Content in context: a data-driven approach

摘要

The dramatic increase in document availability via the Internet means that traditional methods of cataloguing, classifying and retrieving information have been unable to cope with the demand. In order to solve this problem, the emphasis has shifted from the previous labour-intensive methods to technical solutions which rely on statistical and algorithmic means of identifying topics and proper names in documents. While these various solutions have provided a great improvement in retrieval of relevant information, it is still not possible to identify the exact nature of particular terms or relationships between terms and topics. However, using a specially-devised ontology covering topic and type, and a vocabulary set classified using this ontology, it is possible to disambiguate between different senses of terms, and therefore to identify the main topic of a document and the nature of the lexical terms and proper names within it, and to create genuine relationships.