This paper proposes a method of the zero anaphora resolution, which is one of the essential processes in Japanese manual sentences understanding system which uses pragmatic properties of Japanese conditionals. We examined a number of sentences in Japanese manuals according to the classification based on the types of agents and the types of verb phrase. As a result, we obtained the following pattern of usage in main clauses: 1) The connective particles TO and REBA have the same distribution of usage. TARA and NARA have the same distribution of usage. 2) The distribution of usage of TO and REBA and that of TARA and NARA are complementary to each other.

View full abstract

In this paper, we analyze the pragmatic aspects of Japanese verbal suffixes teiru, tearu, tesimau, teoku and temiru used in manual sentences in Japnanese. These suffixes mainly express temporal structure of events but also express some kinds of modality such as a writer's attitude. By analyzing these kinds of modality used in manual, we propose the constraints as for the possible referents of an agent of action described by a verb phrase subcategorized by one of these suffixes. The proposed constraints are proven to be valid by examining a number of manual sentences.

View full abstract

This paper shows a morpheme grammar for Japanese morphological analyzers. The grammar bases on Derivational Grammar in which the inflection of Japanese verbs can be managed systematically. Though Derivational Grammar uses phonological description, this paper shows a method to process Japanese character strings directly. A Japanese morphological analyzer with our grammar is evaluated with EDR corpus.

View full abstract

This paper proposes a semantic classification of Japanese particle “ga” and “ha” based on the viewpoint of speaker's recognition and its implementation on SGLR parser. SGLR parser is a generalized LR parser which uses Tomita's algorithm. We already showed its advantages on parsing speed with an experiment. Not only syntactic processes but also semantic processes are described in DCG rules with augmentations because SGLR parser is implemented in Prolog.
The speaker's recognition process means the process that a speaker recognizes objects faced at some subject, conceptualizes them as words on his language, and tells them from his viewpoint and as a result of his judgement toward concepts. In this process, we especially consider the semantic rules of Japanese particles “ga”, “ha”, and also “ni”, “wo”. According to Miura grammar such Japanese particles have viewpoint of speakers recognition, so we tried to extract them from sentences using semantic rule. In the experiment with these DCG rules, we tried to show the effect of the semantic rules which is a disambiguation of parse tree. The experiment succeeded to show that, semantic rules could remove syntactic ambiguity.

View full abstract