 |
|
| Text Mining extracts human intentions and sentiments from written texts and reveals hidden tendencies and correlations among the concepts in the texts. Such features are hard to detect manually. We have developed two solution systems. One is a text mining system called TAKMI, used for problem recognition, FAQ creation, finding potential customers, etc., based on VOC (voice of the customers) analysis of call center logs. The other is a sentiment analysis system, used for extracting and analyzing positive or negative comments within text (such as questionnaires) for rapid user feedback to product and service development teams. |
|
| |
|
| It has become increasingly important for the health care industry to identify the best clinical courses for specific combinations of diseases and patients by analyzing a vast amount of clinical information. Clinical information mining is a technology for supporting discovery of common clinical patterns and predictive rules from complex sequences of patient statuses and clinical treatment events. It also includes technologies for extracting evidence from a large corpus of biomedical literature, and for exploring relationships among genes, proteins, and diseases. |
|
| |
|
| In cooperation with the T.J. Watson Research Center, TRL has been doing research on Japanese speech recognition and synthesis for more than 20 years. The results of our research are used in solution systems and in some of IBM's software products, such as ViaVoice and Easy Web Browsing. The goal of our project is to enable highly accurate speech recognition and high quality speech synthesis on various devices such as car navigation systems, mobile phones, and personal robots. However, these two capabilities are not sufficient to build a useful spoken dialogue system. Our research foci also include speech understanding and dialog management. |
|
| |
|
|
