Empirical Analysis of Dialogue Relation Extraction with Large Language Models

Authors: Guozheng Li, Zijie Xu, Ziyu Shang, Jiajun Liu, Ke Ji, Yikai Guo

IJCAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct experiments on the two versions (V1 and V2) of Dialog RE dataset [Yu et al., 2020], the first human-annotated DRE dataset, originating from the complete transcripts of the series Friends. We compare the performances of generation-based methods with previous sequence-based and graph-based methods, and conduct extensive experiments to provide valuable insights and guide further exploration.
Researcher Affiliation Academia 1School of Computer Science and Engineering, Southeast University 2Beijing Institute of Computer Technology and Application
Pseudocode No The paper describes methods and processes but does not include any explicitly labeled pseudocode or algorithm blocks.
Open Source Code Yes 3https://github.com/1iguozheng/Landre
Open Datasets Yes We conduct experiments on the two versions (V1 and V2 2) of Dialog RE dataset [Yu et al., 2020], the first human-annotated DRE dataset, originating from the complete transcripts of the series Friends. 2https://dataset.org/dialogre/
Dataset Splits Yes Dialog RE Train Dev Test # Conversations 1,073 358 357 Average dialogue length 229.5 224.1 214.2 # Argument pairs 5,963 1,928 1,858 Average # of turns 13.1 13.1 12.4 Average # of speakers 3.3 3.2 3.3
Hardware Specification Yes all experiments are conducted on a single Geforce GTX 3090 GPU.
Software Dependencies No The paper mentions various models and optimizers (e.g., GPT-2, BART, T5, BLOOM, LLaMA, AdamW, LoRA) but does not provide specific version numbers for the programming languages, libraries, or frameworks used (e.g., Python version, PyTorch version, Hugging Face Transformers version).
Experiment Setup Yes We set the rank r of the Lo RA parameters to 8 and the merging ratio α to 32. The model is optimized with Adam W [Loshchilov and Hutter, 2019] using learning rate 1e-4 with a linear warm up [Goyal et al., 2017] for the first 6% steps followed by a linear decay to 0. We train Landre for 5 epochs with batch size 4