While the sortal constraints associated with Japanese numeral classifiers are well studied, less attention has been paid to the details of their syntax. We describe an analysis implemented within a broadcoverage HPSG that handles an intricate set of numeral classifier construction types and compositionally relates each to an appropriate semantic representation, using Minimal Recursion Semantics.

In this paper, we present a solution to the problem of generating both Japanese and Korean numeral classifiers using semantic classes from an ontology. Most nouns must use a numeral classifier when they are quantified in languages such as Chinese, Japanese, Korean, Malay and Thai. In order to select an appropriate classifier, we propose an algorithm which associates classifiers with semantic classes and uses inheritance to list only exceptional classifiers with individual nouns. The algorithm generates sortal classifiers with an accuracy of 81%. We reuse the ontology provided by Goi-Taikei --- a Japanese lexicon, and show that it is a reasonable choice for this task, requiring information to be entered for less than 6% of individual nouns. Keywords: multilingual generation, numeral classifiers, Japanese, Korean 1

A severe problem for NLP applications dealing with multilingua language resources is the acquisition of knowledge that is obligatory in one language but not explicitly expressed in another language. In this paper, we focus on numeral classifiers, which are required in languages like Japanese but are usually not explicitly used in languages like English, which don't have such a classifier system.

We present a method for combining two bilingual lexicons to make a third, using one language as a pivot. In this case we combine a Japanese-English lexicon with a Malay-English lexicon, to produce a Japanese-Malay lexicon suitable for use in a machine translation system. Our method di#ers from previous methods in its use of semantic classes to rank translation equivalents: word pairs with compatible semantic classes are preferred to those with dissimilar classes. We have made a prototype lexicon of over 75,000 pairs, and are now in the process of removing inappropriate entries by hand. 1

We examine the capacity of Web and corpus frequency methods to predict preferred count classifiers for nouns in Malay. The observed F-score for the Web model of 0.671 considerably outperformed corpus-based frequency and machine learning models. We expect that this is a fruitful extension for Web–as–corpus approaches to lexicons in languages other than English, but further research is required in other South-East and East Asian languages. 1

The so-called floating quantifier constructions in languages like Korean display intriguing properties whose successful processing can prove the robustness of a parsing system. This paper shows that a constraint-based analysis, in particular couched upon the framework of HPSG, can offer us an efficient way of analyzing these constructions together with proper semantic representations. It also shows how the analysis has been successfully implemented in the LKB (Linguistic Knowledge Building) system. 1