We have previously designed a methodology for semi-automatic grammar induction from un-annotated corpora belonging to a restricted domain. The induced grammar contains both semantic and syntactic structures, and experiments with the Air Travel Information Service (ATIS-3) corpus demonstrated the viability of our approach [1] for natural language understanding. This work explores the portability of our grammar induction approach to Chinese, based on a corpus of translated ATIS-3 queries. To assess grammar quality, we developed a framework bi-directional English-Chinese example-based machine translation using the induced grammars. Our translation framework can handle word order differences between the language pair during translation. Translations based on the ATIS-3 test sets showed a high percentage (76% to 91%) of user-accepted translations.