All processes of formulating and implementing development cooperation policies create text: proposals describe means and ends of interventions; progress reports monitor the steps towards desired outcomes; and evaluations provide evidence on the effects of development interventions. In addition, press and social media reflect the public opinion on development cooperation and, in some cases, are even indicative of outcomes.
The increased availability of digital text has rendered detailed reading of all relevant sources impossible. Text mining combines a qualitative appraisal of meaning and complex statistical approaches to extract relevant information from text. It provides a methodological tool to cope with the abundance of digital information by enabling evaluators to efficiently analyse large document collections.
The German Institute of Development Evaluation applies text mining in its evaluations and develops tools to facilitate the application of text mining by evaluators. This includes the development of a tool to geocode project locations based on project documents, machine learning for analysing human rights mainstreaming in development project documents, natural language processing and sentiment analysis of newspaper articles and tweets, and supporting the development of a text mining infrastructure that provides a web-interface and integrated workflows to simplify the use of text mining for evaluators.