A housing equipment company under consideration typically receives about 110000 phone calls per month at three call centers. The resulting data, called VOC data, involve a variety of contents including complaints about products, requests for product information and questions about maintenance services. In this paper, we analyze the VOC data for the month of March 2015, through text mining combined with other data mining techniques, so as to assess the corporate image of the company collectively perceived by the callers. Focusing on calls related to toilets, bath modules, sash windows and bathroom sinks and further eliminating meaningless data such as sentences too short, 22838 calls are extracted for the study. Then 1956 calls are chosen at random, each of which is evaluated by three people along the following binary axes: 1) sales potential; 2) negative attitude; and 3) urgency. Using 1304 calls as a learning data set and the remaining 652 calls as a testing data set, an algorithmic procedure is developed for establishing a special dictionary and quantifying the three scores of each call. The procedure is applied to the entire VOC data, thereby enabling one to capture the corporate image of the company represented by the VOC data.