[人工智能]论文阅读 2022 AAAI Hybrid Curriculum Learning for Emotion Recognition in Conversation

Hybrid Curriculum Learning for Emotion Recognition in Conversation

改论文发表于2022 AAAI，阿里巴巴团队，旨在用课程学习解决对话情感识别

动机：

Motivated by recent studies which have proven that feeding training examples in a meaningful order rather than considering them randomly can boost the performance of models, we propose an ERCoriented hybrid curriculum learning framework.

贡献：

1. We propose a hybrid curriculum learning framework to tackle the task of ERC. At conversation-level curriculum, we utilize an emotion-shift frequency to measure the difficulty of each conversation.
2. We propose emotion-similarity based curriculum learning to achieve utterance-level curriculum learning. It implements the basic idea that at early stage of training it is less important to distinguish between similar emotions compared to separating very different emotions
3. 理解一下：在训练的早期，对于识别不同的情感比识别区分相似的情感更重要。所以文章的方法为啥能达到这种效果？
4. 实验结果的有效性，实现了SOTA

方法：

(1) conversation-level curriculum (CC) – we construct a difficulty measurer based on “emotion shift”frequency. The conversations are scheduled in an “easy to hard” schema according to the difficulty score returned by the difficulty measurer.

(2) utterance-level curriculum (UC). it is implemented from an emotion-similarity perspective, which progressively strengthens the model’s ability in identifying the confusing emotions

UC很容易理解，就是常规的课程学习方法；UU的动机是说一个对话中的句子是具有逻辑性的，顺序是无法改变的，怎么解决这个问题呢？

作者assuming that the utterances with confusing emotion labels are more difficult for prediction;
and our utterance-level curriculum is based on the pairwise similarities between the emotion labels.

见下图：

首先计算一个情感矩阵：

进行归一化之后，得到目标情感矩阵概率分布，然后计算loss：

我理解的这个loss和之前的model相比应该就是增加了一个$M_{target}$，可是和动机的关系是？没太理解。

整个过程的训练算法如下：

问题总结：

第一，对于动机2着实没理解透彻，为啥引入一个$M_{target}$，可以起到课程学习的作用？解决方法和motivation之间的关联性是什么？
第二，对于算法中的第9~11行之间的矩阵的更新，我的理解是让这个矩阵对角线上逐步趋近于1，然而这个更新的方式也是没太明白。

有看明白的同学可以解惑一下，感谢！

更多有趣文章见：
利用逆向思维的机器阅读理解
证据推理网络
Bert预训练模型-中文文本分类