Is Discourse Role Important for Emotion Recognition in Conversation?
Authors: Donovan Ong, Jian Su, Bin Chen, Anh Tuan Luu, Ashok Narendranath, Yue Li, Shuqi Sun, Yingzhan Lin, Haifeng Wang11121-11129
AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experiments show that our proposed method beats the best-reported performances on three public Emotion Recognition in Conversation datasets. This proves that the discourse role information of an utterance plays an important role in the emotion recognition task, which no previous work has studied. |
| Researcher Affiliation | Collaboration | Donovan Ong1,2,3, Jian Su1, Bin Chen1, Anh Tuan Luu3*, Ashok Narendranath1, Yue Li1, Shuqi Sun4, Yingzhan Lin4, Haifeng Wang4 1Institute for Infocomm Research, A*STAR, Singapore 2CNRS@CREATE LTD, Singapore 3School of Computer Science and Engineering, Nanyang Technological University, Singapore 4Baidu Inc., China |
| Pseudocode | No | The paper describes the model architecture using mathematical equations and text, but it does not include any pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any concrete statements or links indicating that the source code for the described methodology is publicly available. |
| Open Datasets | Yes | We evaluate our model on three publicly available datasets, differing in magnitudes of size. We present the summary of statistics for the datasets used in our experiments in Table 1. Dialy Dialog (Li et al. 2017) is a dyadic text-based dialog dataset based on daily written communications. MELD (Poria et al. 2019) is a multiparty multi-modal dialog dataset from the Friends TV series. IEMOCAP (Busso et al. 2008) is dyadic multi-modal dialog dataset based on videos of two-way conversations. |
| Dataset Splits | Yes | Table 1: Dataset # dialogues / # utterances train val test Daily Dialog 11,118 / 87,832 1,000 / 7,912 1,000 / 7863 7* Micro-F1 MELD 1,039 / 9,989 114 / 1,109 280 / 2,610 7 Weighted Avg. F1 IEMOCAP 108 / 5,236 12 / 574 31 / 1,623 6 Weighted Avg. F1 |
| Hardware Specification | No | The paper does not specify any details about the hardware used to run the experiments, such as GPU models, CPU types, or memory. |
| Software Dependencies | No | The paper mentions 'PyTorch' and 'Spacy' tokenizers, and initializes with 'RoBERTa BASE weights', but it does not provide specific version numbers for any of these software components. |
| Experiment Setup | Yes | Table 2: Hyperparameter settings. C is the hidden size of the conversation-level bidirectional GRU, D is the number of latent discourse roles, M is the discourse role and word embedding dimension. Table includes specific values for C, D, M, λdiscourse, and learning rate for each dataset. |