Customer language processing: Extended abstract

Abstract

The research presented is the first working step towards the goal of developing a domain-independent method for sentiment analysis of German customer feedback in social media. The approach proposes to apply the concept of natural language processing (NLP) to customer language processing (CLP). In this context we hypothesize an indifference in annotator ability in assigning customer reviews of tangible vs. intangible goods and an indifferences within customers' writing styles within their evaluation of these goods. To test these hypotheses, a study was conducted where participants had to assign the sentiment as well as the subject of customer reviews and its evaluative attribute. The results reveal that the inter-rater reliability of annotators does not differ significantly with respect to product groups. However a slight difference with respect to product categories could be observed. Moreover, there occur variations within the inter-rater ability according to the emotional commitment towards products.