The paper presents two persona-based models for improving speaker consistency in neural response generation: the Speaker Model and the Speaker-Addressee Model. The Speaker Model encodes individual characteristics such as background information and speaking style into distributed embeddings, while the Speaker-Addressee Model captures interaction patterns between two interlocutors. These models are integrated into a sequence-to-sequence (Seq2Seq) framework, enhancing the generation of personalized and coherent responses. Experiments on Twitter conversations and TV series scripts show that these models improve BLEU scores by up to 20% and perplexity by up to 12%, with human judges also noting significant improvements in speaker consistency. The models leverage conversational data and side-information to learn speaker embeddings, which are then used to generate more contextually appropriate responses. The paper discusses related work, details the Seq2Seq models, and presents experimental results on various datasets, demonstrating the effectiveness of the proposed models in generating more human-like and consistent responses.The paper presents two persona-based models for improving speaker consistency in neural response generation: the Speaker Model and the Speaker-Addressee Model. The Speaker Model encodes individual characteristics such as background information and speaking style into distributed embeddings, while the Speaker-Addressee Model captures interaction patterns between two interlocutors. These models are integrated into a sequence-to-sequence (Seq2Seq) framework, enhancing the generation of personalized and coherent responses. Experiments on Twitter conversations and TV series scripts show that these models improve BLEU scores by up to 20% and perplexity by up to 12%, with human judges also noting significant improvements in speaker consistency. The models leverage conversational data and side-information to learn speaker embeddings, which are then used to generate more contextually appropriate responses. The paper discusses related work, details the Seq2Seq models, and presents experimental results on various datasets, demonstrating the effectiveness of the proposed models in generating more human-like and consistent responses.