Japanese documents have noun synonyms. These use kanji notation, hiragana notation, and katakana notation for words. Sometimes words have alternate kanji expressions: alternate names for an object, different suffixes for kanji, etc. This is why noun synonym sets are formed for Japanese nouns. Thesauruses and dictionaries can be used to select a representative expression from a noun synonym set. However, these references do not consider the type of document. Representative nouns are often different depending on the type of articles. For example, in articles in newspapers, kanji is preferred. In contrast, in articles in encyclopedias, katakana is preferred. The problem is to form a rule set to select a representative noun from a noun synonym set, and the rule set must consider the type of document. We propose a rule set arranged for the WEB Fish Encyclopedia (in Japanese, Sakanazukan). We introduce a keyword category in the rule set to increase the correctness of the selected representative noun. As a result, most of the representative expressions were selected appropriately from noun synonyms. We expressed these noun synonyms as feature vectors. By using three numerical values and four Boolean values, all noun synonyms were expressed.