Skip to content

Latest commit

 

History

History

CQAS

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
data-prep
data-prep
 
 
Dockerfile
Dockerfile
 
 
README.md
README.md
 
 
demo.py
demo.py
 
 
docker-compose.yaml
docker-compose.yaml
 
 
main.py
main.py
 
 
raytune.py
raytune.py
 
 
requirements.txt
requirements.txt
 
 
summarizer.py
summarizer.py
 
 
test.py
test.py
 
 
test_pegasus.py
test_pegasus.py
 
 
test_t5.py
test_t5.py
 
 
train.py
train.py
 
 
train_from_checkpoint_multi_news.py
train_from_checkpoint_multi_news.py
 
 
train_multi_news.py
train_multi_news.py
 
 
train_pegasus.py
train_pegasus.py
 
 
train_t5.py
train_t5.py
 
 

README.md

Comparative-Question-Answering-Summarization

UHH -

The Comparative Question Answering Summarization is an abstractive text summarization model used for summarizing CQAM (Comparative Question Answering Machine) outputs.

Dataset

Models are trained on a small dataset obtained by feeding CQAM outputs to ChatGPT to obtain ground truth summaries. The dataset is then pre-processed by modifying the model input to the form appropriate to HuggingFace models. The dataset consists of 1602 inputs, with 80% dedicated to training, 10% to validation, and 10% to testing.

Evaluation Metrics

For evaluation, following metrics were used:

  • Rouge1
  • Rouge2
  • RougeL
  • RougeLSum

Training

After installing requirements in requirements.txt, you can obtain the best-performing model by running following command:

python train.py

Hyperparameter tuning can be done via:

python raytune.py

API

The service can be started with:

uvicorn service:app --reload

Summarization is available via endpoint /summary

Demo

You can try generating summaries via demo:

python demo.py