Artificial intelligence and content analysis - Problems of and strategies for computer text analysis

Content analysis is a very tedious method of data collection. This paper addresses once more the question of whether and how computers may be used to facilitate content analysis in the coding stage. To interpret natural language automatically a computer program must be able to unravel the syntactical structure of sentences ('parsing') and to trace their semantical meaning by dealing with textual context (= 'co-text'), prior knowledge (context outside the text) and semantic variability (different or ambiguous meanings of words and phrases). Several approaches to enable computer programs to perform these tasks are discussed, including approaches from the fields of cognitive psychology and artificial intelligence. The conclusion should be that it is still impossible to enable computer programs to perform all these tasks. But there are possibilities for using computer programs to support human coders both in the coding stage and the data-analysis stage of content analysis. As an example the program CETA to perform Computer-aided Evaluative Textual Analysis is discussed.