ECRC: Emotion-Causality Recognition in Korean Conversation for GCN

TL;DR

The emotion-causality recognition in conversation model is proposed, which is based on a novel graph structure, which is based on a novel graph structure, thereby leveraging the strengths of both embedding methods.

Abstract

In this multi-task learning study on simultaneous analysis of emotions and their underlying causes in conversational contexts, deep neural network methods were employed to effectively process and train large labeled datasets. However, these approaches are typically limited to conducting context analyses across the entire corpus because they rely on one of the two methods: word- or sentence-level embedding. The former struggles with polysemy and homonyms, whereas the latter causes information loss when processing long sentences. In this study, we overcome the limitations of previous embeddings by utilizing both word- and sentence-level embeddings. Furthermore, we propose the emotion-causality recognition in conversation (ECRC) model, which is based on a novel graph structure, thereby leveraging the strengths of both embedding methods. This model uniquely integrates the bidirectional long short-term memory (Bi-LSTM) and graph neural network (GCN) models for Korean conversation analysis. Compared with models that rely solely on one embedding method, the proposed model effectively structures abstract concepts, such as language features and relationships, thereby minimizing information loss. To assess model performance, we compared the multi-task learning results of three deep neural network models with varying graph structures. Additionally, we evaluated the proposed model using Korean and English datasets. The experimental results show that the proposed model performs better in emotion and causality multi-task learning (74.62% and 75.30%, respectively) when node and edge characteristics are incorporated into the graph structure. Similar results were recorded for the Korean ECC and Wellness datasets (74.62% and 73.44%, respectively) with 71.35% on the IEMOCAP English dataset.

Figures (7)

• Figure 1: Illustration of homonyms and homographs. Homophones refer to words that have the same pronunciation but different meanings, whereas homonyms are words that have multiple meanings but the same form.
• Figure 2: Illustration of ECRC model structure. Initially, the embeddings are processed in the embedding layer. Then, in the GCN Layer, the processed data are transformed into a graph structure to simultaneously learn emotions and the causality that triggers them, for subsequent prediction of both classes. The GCN takes a graph structure with 1627 dimensions as input for training. In the hidden layer, information from neighboring nodes is combined to update the graph's hidden state. The readout step transforms the graph's features into a single vector, which serves as the output. This output predicts the label each class belongs to in the graph.
• Figure 3: Illustration of node and edge features. Each utterance is represented as $u_{i}$, and sequentially transformed into a graph structure $g_{i}$, where node and edge features are added. The dimensions of each node and edge feature are indicated, with the higher weight placed on the last sentence of the corpus, where labeling for emotion and causality is performed.
• Figure 4: Illustration of adjacency matrix of graph in Figure 3.
• Figure 5: Illustration of proportions of labels. This represents the labeling ratio of emotions and causality in the training data of the emotional conversation corpus, consisting of 5,312 samples. In the emotion category, 'Anger' accounts for 19.2%, while in the causality category, 'Personal Relationship' has the highest proportion at 20.0%. To address data bias, regularization and dropout techniques were applied during training.