Eliciting Considerable Requirements with Word and Customer Graphs

We propose a graph-based method to elicit considerable requirements from customer survey results. Requirements usually have relationships with each other, and understanding such relationships gives us deeper insight. One useful approach to investigate the relationships is making a co-occurrence graph of words where links are connected if words co-occur in a specific requirement. If we find uncommon links from the word graph, they help us to elicit considerable requirements. In our method, we make a customer graph in addition to a word graph. This method indicates some uncommon relationships by reflecting centrality analysis of the customer graph to the links of the word graph. We have applied our method to the result of a survey using a questionnaire containing open-ended answer question. Links in the word graph made from the survey results include information of customers. We have shown we can elicit some considerable contexts, by reflecting the information of customers who bridge majorities and minorities.