Text Mining

Given that ever more consumer-generated text is available (on blogs, on forums, from companies websites, etc.), analyzing this wealth of information in an intelligent way will become increasingly more important. We conduct text mining in a series of different steps. In the first step, we simply count how often each word occurs in the open answers. This can be done for different topics, different sources and even for different socio-demographic variables if the text can be linked to personal data (e.g. for open answers in surveys). In a second step, we look for strong associations between words. Instead of looking for associations between bigrams (two words), topic detection is looking for associations between all words. This technique provides a clustering of words according to their association. By looking at the specific words in a cluster you can assign a meaning to it (a topic). The specific topics can tell you what is currently on the mind of the customers. Another step is to find out the general sentiment underlying the texts that were provided.

If you're interested in this or any other service we're offering, please contact us.