Topic-Driven and Knowledge-Aware Transformer for Dialogue Emotion Detection

Zhu, Lixing; Pergola, Gabriele; Gui, Lin; Zhou, Deyu; He, Yulan

doi:10.48550/arxiv.2106.01071

Cited by 10 publications

(16 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The code framework and initial weight of Sim-CSE come from Huggingface's Transformers(Wolf Models IEMOCAP MELD EmoryNLP COSMIC (Ghosal et al, 2020) 65.28 65.21 38.11 DialogueCRN (Hu et al, 2021) 66.46 63.42 38.91 DAG-ERC 68.03 63.65 39.02 TODKAT (Zhu et al, 2021) 61.33 65.47 38.69 Cog-BART (Li et al, 2021) 66.18 64.81 39.04 TUCORE-GCN_RoBERTa (Lee and Choi, 2021) -65.36 39.24 SGED + DAG-ERC (Bao et al, 2022) 68 et al, 2020). We use the AdamW optimizer and cosine learning rate schedule strategy.…”

Section: Methodsmentioning

confidence: 99%

“…DAG-ERC uses a directed acyclic graph (DAG) to model the intrinsic structure within a conversation. Knowledgeenhanced methods (Zhong et al, 2019;Zhu et al, 2021;Ghosal et al, 2020; usu-ally utilize external knowledge from ATIMOC (Sap et al, 2019) or ConceptNet (Liu and Singh, 2004). Besides individual models, several frameworks have also been proposed.…”

Section: Emotion Recognition In Conversationmentioning

confidence: 99%

“…With the development of online social networks, capturing emotions during conversations has gained increasing attention in both academia and industry (Li et al, 2020;Ghosal et al, 2020;Song et al, 2022;Zhu et al, 2021). Emotion recognition in conversation (ERC) is critical in many scenarios, such as chatbots, healthcare applications, mining opinions on social media, and so on (Poria et al, 2019b).…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Supervised Prototypical Contrastive Learning for Emotion Recognition in Conversation

Song¹,

Huang²,

Han³

et al. 2022

Preprint

View full text Add to dashboard Cite

Capturing emotions within a conversation plays an essential role in modern dialogue systems. However, the weak correlation between emotions and semantics brings many challenges to emotion recognition in conversation (ERC). Even semantically similar utterances, the emotion may vary drastically depending on contexts or speakers. In this paper, we propose a Supervised Prototypical Contrastive Learning (SPCL) loss for the ERC task. Leveraging the Prototypical Network, the SPCL targets at solving the imbalanced classification problem through contrastive learning and does not require a large batch size. Meanwhile, we design a difficulty measure function based on the distance between classes and introduce curriculum learning to alleviate the impact of extreme samples. We achieve state-of-the-art results on three widely used benchmarks. Further, we conduct analytical experiments to demonstrate the effectiveness of our proposed SPCL and curriculum learning strategy. We release the code at https://github.com/caskcsg/SPCL.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Emotion Recognition In Conversationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Supervised Prototypical Contrastive Learning for Emotion Recognition in Conversation

Song¹,

Huang²,

Han³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…A directed acyclic graph (DAG) based ERC was introduced by Shen et al in [27] which is an attempt to combine the strengths of conventional graph-based and recurrence-based neural networks. In [38], Zhu et al propose a new model in which the transformer model fuses the topical and commonsense information to predict the emotion label. Recently, Song et al [29] proposed the EmotionFlow model, which encodes the user's utterances via concatenating the context with an auxiliary question, and then, a random field is applied to capture the sequential information at the emotion level.…”

Section: Text-based Methodsmentioning

confidence: 99%

M2FNet: Multi-modal Fusion Network for Emotion Recognition in Conversation

Chudasama¹,

Kar²,

Gudmalwar³

et al. 2022

Preprint

View full text Add to dashboard Cite

Emotion Recognition in Conversations (ERC) is crucial in developing sympathetic human-machine interaction. In conversational videos, emotion can be present in multiple modalities, i.e., audio, video, and transcript. However, due to the inherent characteristics of these modalities, multi-modal ERC has always been considered a challenging undertaking. Existing ERC research focuses mainly on using text information in a discussion, ignoring the other two modalities. We anticipate that emotion recognition accuracy can be improved by employing a multi-modal approach. Thus, in this study, we propose a Multi-modal Fusion Network (M2FNet) that extracts emotion-relevant features from visual, audio, and text modality. It employs a multi-head attention-based fusion mechanism to combine emotion-rich latent representations of the input data. We introduce a new feature extractor to extract latent features from the audio and visual modality. The proposed feature extractor is trained with a novel adaptive marginbased triplet loss function to learn emotion-relevant features from the audio and visual data. In the domain of ERC, the existing methods perform well on one benchmark dataset but not on others. Our results show that the proposed M2FNet architecture outperforms all other methods in terms of weighted average F1 score on well-known MELD and IEMOCAP datasets and sets a new state-of-theart performance in ERC.

show abstract

“…In their experimentation, they established that BERT is competitive with other language models such as XLNet, however, can be improved. They showed RoBERTa performed better than BERT on GLUE and SQuAD Zhu et al (2021). in their study, inserted topic layer into fine-tuned RoBERTa and thus, proposed topic augmented language model for topic extraction.…”

mentioning

confidence: 92%

Statistical learning approaches to sentiment analysis in the Nigerian banking context

Ogunleye¹

View full text Add to dashboard Cite

Banking is an essential component of our day-to-day activity, and the sector contributes to the development of every Nation's economy. To ensure bank stability, the banks have gone through several deregulation and reformation processes. Unfortunately, these processes and technological advancement have led to increased competitiveness, saturated market, and low profitability. Thus, the need for banks to build customer centric service to gain profitability and stability is vital. This opens the need to investigate customers attitude towards banking.With the growing usage of social network sites (SNS), the user generated content (UGC) has opened opportunities for banks to mine customers opinion . This can help the banks to generate insight into their product and services, create marketing strategies and manage their reputation. To the customers, this serves as source to information that can help in supporting decision making on product or service purchase. In this study, sentiment analysis (SA) techniques were employed to investigate customers attitude towards banking us ing Twitter data. However, the unstructured nature of the data and word ambiguity made sentiment analysis complicated. In the context of this study, SA is more difficult because it involves the natural language processing of Pidgin English and English words in the bank domain.Unfortunately, there are limited or no resources for this purpose.

show abstract

Topic-Driven and Knowledge-Aware Transformer for Dialogue Emotion Detection

Cited by 10 publications

References 26 publications

Supervised Prototypical Contrastive Learning for Emotion Recognition in Conversation

Supervised Prototypical Contrastive Learning for Emotion Recognition in Conversation

M2FNet: Multi-modal Fusion Network for Emotion Recognition in Conversation

Statistical learning approaches to sentiment analysis in the Nigerian banking context

Contact Info

Product

Resources

About