Table 1 shows an example Japanese sentence of the speech input data.
「私は彼の高校を受験する.」 |
``Watashiwa kareno koukouwo jyukeNsuru.'' |
` I will take an entrance examination for his high school.' |
We used the word bigram model to generate the 2-best output sentences (table 2).
order | output sentence | |||
1 | 「私からの高校と受験する.」 (UGS) | |||
``watashikarano koukouto jyukeN suru.'' (UGS) | ||||
`I take examination with a senior high school from me.' | ||||
(UGS) | ||||
2 | 「私は彼の高校を受験する.」 | |||
``watashiha kareno koukouwo jyukeN suru.'' | ||||
`I take an examination to his senior high school.' |
Table 3 shows the valency patterns for the Japanese verb 「受験する」('take an examination'). In this table, Japanese or English words in parentheses mean the semantic categories for the N's.
1 | 「 ( 人) [は/が] (362 組織) を 受験する」 |
` ( human ) [ha/ga] (362 Organization ) take an examination ' | |
2 | 「 ( 人) [は/が] (1426 試験) を 受験する」 |
` ( human) [ha/ga] (1426 Examination ) take an examination ' | |
3 | 「 ( 人) [は/が] 受験する」 |
` ( human) [ha/ga]take an examination ' |
Because none of the valency patterns in Table 3 match the first candidate, it must be discarded as an ungrammatical sentence. Whereas, since one of the valency patterns in Table 3 matches the second candidate, it becomes the first candidate.