https://github.com/thefloatingstring/agora
Automated Generation and Omission Recurrent Architecture (AGORA). This model inputs speech (audio recording) and replaces hate speech and profanity with generated textual content. (Speech to text model.) McGill's submission to Project X, 2022-23.
https://github.com/thefloatingstring/agora
generative-model hate-speech-detection speech-to-text
Last synced: 11 months ago
JSON representation
Automated Generation and Omission Recurrent Architecture (AGORA). This model inputs speech (audio recording) and replaces hate speech and profanity with generated textual content. (Speech to text model.) McGill's submission to Project X, 2022-23.
- Host: GitHub
- URL: https://github.com/thefloatingstring/agora
- Owner: TheFloatingString
- Created: 2022-12-14T05:48:08.000Z (about 3 years ago)
- Default Branch: main
- Last Pushed: 2024-04-02T13:45:39.000Z (almost 2 years ago)
- Last Synced: 2025-01-13T08:12:44.414Z (about 1 year ago)
- Topics: generative-model, hate-speech-detection, speech-to-text
- Language: Python
- Homepage:
- Size: 560 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 18
-
Metadata Files:
- Readme: README.md
- Contributing: CONTRIBUTING.md
Awesome Lists containing this project
README
# AGORA - Automated Generation and Omission Recurrent Architecture
Given a speech input (audio recording), this model replaces harmful speech with generated textual content. (Speech to text model.)
### Installation and Setup
Configure environment variables:
```
set OPENAI_API_KEY_AGORA=
```
Setup
```
git clone https://github.com/TheFloatingString/agora.git
cd agora
pip install -r requirements.txt
```
In a Python file:
```python
from src.agora import Agora
agora_model = Agora()
response = agora_model.transcribe_audio("filepath_to_speech_audio.wav")
print(response["outputText"])
```
### Quickstart examples
```
python -m quickstart.run_sample
```
### Analyze AGORA's ability to Recognize Offensive Content in the Jigsaw Dataset
Note: move `train.csv` into `data/jigsaw-data` from the Jigsaw dataset on Kaggle (https://www.kaggle.com/competitions/jigsaw-toxic-comment-classification-challenge/data)
```
python -m src.run_jigsaw_data
python -m src.analyze_results
```
### Filter and Paraphrase the Speech-to-Text Functionality for Offensive Content
Run the folowing, while making sure to change the filename from `1` to `10` at each new run.
**Warning: the audio files contain explicit content.**
```
python -m src.run_audio_files
```