Discovering Relationships among Word Senses

The direction that future development of natural language processing systems will take is toward increased semantic capability. Systems will need to behave as though they understand the texts that they process. An important prerequisite for this type of behavior is the creation of lexical knowledge bases in which word senses are clearly identified, endowed with appropriate lexical information, and correctly related to one another. This paper discusses issues that arise in creating such a knowledge base, paying particular attention to the process of discovering relationships among the word senses it contains.