Building controlled bilingual terminologies for the municipal domain and evaluating them using a coverage estimation approach

Abstract

This article examines the status of constructed controlled terminologies from the perspective of the coverage of terms/concepts. To facilitate controlled authoring of Japanese texts of the municipal domain and promote machine translatability into English, we constructed terminologies in the following way: (1) Japanese-English term pairs are extracted from aligned texts; (2) term variations are controlled by defining preferred and proscribed terms for both languages. To assess the coverage of the constructed terminologies, we propose a quantitative extrapolation method that estimates the potential vocabulary size. The coverage estimations show that the coverage of terms for Japanese is higher than that for English by about 10%, which reflects the greater diversity of the translated English terms. The coverage of concepts reaches around 60% for both Japanese and English. The method also enables us to quantitatively estimate how much effort is needed to further increase the coverage.