부산대학교 도서관

This article shows that the unambiguous meaning representation of machine language must be inspired by human cognition to find a universal label for Natural Language Processing resources. We show, that the meaning representation is linked to the quality and not to the quantity of data, since the architecture of human symbolization, taken as a model, incorporates contextual elements in its dynamics. We argue that language, when conceived under this dynamic concept, must be analyzed in terms of sequence-by-sequence transformations, focusing on the structural bases of the meaning construction that: i) offers relative certainty for different possible responses; ii) provides more accurate and reliable input data; and iii) avoids the curse of dimensionality. We conclude by suggesting that the construction of databases should indicate the upper hierarchical structure based on the lower structures, referring to the string-to-meaning derivation in the fundamental tree of the language. This strategy, applied to meaning representation, frees the operator from dictating a specific sequence for each application. Thus, the construction of meaning in natural language processing becomes more contextualized and unambiguous. Based on the dynamic structure of the language, therefore, it is possible to develop semantic databases in any spoken language. [ABSTRACT FROM AUTHOR]