Domain knowledge
The NewsTools Project
Obtaining and using background domain knowledge
Obviously, the NewsTools approach is dependent on the quality of the domain knowledge. Given the sophistication of the domain knowledge, we need to represent it in logic. Only by using the expressibility of a first-order logic can we efficiently, and precisely, capture the generalities in domain knowledge.
Structured text can be translated into classical predicate logic. Consistency checking is then conducted using specialized logic-based techniques. We have assumed the words and phrases are sufficiently simple and restricted to not require natural language processing. In most cases, we will represent each word or phrase by a constant symbol in the logic. Once we represent items of structured text and domain knowledge in logic, we can use algorithmic techniques for analysing news reports.
We are currently developing domain knowledge interfaces. This include interfaces to the following kinds of background knowledge: (1) relational database resources (some of our collaborators are providing us with access to their extensive database resources which includes relational data on companies, with subsidiaries, key personnel, turnover, business activities, etc); (2) a set of integrity constraints (for example a company listed on the London Stock Exchange is a plc); (3) a set of subsidiary axioms for reasoning with time, space, and other units that arise in many business news domains. (e.g. axioms that caputure the sequence of days in a month, the fact that January 31st is followed by 1 February, and the equivalence between 31 January 2001 and 31st January 2001); and (4) a set of defeasible rules rules in the domain knowledge (for example company X is a white knight for company Y only after there is has been a company Z making an aggressive bid for company Y). We are investigating the use of Prolog for representing and reasoning with integrity constraints, subsidiary axioms, and defeasible rules.
For more information about the NewsTools Project, contact Tony Hunter (a.hunter@cs.ucl.ac.uk or +44 20 7679 7295), or look at the NewsTools Project
home page.