The evidence terms are the keywords and phrases that serve as evidence of the category.

The preferred evidence term is the category name. Other evidence terms can be synonyms of the category name to which you give the same confidence value or related terms to which you give a lower confidence value.

If you use only terms that are unique to that category, CIS server does not recognize the category in documents that relate to it in an indirect way. But if you choose common words as evidence terms, CIS server can recognize the category when the document does not belong to it.

The challenge is to create category definitions that are complete enough to trigger category recognition without introducing ambiguity. It is as important to keep misleading terms out of category definitions as it is to make sure that all viable terms are included.

Start with including proper nouns as evidence terms. When the proper noun is made up of several commonly occurring words, such as Internet Service Provider, define the term as a phrase. Collect the vocabulary from a set of documents representative of each category: synonyms, abbreviations, acronyms, antonyms, related terms that appear in the text.

Related topics: