First we should define what we mean by computer-assisted coding: it is a situation where a person codes an answer using the computer to search for possible classifications based on some search text. Compared to automatic coding the demands for such a program are less strict: the program may return multiple results, ordered by relevance.
Computer-assisted coding can be used:
- During the interview if a question arises that requires coding: the coding can be done either by the respondent (e.g., CAWI) or by the interviewer (e.g., CAPI or CATI).
- After the interview has taken place and some of the variables need to be coded at the statistical office by a coding expert.
Obviously, these situations require different approaches depending on the knowledge of the person that codes the question: if a respondent fills in a coding question, one must assume he has little or no knowledge of the targeted classification. On the other hand, a coding expert trying to code an open answer from the interview just has the information supplied in the open text answer as a basis for coding. Both situations require a different interaction with the computer: an unknowledgeable respondent needs to be taken by the hand to arrive at the classification, whereas the expert needs to be able to formulate a detailed search.
In the following sections we will describe two situations with a different degree of interaction. It will depend on the underlying search system how much interaction is possible.
All the pre-processing steps, such as stop word removal are applicable here as well, but will not be described; for more detail on these steps see Hacking and Willenborg (2012) and Sebastiani (2001).
To read the entire document, please access the pdf file (link under "Related Documents" on the right-hand-side of this page).
Your feedback is appreciated. Please send your remarks, suggestions for improvement, etc. to firstname.lastname@example.org.