Abstract:
|
Currently, most surveys ask for occupation with open-ended questions. The verbatim responses are coded afterwards, which is error-prone and expensive. When textual answers have a low level of detail, exact coding may be impossible. We describe an alternative approach that was tested in a telephone survey: A supervised learning algorithm searches for candidate job categories at the time of the interview. Those job categories that have the highest probability to be correct are then presented to the interviewer who asks in turn the respondent to select the most adequate job title. Respondents can also choose that no suggested job category is adequate. 72.4% of the respondents did select an occupation during the interview making additional manual coding superfluous. The quality of interview-coded occupations is compared with two independent codings from professional coders. It is comparable to one of them but slightly worse than the second. While these results are already highly promising, a number of factors have been identified how the process can even be improved.
|