基于智能语音交互的文言文翻译机器人关键技术

Key technologies of classical Chinese translation robot based on intelligent voice interaction

以语音交互为基础的翻译机器人,依托于固定关键词生成交互命令,使得文言文翻译结果双语替换测评(Bilingual Evaluation Understudy,BLEU)值较低。因此,提出基于智能语音交互的文言文翻译机器人关键技术。运用自然语言处理机制,对原始语音进行分词、词性标注等多方面处理。提出自动语音识别技术,结合语言模型和声学模型,对待翻译文言文语义进行准确预测,生成相应的交互命令。以循环神经网络为核心,构建机器翻译模型,并引入注意力机制,将网络单元改为双向循环模式,获取精准翻译结果。实验结果显示:所提文言文翻译机器人关键技术在实际应用中,使得BLEU值提升了31.47%、22.01%,更好地满足文言文翻译质量要求。

The translation robot based on voice interaction generates interactive commands based on fixed keywords,which makes the Bleu value of classical Chinese translation results low.Therefore,the key technology of classical Chinese translation robot based on intelligent voice interaction is proposed.Using natural language processing mechanism,the original voice is processed in many aspects,such as word segmentation,part of speech tagging and so on.Automatic speech recognition technology is proposed,which combines language model and acoustic model to accurately predict the semantics of translated classical Chinese and generate corresponding interactive commands.Taking the cyclic neural network as the core,this paper constructs the machine translation model,introduces the attention mechanism,and changes the network unit into a two-way cyclic mode to obtain accurate translation results.The experimental results show that the proposed key technology of classical Chinese translation robot has improved the Bleu value by 31.47% and 22.01% in practical application,which can better meet the quality requirements of classical Chinese translation.