User Profile: Working with WordStat in the Netherlands July 11, 2016 - Blogs on Text Analytics

 

Dr. Pepijn Van Houwelingen is a researcher at the Netherlands Institute for Social Research. It is an interdepartmental, scientific institute that conducts social science research to help the government with its planning and policy development. The main fields studied are health, welfare, social security, the labour market and education with a particular focus on the relationships between these areas. Dr. Van Houwelingen has worked as a researcher at the institute for the last seven years.  He received his Phd at Hiroshima City University.  His thesis was titled: Social capital in Japan. Prior to that he received a Master of Science in Economics and Business in Pacific Asia, a Master of Arts in Philosophy of Economics from Erasmus University Rotterdam and a Master of Science in Industrial Engineering & Management from  University of Twente.

In his work at the Netherlands Institute for Social Research Dr. Van Houwelingen is involved in the research of social and cultural issues. He is part of a group that conducts a quarterly public opinion survey of

1,000 people.  One question that is asked each quarter is, Do you think the Netherlands is heading in the right direction and if so why and if not why not. Since the question has been asked every quarter beginning in 2008 there are now more than 25,000 responses. Respondents comments and the issues they raise change over time usually depending on critical issues of the day. For example, in one quarter a certain aspect of the nation’s health care system may be at the forefront, on another survey it could be public anger at bonuses paid to public servants.

Dr. Van Houwelingen uses WordStat to analyze the results and compares them with previous surveys. The exploratory text mining features of WordStat help him find themes by using the cluster extraction and topic modelling tools and trends by using the crosstab and correspondence analysis plot in order to see relationships between time and topics in the responses and track those changes over time.  In order to have a more comprehensive analysis of the results, Dr. Van Houwelingen built a categorization dictionary with several categories based on the themes defined by his team and those automatically identified by WordStat. In order to continue to improve the dictionary and analyze all responses, Dr Houwelingen uses features like “leftover words” and “show related words” in order to quickly identify non-classified words and assign them to the appropriate categories. Once the categories are built, the crosstab and correspondence analysis plot are used to identify relationship between time and categories and identify any trends and patterns.