Yumi Ichimura

Inductive Learning of a Knowledge Dictionary for a Text Mining System

A text mining system using domain-dependent dictionaries efficiently analyzes text data. The dictionaries store not only important words for the domains, but also rules composed of some important words. The paper proposes a method that automatically acquires the rules from the text data and their classes by using a fuzzy inductive learning method. Also, in order to infer a class corresponding to new text data, the paper proposes an inference method based on the acquired fuzzy decision tree. Moreover, the efficiency of the methods is verified through numerical experiments using more than 1,000 daily business reports concerning retailing.

Acquisition of a Knowledge Dictionary from Training Examples Including Multiple Values

A text mining system uses two kinds of background knowledge: a concept relation dictionary and a key concept dictionary. The concept relation dictionary consists of a set of rules. We can automatically acquire it by using an inductive learning algorithm. The algorithm uses training examples including concepts that are generated by using both lexical analysis and the key concept dictionary. The algorithm cannot deal with a training example with more than one concept in the same attribute. Such a training example is apt to generate from a report, when the concept dictionary is not well defined. It is necessary to extend an inductive learning algorithm, because the dictionary is usually not completed. This paper proposes an inductive learning method that deals with the report. Also, the paper shows the efficiency of the method through some numerical experiments using business reports about retailing.

Kana-kanji conversion system with input support based on prediction

We propose a kana-kanji conversion system with input support based on prediction. This system is composed of two parts: prediction of succeeding kanji character strings from typed kana ones, and ordinary kana-kanji conversion. It automatically shows candidates of kanji character strings which the user intends to input. Our prediction method features: (i)Arbitrary positions of typed kana character strings are regarded as the top of words. (ii)A system dictionary and a user dictionary are used, and each entry in the system dictionary has certainly factor calculated from the frequency of words in corpora. (iii)Candidates are estimated by certainty factor and usefulness factor, and likely ones with greater factors than thresholds are shown. The proposed system could reduce the users key input operations to 78% from the original ones in our experiments.