Analysing Parliamentary Debate with Computer Assistance

The analysis of parliamentary debates is at the confluence of a number of developments in political science. What light can automated and semi-automated techniques throw on such analysis? In this paper we compare two such approaches, one semi-automated (Hamlet) and the other fully automated (Alceste). We use both approaches to identify the prominent themes in debate and to assess how far speakers who favour different positions adopt a distinct pattern of discourse. We seek to assess how far the two approaches yield convergent or divergent analyses. Selecting a second reading debate from the UK House of Commons on a private member's bill on abortion in July 1966, we are able to show similarities of analysis despite the detailed differences between the two approaches. In particular, the analysis in Hamlet al.lows identification of the extent to which individual speakers employ one type of vocabulary rather than another. Alceste is able to provide a statistical basis for the different classes of vocabulary that occur in the debate. However, the two programs rest upon quite different assumptions about the relationship between syntax and meaning, with implications for the practice of political science.