &amp;quot;has-web-address&amp;quot; has been learned by the Learning Mechanism. When the RSS comes across a query like that it has to access to the ontology information to recreate the context and complete the ontology triples. In that way, it realizes that &amp;quot;who has an interest on the Semantic Web&amp;quot; is a modifier of the term &amp;quot;Peter&amp;quot;.</Paragraph>
  </Section>
class="xml-element"></Paper>