Abstract

When surveys ask for occupation, they use open-ended questions. For further statistical analysis, the textual answers need to be assigned into occupational classifications (KldB or ISCO). Currently, this process is partly automated using a database. If entries are not available in the database, manual coding is necessary. This is time-consuming, error-prone and expensive. As an alternative, we propose to suggest possible answer option by using machine learning and let the respondents choose the most appropriate occupation themselves.