本文選題：交互式機(jī)器翻譯 + 詞對(duì)齊　； 參考：《中文信息學(xué)報(bào)》2017年04期

【摘要】：在傳統(tǒng)的機(jī)器翻譯(machine translation,MT)與計(jì)算機(jī)輔助翻譯(computer aided translation,CAT)中,譯員與翻譯引擎之間的交互受到很大限制,于是產(chǎn)生了交互式機(jī)器翻譯(interactive machine translation,IMT)技術(shù)。但傳統(tǒng)的模型只考慮當(dāng)前源語(yǔ)與部分翻譯的目標(biāo)語(yǔ)的信息,沒(méi)有將用戶交互后的對(duì)齊信息加入到之后的預(yù)測(cè)模型中。該文基于詞預(yù)測(cè)交互式機(jī)器翻譯的研究思路,將用戶交互翻譯過(guò)程中的鼠標(biāo)點(diǎn)選行為轉(zhuǎn)化為中間譯文的詞對(duì)齊信息,進(jìn)而在翻譯交互過(guò)程中實(shí)現(xiàn)了對(duì)譯文的動(dòng)態(tài)詞對(duì)齊標(biāo)注,并在詞對(duì)齊信息和輸入譯文的約束下提高了傳統(tǒng)詞預(yù)測(cè)的準(zhǔn)確性。
[Abstract]:In the traditional Machine Translation (machine translation, MT) and computer aided translation (CAT), the interaction between the interpreter and the translation engine is greatly limited, so the interactive Machine Translation (Interactive machine translation, IMT) technology is produced. But the traditional model only considers the current source language and the partial translation. The information of the target language is not added to the prediction model after the user interaction. Based on the research idea of the word prediction interactive Machine Translation, this paper transforms the mouse point selection behavior in the interactive translation process into the word alignment information in the intermediate translation and realizes the translation in the process of translation interaction. State word alignment and annotation improves the accuracy of traditional word prediction under the constraints of word alignment information and input translation.
【作者單位】： 沈陽(yáng)航空航天大學(xué)人機(jī)智能研究中心;
【基金】：國(guó)家自然科學(xué)基金(61403262,61402299)
【分類號(hào)】：H085
，

本文編號(hào)：2079538