Marc Tilbrook

Dynamic semantics at work

In this case study we show how an unambiguous semantic representation can be constructed dynamically in left-to-right order while a text is written in PENG, a controlled natural language designed for knowledge representation. PENG can be used in contexts where precise texts (e.g. software specifications, axioms for formal ontologies, legal documents) need to be composed. Texts written in PENG look seemingly informal and are easy to write and to read for humans but have first-order equivalent properties that make these texts computer-processable.

Key element summarisation: extracting information from company announcements

In this paper, we describe KES, a system that integrates text categorisation and information extraction in order to extract key elements of information from particular types of documents, with these informational elements being presented in such a way as to provide a concise summary of the input document We describe the overall architecture of the system and its components, with a particular focus on the problems involved in handling the names of companies and individuals in this domain.

Summarising company announcements

This paper describes work that attempts to use language technology as a solution to the problem of information overload. The specific domain of application is the database of company announcements accessible via the Web site of the Australian Stock Exchange to meet regulatory requirements, over 100,000 documents a year are made available via this site, with only limited search facilities. We use a variety of techniques from language technology to make it easier to explore and manage the information in this data set. In this paper, we focus on our use of information extraction, which identifies and extracts important elements of information from a document, and text compaction, which applies linguistically-motived substitutions to reduce potential summary sentences to more compact forms. Together, these techniques provide a way of producing summaries of a significant proportion of the document base.

Information Extraction via Path Merging

In this paper, we describe a new approach to information extraction that neatly integrates top-down hypothesis driven information with bottom-up data driven information. The aim of the kelp project is to combine a variety of natural language processing techniques so that we can extract useful elements of information from a collection of documents and then re-present this information in a manner that is tailored to the needs of a specific user. Our focus here is on how we can build richly structured data objects by extracting information from web pages; as an example, we describe our methods in the context of extracting information from web pages that describe laptop computers. Our approach, which we call path-merging, involves using relatively simple techniques for identifying what are normally referred to as named entities, then allowing more sophisticated and intelligent techniques to combine these elements of information: effectively, we view the text as providing a collection of jigsaw-piece-like elements of information which then have to be combined to produce a representation of the useful content of the document. A principle goal of this work is the separation of different components of the information extraction task so as to increase portability.