Budget Amount *help |
¥1,700,000 (Direct Cost: ¥1,700,000)
Fiscal Year 1991: ¥400,000 (Direct Cost: ¥400,000)
Fiscal Year 1990: ¥1,300,000 (Direct Cost: ¥1,300,000)
|
Research Abstract |
The aim of the present study is to find the semantic constraints that regulate the possibilities of verbalizing of nouns in Japanese and English, and to study the process in a more precise fashion by simulating the process on a computer utilizing neural network models, or connectionist models. We could define the semantic constraints that regulate the verbalizing processes in rather subjective terms, but we are still in the final stage of building a suitable neural network model that can learn and associate semantic information of nouns and verbs. Since previous studies in linguistics utilizing neural network models were rare, we needed to apply the methodology to a simpler problem and acquire techniques how to apply the model to problems in natural languages. Currently, our model is able to learn polysemous words and associate to quite different meanings according to the context units that the words co-occur. However, it is found necessary for the network model to handle temporal information in order to deal with the problem of verbalization in a meaningful way. Thus, we are now working on a new model that can handle temporal information and trying to implement this model on a computer. The semantic constraints we found are as follows. As for English verbalizing processes, when the noun has a self-controllable meaning, then the word can be verbalized by one of the three suffixes, -ize, -ify, -en. If the noun does not have a self-controllable meaning but has a semantic element strongly related to action in the lexicon can be used as a verb without changing its form. In the case of Japanese, if the noun can occur in a frame "the way of doing X", then it can be verbalized by attaching"-- suru". On the other hand, if a noun can satisfy a frame "the characteristic feature of X", then it can be verbalized by using the suffix "--kasuru".
|