Code and results data underlying Chapter 4 of the PhD thesis: "Towards predicting memory in multimodal group interactions"

This repository contains the complete code and experiment outputs for Chapter 4 of M. Tsfasman's PhD thesis "Towards predicting memory in multimodal group interactions". It provides Python scripts and notebooks for preprocessing multimodal data, training and evaluating machine learning models to predict memorable moments in conversations using features such as eye-gaze and speaker activity, and performing feature ablation studies to assess the importance of each input. The repository includes robust session-based cross-validation, hyper-parameter optimization, and tools for result visualization. It is designed for reproducible research and is an extended, methodologically improved version of a published ICMI 2022 paper*, supporting both local and cluster-based execution.<br><em>*M Tsfasman, K Fenech, M Tarvirdians, A Lorincz, C Jonker, C Oertel, "Towards creating a conversational memory for long-term meeting support: predicting memorable moments in multi-party conversations through eye-gaze,'' in Proc. International Conference on Multimodal Interaction (ICMI)}, pp. 94–104, 2022.</em>

本仓库收录了M. Tsfasman博士学位论文《面向多模态群体交互的记忆预测》（Towards predicting memory in multimodal group interactions）第4章的完整代码与实验成果。其提供了用于预处理多模态（multimodal）数据的Python脚本与Notebook，可基于眼动注视（eye-gaze）、说话者活跃度等特征，训练并评估机器学习模型以预测对话中的难忘时刻，同时支持开展特征消融实验以评估各输入特征的重要性。本仓库集成了鲁棒的基于会话的交叉验证、超参数优化工具以及结果可视化工具，旨在支持可复现研究，同时作为已发表的国际多模态交互会议（International Conference on Multimodal Interaction, 简称ICMI）2022会议论文*的扩展与方法论优化版本，支持本地与集群环境运行。<br><em>*M Tsfasman, K Fenech, M Tarvirdians, A Lorincz, C Jonker, C Oertel, "Towards creating a conversational memory for long-term meeting support: predicting memorable moments in multi-party conversations through eye-gaze," in Proc. International Conference on Multimodal Interaction (ICMI), pp. 94–104, 2022.</em>