Semantic Text analysis of parliamentary speeches in the state duma of the Russian Empire: methodological aspects

Abstract

The essay focuses on contemporary method of text analysis of political narrative. Traditional approach to content analysis assumes creation of frequency dictionaries which means that each word is considered as an isolated element of the text out of connection with other words. It raises a problem of incorrect interpretation of the results which can be solved by using semantic text analysis and network analysis in particular. Network technologies can be useful in two different cases: the first one is the opportunity of detailed reconstruction of an author’s views. In this case semantic categories (nodes of network) are retrieved from frequency dictionary, then, co-occurrence matrix is created on which network model is based. Another way of research is connected with preliminary determination of the list of categories. This can be productive in a comparative analysis of different texts or authors. In our opinion, betweenness centrality is the best network metric for detection of key categories because it displays nodes close to the majority of network segments. The chosen methodology was verified through two examples from Russian parliament history. In the first example, the author considered the case of a deputy’s change of the faction and concluded on the basis of his speeches that ideologically he was ready for the change of the group. In the second example the author investigated the views of ideological opponents within the same faction and made the conclusion that they took different conflicting ideas from the party program. In general, semantic text analysis proved its practical applicability.