Skip to content

yanpengt06/explore-chatGPT-in-Affective-Computing

Repository files navigation

Explore ChatGPT in Affective Computing

Introduction

We want to test the capacity of ChatGPT on Affective Computing. This repository contains the result of Dialogue Act Classification(DAC) for now and we will update more results in few weeks.

Result for DAC

Model Acc weighted-F1
Co-GAT - 79.4
ChatGPT, oneshot 0.67 0.65
ChatGPT, oneshot+prompt-engineering 0.71 0.70
ChatGPT, fewshot 0.73 0.72
ChatGPT, fewshot+prompt-engineering 0.74 0.73

Project Structure

  • cal_prf.py

    input: two label files which are chatGPT prediction and golden answer file

    output: the classification report using sklearn

  • check_utts

input: output file of chatGPT and golden answer file

output: the indexes of dialogue which are obviously wrongly annotated.

  • dd_label_extract.py: extract the labels from output file of chatGPT into a label file.

  • dd-dialogue-act.py: sample 64 dialogs in daily dialogue and transform the data format.

  • main.py: use ReverseChatGPT API to use chatGPT

About

test the capacity of ChatGPT in Dialogue Act Classification

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages