abhisek-ai / cogmen Goto Github PK

COntextualized GNN based Multimodal Emotion recognitioN Picture: COGMEN Model Architecture

This repository contains the official Pytorch implementation of the following paper:

COGMEN: COntextualized GNN based Multimodal Emotion recognitioN

Paper: https://arxiv.org/abs/2205.02455

Authors: Abhinav Joshi, Ashwani Bhat, Ayush Jain, Atin Vikram Singh, Ashutosh Modi

Abstract: Emotions are an inherent part of human interactions, and consequently, it is imperative to develop AI systems that understand and recognize human emotions. During a conversation involving various people, a person’s emotions are influenced by the other speaker’s utterances and their own emotional state over the utterances. In this paper, we propose COntextualized Graph Neural Network based Multimodal Emotion recognitioN (COGMEN) system that leverages local information (i.e., inter/intra dependency between speakers) and global information (context). The proposed model uses Graph Neural Network (GNN) based architecture to model the complex dependencies (local and global information) in a conversation. Our model gives state-of-theart (SOTA) results on IEMOCAP and MOSEI datasets, and detailed ablation experiments show the importance of modeling information at both levels

• We use PyG (PyTorch Geometric) for the GNN component in our architecture. RGCNConv and TransformerConv

• We use comet for logging all our experiments and its Bayesian optimizer for hyperparameter tuning.

• For textual features we use SBERT.

• Install PyTorch Geometric

• Install Comet.ml

• Install SBERT

    python preprocess.py --dataset="iemocap_4"

    python train.py --dataset="iemocap_4" --modalities="atv" --from_begin --epochs=55

    python eval.py --dataset="iemocap_4" --modalities="atv"

    @inproceedings{COGMEN-2022,
    title = “{COGMEN:} {CO}ntextualized {G}NN based {M}ultimodal {E}motion recognitio{N}”,
    author = {Joshi, Abhinav and Bhat, Ashwani and Jain, Ayush and Singh, Atin Vikram and Modi, Ashutosh},
    booktitle = "Proceedings of the 2022 Conference of the North {A}merican Chapter of the Association for Computational Linguistics: Human Language Technologies, {NAACL-HLT}",
    month = jul,
    year = “2022”,
    address = "Seattle, Washington",
    publisher = "Association for Computational Linguistics",
    }

The structure of our code is inspired by pytorch-DialogueGCN-mianzhang.