Abstract
Human can recognize the subject of document fields by reading only some relevant specific word called Field Association (FA) terms in that field. Various researches focused on how to extract FA terms depending on levels or frequency information. Therefore, the extracted FA terms are connected to several different fields. The traditional method causes misleading irrelevant terms to be registered because the quality of the resulting FA terms depends on levels and frequency information only. As a consequence, ambiguity occurs. This paper introduces two cases of ambiguity (1) In-Field Ambiguity and (2) Cross-Field Ambiguity to analyze the criteria of ambiguity. To treat these disadvantages this paper proposes a new technique to disambiguate FA terms using co-occurrence information. From the experimental results, Recall and Precision are achieved 80 similar to 84% and 82 similar to 88% respectively. By using the presented algorithm, Recall and Precision are about 17% similar to 20% higher than the traditional methods.