dialogue-summarization

Code and dataset for our Bioinformatics 2022 paper: "A Benchmark for Automatic Medical Consultation System: Frameworks, Tasks and Datasets"

MEDIQA-Chat Shared Tasks @ ACL-ClinicalNLP 2023

EMNLP 2022: ClidSum: A Benchmark Dataset for Cross-Lingual Dialogue Summarization

Revisiting Cross-Lingual Summarization: A Corpus-based Study and A New Benchmark with Improved Annotation

[EMNLP 2023] FaMeSumm: Investigating and Improving Faithfulness of Medical Summarization. Support BART, PEGASUS, T5, mT5, BioBART, etc.

Dialogue Summarization application hosted using AWS and CICD deployment with docker and FASTAPI. Model card created in HuggingFace and a deployed on HuggingFace Spaces.

A dataset focused on summarization of dialogs, which represents the rich domain of Twitter customer care conversations.

For this project, I fine-tuned two separate models for three tasks: document summarization, dialogue summarization and text classification

GEN AI use case: dialogue summary. This notebook is extracted from the course Generative AI with Large Language Models. It is used to understand how input text can affect model performance.

This notebook fine-tunes the FLAN-T5 model for dialogue summarization, comparing full fine-tuning with Parameter-Efficient Fine-Tuning (PEFT). It evaluates performance using ROUGE metrics, demonstrating PEFT's efficiency while achieving competitive results.

💬 An end-to-end NLP project where I fine-tuned Google's T5-Small transformer on the SAMSum dataset to generate abstractive dialogue summaries, served locally with FastAPI.

This repository explores enhancing dialogue summarization with commonsense knowledge through the SICK framework, evaluating models on dialogue datasets to assess commonsense's impact on summarization quality.

Instruction fine tuning BART for Dialogue Summarization | IT4772E | NLP Project 20232

A FastAPI web application that generates concise summaries from long text using a fine-tuned T5 Transformer model.

FastAPI app that summarizes dialogue and text using a pretrained Transformer model (BART/T5, via Hugging Face). Includes a clean two-pane web UI with live compression stats, sample prompts, and a REST API.

Successfully developed a text summarization model using Seq2Seq with attention to condense multi-turn dialogues from the SAMSum dataset into coherent and informative summaries.

Fine-tuning FLAN-T5 with PPO and PEFT to generate less toxic text summaries. This notebook leverages Meta AI's hate speech reward model and utilizes RLHF techniques for improved safety.

Fine-tuning de T5 com LoRA para sumarização de diálogos, com avaliação ROUGE, ablação de rank, demo, testes, CI e Docker