Placement of Nouns in a Multi-Dimensional Space Based on Words' Cooccurrency

Abstract

The semantic similarity (or distance) between words is one of the basic knowledge in Natural Language Processing. There have been several previous studies on measuring the similarity (or distance) based on word vectors in a multi-dimensional space. In those studies, high dimensional feature vectors of words are made from words' cooccurrence in a corpus or from reference relation in a dictionary, and then the word vectors are calculated from the feature vectors through the method like principal component analysis. This paper proposes a new placement method of nouns into a multi-dimensional space based on words' cooccurrence in a corpus. The proposed method doesn't use the high dimensional feature vectors of words, but is based on the idea that "vectors corresponding to nouns which cooccur with a word w in a relation f constitute a group in the multi-dimensional space". Although the whole meaning of nouns isn't reflected in the word vectors obtained by the proposed method , the semantic similarity (or distance) between nouns defined with the word vectors is proper for an example-based disambiguation method.

Image estimation of words based on adjective co‐occurrences

“…As a result, we tend to avoid these kinds of expressions. On the other hand, the examples in (1) feel more natural than those in (2). This is because the kind of collision that was mentioned earlier does not occur.…”

“…This is similar for an "and" expression in English. On the other hand, the examples in (1) feel more like natural expressions than those in (2). "akarui (bright)" and "genki (healthy)" exhibit a strong positive correlation, and much of the "genki (healthy)" component has already been declared at the moment "akarui (bright)" is declared.…”

“…If "kurai (dark)-akarui (bright)" are taken as the image words, spatial positioning along image words means that related words that exhibit a strong disposition toward "kurai (dark)" and related words that exhibit a strong disposition toward "akarui (bright)" are greatly separated. Techniques for spatially positioning words based on similarity include techniques that use quantification method 4 [1] or vector variance minimization [2]. However, with these techniques, spatial positioning is performed only according to the similarity between words, and an arbitrary image word cannot be specified as the basis of the spatial axis.…”

“…Various research projects have been conducted in which similarities were obtained based on co-occurrence frequencies, and these were used for feature representations of words [1][2][3][4][5]. Omachi and colleagues [1] conducted research in which they obtained similarities based on co-occurrence relationships from a corpus and the EDR concept classification dictionary to position nouns in a multidimensional space.…”

“…It has been reported that expected results are not necessarily obtained even if similarity is measured from a corpus since the frequency of synonyms such as "bijin (beautiful woman)" and "bijo (beautiful woman)" is low [2]. Attempts have been made to obtain word information by using text on the Web instead of a corpus [6][7][8][9].…”