Extraction and Classification of Employment-Related Information from Web Data

On the web, there is large amount of information pertaining to various job-hunting activities, and automatically collecting this information would be efficient and convenient. In this study, information related to job hunting was extracted from massive web data and was classified. For extracting employment-related information, the performance of the following three methods was compared: a supervised machine learning method called a support vector machine (SVM), a rule-based method, and a baseline method. The rule-based method obtained superior performance with an F-measure of 0.7. To classify employment-related information, the following seven categories were used: “qualification information,” “vocational information,” “job seeker information,” “job offer information,” “no relationship,” “job hunting support information,” and “job change information.” The results of the comparison between the three methods revealed the rule-based method displayed superior performance, with an average F-measure of approximately 0.6. In addition, the rule-based method had an F-measure of approximately 0.8 for the categories “qualification information,” “job offer information,” and “job change information.”