Interactive Acquisition of Terminology Describing Job Applicants in Job Advertisements

There are several computer based job classification systems for the labour domain and the purpose of which mainly is to enable search and to collect statistics. The systems typically encode information about the job, the workplace , the employer, the salary, necessary education and experience, duration, hours, need for a driving licence etc. But actual job advertisements, for example such that can be found in the Swedish Job Bank, a set of web pages containing information about job vacancies provided by AMS, the Swedish government agency for labour market activities, often contain requirements that are not included in these classification schemes. Examples of the following sort are common, the applicant should enjoy working with people, be able to pick up new things quickly or be creative. In an effort to extend existing schemes we have developed a taxonomy of properties like these and a computer tool that can be used to analyse text based on the categories in the taxonomy. The tool can also be used to construct future taxonomies. In this paper, we will present the taxonomy, the computer tool and some results based on using the tool. The paper has three purposes; the first is to present a taxonomy of personal properties of job applicants in job advertisements, the second is to present a tool that can be used both to develop such a taxonomy and to analyse text using the taxonomy, and the third is to describe some results of having used the tool and the taxonomy to analyse job advertisements In Human IT, 1999, No. 2 2 Background Our first task was to find suitable categories to classify words describing job applicants. According to Allwood (1989), several kinds of empirical data can be used to support conceptual analysis and construction of taxonomies, e. g. interviews, intuitions or relevant linguistic material. Given that we wanted our categories to cover a large and varied set of job descriptions, we suspected that intuitions and interviews would not alone provide us with sufficient variation. We therefore decided to develop a taxonomy through a direct analysis of a written corpus of job advertisements written by more than one thousand authors from different labour sectors. The following steps were taken: