Affective Neural Response Generation
Nabiha Asghar†, Pascal Poupart†, Jesse Hoey†, Xin Jiang‡, Lili Mou† †Cheriton School of Computer Science, University of Waterloo, Canada ‡Noah’s Ark Lab, Huawei Technologies, Hong Kong
Abstract
Existing neural conversational models process natural language primarily on a lexico-syntactic level, thereby ignoring one of the most crucial components of human-to-human dialogue: its affective content. We take a step in this direction by proposing three novel ways to incorporate affective/emotional aspects into long short term memory (LSTM) encoder-decoder neural conversation models: (1) affective word embeddings, which are cognitively engineered, (2) affect-based objective functions that augment the standard cross-entropy loss, and (3) affectively diverse beam search for decoding. Experiments show that these techniques improve the open-domain conversational prowess of encoder-decoder networks by enabling them to produce emotionally rich responses that are more interesting and natural.
中文速览
现有神经对话模型只关注词汇和句法层面，完全忽视了人与人交流中至关重要的情感信息，导致生成的回复往往平淡、无聊甚至情绪脱节。为此，研究者提出了三种将情感注入LSTM编解码对话模型的方法：一是将每个词映射到由心理学实验标注的"效价-唤醒度-支配度"（VAD）三维情感空间，作为额外输入；二是在标准交叉熵损失之外叠加情感目标函数，让模型在训练时被明确引导去生成有情感倾向的回复；三是在解码阶段采用情感多样化的束搜索，主动筛选出更具情感色彩的候选回复。实验表明，这三种方法单独使用或组合使用，都能让模型产出更自然、更丰富、更有感染力的对话回复。这项工作的意义在于，它开创性地把认知心理学中的情感评分体系与神经对话生成框架打通，为构建真正能与人进行情感共鸣交流的对话系统迈出了重要一步。
原文 arXiv:1709.03968;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1709.03968v1