<?xml version="1.0" standalone="yes"?>
<Paper uid="P03-1018">
  <Title>Orthogonal Negation in Vector Spaces for Modelling Word-Meanings and Document Retrieval</Title>
  <Section position="5" start_page="0" end_page="0" type="concl">
    <SectionTitle>
5 Conclusions
</SectionTitle>
    <Paragraph position="0"> Traditional branches of science have exploited the structure inherent in vector spaces and developed rigourous techniques which could contribute to natural language processing. As an example of this potential fertility, we have adapted the negation and disjunction connectives used in quantum logic to the tasks of word-sense discrimination and information retrieval.</Paragraph>
    <Paragraph position="1"> Experiments focussing on the use of vector negation to remove individual and multiple terms from queries have shown that this is a powerful and efficient tool for removing both unwanted terms and their related meanings from retrieved documents.</Paragraph>
    <Paragraph position="2"> Because it associates a unique vector to each query statement involving negation, the similarity between each document and the query can be calculated using just one scalar product computation, a considerable gain in efficiency over methods which involve some form of post-retrieval filtering.</Paragraph>
    <Paragraph position="3"> We hope that these preliminary aspects will be initial gains in developing a concrete and effective system for learning, representing and composing aspects of lexical meaning.</Paragraph>
    <Paragraph position="4"> Demonstration An interactive demonstration of negation for word similarity and document retrieval is publicly available at http://infomap.stanford.edu/webdemo.</Paragraph>
  </Section>
class="xml-element"></Paper>