石旺君



摘 要:為了提高機(jī)器翻譯中英語語言轉(zhuǎn)換的自動化和智能化水平,提出了一種基于深度學(xué)習(xí)和語言轉(zhuǎn)換錯(cuò)誤特征提取的機(jī)器翻譯錯(cuò)誤消除方法,其核心差分語義修正方法。建立了英語語言轉(zhuǎn)換翻譯中錯(cuò)誤排除的語義相關(guān)檢測模型,并通過語法分析,建立了英語語言轉(zhuǎn)換翻譯中錯(cuò)誤排除的語義樹。提取了英語語言轉(zhuǎn)換的語義相似度特征。根據(jù)語義相似度的不同組合,對英語語言轉(zhuǎn)換中的語義分配和機(jī)器翻譯錯(cuò)誤特征進(jìn)行了分析。采用深度學(xué)習(xí)方法建立了英語轉(zhuǎn)換的樹形主題詞表,并根據(jù)樹形主題詞表中的語義修改目標(biāo)調(diào)整了英語轉(zhuǎn)換的句子結(jié)構(gòu)。為了消除英語轉(zhuǎn)換過程中的翻譯錯(cuò)誤,計(jì)算出每個(gè)小句的最佳語義相關(guān)特征,并采用深度學(xué)習(xí)算法對英語轉(zhuǎn)換過程中的翻譯錯(cuò)誤進(jìn)行自動優(yōu)化。仿真結(jié)果表明,該方法精度高。
關(guān)鍵詞:深度學(xué)習(xí);語言翻譯;英語;語義學(xué);英語機(jī)器翻譯;翻譯錯(cuò)誤;錯(cuò)誤消除
中圖分類號:TP 391
文獻(xiàn)標(biāo)志碼:A
文章編號:1007-757X(2020)11-0055-04
Abstract:In order to improve the level of automation and intelligence of English language conversion in machine translation, a machine translation error elimination method based on deep learning and language conversion error feature extraction is proposed. Its core bit difference semantic correction method establishes a semantic correlation detection model for error elimination in English language translation. Through grammatical analysis, it establishes a semantic tree for error elimination in English language translation. The semantic similarity features of English language conversion are extracted. According to different combinations of semantic similarity, the semantic distribution and machine translation error features in English language conversion are analyzed. A deep-learning method is used to establish a tree-shaped thesaurus in English conversion, and the sentence structure of English conversion is adjusted according to the semantic modification target in the tree-shaped thesaurus. In order to eliminate translation errors during the English conversion process, the best semantically relevant features of each clause are calculated, and deep learning algorithms are used to automatically optimize the translation errors during the English conversion process. Simulation results show that the method has high accuracy.
Key words:deep learning;language translation;English;semantics;English machine translation;translation errors;error elimination
0?引言
隨著自動翻譯軟件的發(fā)展和成熟,對機(jī)器翻譯軟件的翻譯和校準(zhǔn)的準(zhǔn)確性要求更高。在機(jī)器翻譯環(huán)境中,語義分析需要與自動翻譯軟件的上下文特征結(jié)合起來[1-2]。根據(jù)語義相似度,進(jìn)行自動翻譯和標(biāo)定,以提高自動翻譯的語義分配精度。在語義異構(gòu)的條件下,機(jī)器翻譯的自動校準(zhǔn)主要通過語義相似度的概念分析來實(shí)現(xiàn)[3-4]。摘要自動翻譯文本的相關(guān)性特征和語義相似性特征,根據(jù)相似語義對存在的語義異構(gòu)性進(jìn)行消除,實(shí)現(xiàn)英語翻譯過程中的知識結(jié)構(gòu)圖分析,并構(gòu)建語義概念樹[5-6]。……