We propose an approach to gap handling, which is based on GLR parsing method. The idea of gaps has been introduced in Noam Chomsky's transformational grammar. A gap occurs in a sentence when a constituent of the sentence transfers to another position of the sentence. In cases where a parser finds a gap in a sentence and fills it with the corresponding constituent, it is easier to write and maintain grammar because additional grammar rules for gaps are needless. GLR parsing method has no gap handling, although it is known as an efficient parsing method. We make it clear what prevents GLR parsing method from handling gaps, and give a solution to gap handling in GLR parsing method. The main problem is that a gap described in a grammar rule has bad effect on parsing with other grammar rules (LR items) in the same state. We solve this problem by splitting a state into more than one state.

View full abstract

We exploit and extend the Generative Lexicon Theory to develop a formal description of adnominal constituents in a lexicon which can deal with linguistic phenomena found in Japanese adnominal constituents. We classify the problematic behavior into “static disambiguation” and “dynamic disambiguation” tasks. Static disambiguation can be done using lexical information in a dictionary, whereas dynamic disambiguation requires inferences at the knowledge representation level.

View full abstract

This paper takes up the problem of tokenization and part-of-speech tagging of segmented and non-segmented languages, and proposes a simple framework that enables an efficient and uniform treatment of tokenization for both types of languages. We also reports a language independent morphological analysis system based on the proposed idea, and shows running systems for three different languages, English, Japanese and Chinese.

View full abstract

Since both Japanese and Uighur languages are agglutinative languages, they have a lot of syntactic similarities. Thus we can translate from Japanese into Uighur by replacing Japanese words with corresponding Uighur words after morphological analysis of Japanese sentences. However, Japanese verbs have been said to conjugate but Uighur verbs do not, so that we have to analyze Japanese conjugation. In this paper, by using the derivational grammar which formalizes Japanese verbs syntax without conjugation, we propose a simple and systematical translation system.Since there are some differences between Japanese and Uighur, this simple translation sometimes makes output sentences unnatural. To solve this problem, we use a word replacement table which is based on word connection relations on Uighur. The system consists of three independent modules and so we can apply this approach to the translation between other agglutinative languages. We also show the performance evalution of the system. In this paper, we will focus on the translation method of verbal clauses which play an important role in sentences.

View full abstract

This paper presents a method for resolving ambiguity in translating a list of content words (termlist), such as a query for information retrieval or a keyword list that represents a document. The entire translation begins with retrieving a potential translation alternatives for each input word from a bilingual dictionary. Our disambiguation method, then, chooses one translation for each input word in such a way that the resulting list of translations are most coherent or related. The relatedness of translations is defined as the proximity among multi-dimensional vectors produced from the words on the basis of co-occurrence statistics in the target langauge corpora. The method was applied to term-lists extracted from newspaper articles and achieved 77.4% translation accuracy for ambiguous words.

View full abstract