- 學位論文
人工智慧情感對話機器人
Artificial Intelligence Affective Conversational Robot
摘要
近年來交談機器人(ChatBot)已成為各領域所廣泛運用的技術之一。為了與用戶有更好要互動，提升交談機器人在對話時的溫度，也變成各交談機器人的課題之一。 本研究的對話模型使用檢索式模型與生成式模型做為主要對話訓練模型，情感分析模型則是使用MLP、LSTM與BiLSTM三種模型，以Word2Vec與Semantic做為相似度模型的比較，最後將對話模型、情感分析模型與相似度模型三種模型整合比較。本研究也提出情感對話機器人指數(Affective Conversational Robot Index; ACR Index)做為評估情感對話機器人的標準，最後實驗結果顯示情感分析預測上情感分析模型使用BiLSTM，相似度模型使用Word2Vec，對話模型使用檢索式模型的對話效果最為出色。
並列摘要
The ChatBot has become one of technologies using in various applications. For better interaction between ChatBot and user, programing the humanity in Chatbot is one of main subject in this application. This research uses the Retrieval-base model & Generative Model for main dialogue developing model. In sentiment analysis model, use MLP, LSTM and BiLSTM for training. Comparison between similarity models are based on Word2Vec and Semantic. At the end, summarize the dialogue model, sentiment analysis model and similarity model. Besides, this study applies ACR Index as standard for evaluating affective dialogue. The study result demonstrates sentiment analysis model using BiLSTM has the most outstanding feedback in sentiment analysis prediction. Similarity model using Word2Vec and Dialogue model using Retrieval-base model have better dialogue effect.