本文選題：困惑度 + 病句�。� 參考：《計(jì)算機(jī)應(yīng)用研究》2017年12期

【摘要】：目前用來(lái)評(píng)價(jià)機(jī)器翻譯系統(tǒng)譯文質(zhì)量的方法主要有IBM提出的BLEU、TER和METEOR等方法,分別以詞匯的重現(xiàn)率、譯文與參考譯文之間的編輯距離和語(yǔ)言學(xué)知識(shí)等特征作為評(píng)價(jià)依據(jù),在判定中文句子的困惑度方面具有一定局限性。提出在依存語(yǔ)法分析的基礎(chǔ)上,通過(guò)對(duì)中文句子及其句子主干的語(yǔ)法和語(yǔ)義兩方面進(jìn)行分析得出中文句子的困惑度。實(shí)驗(yàn)證明這種方法比通過(guò)譯文加權(quán)改進(jìn)后的BLEU方法準(zhǔn)確率高出4%。
[Abstract]:At present, the methods used to evaluate the quality of the Machine Translation system mainly include the methods of BLEU, TER and METEOR proposed by IBM, which are based on the repetition rate of vocabulary, the editing distance between the translation and the reference translation and the linguistic knowledge, and have some limitations in determining the confusion of Chinese sentences. On the basis of grammatical analysis, the perplexity of Chinese sentences is obtained by analyzing the two aspects of the grammar and semantics of the Chinese sentence and the main sentence of the sentence. The experiment proves that this method is more accurate than the BLEU method which is improved by the translation weighted 4%..

【作者單位】： 山東師范大學(xué)信息科學(xué)與工程學(xué)院;山東省分布式計(jì)算軟件新技術(shù)重點(diǎn)實(shí)驗(yàn)室;
【基金】：國(guó)家自然科學(xué)基金資助項(xiàng)目(61672329,61373149,61472233,61572300,81273704) 山東省科技計(jì)劃資助項(xiàng)目(2014GGX101026) 山東省教育科學(xué)規(guī)劃資助項(xiàng)目(ZK1437B010) 山東省泰山學(xué)者基金資助項(xiàng)目(TSHW201502038,20110819) 山東省精品課程資助項(xiàng)目(2012BK294,2013BK399,2013BK402)
【分類號(hào)】：TP391.1
，

本文編號(hào)：1803728