Airbnb Reviews

For this project we have different components as follows:

ETL:

used to extract the data form the website and load it locally and into Kaggle Private Dataset (feel free to contact the owners to add you to the dataset)

To run the etl use the following snippet etl.main(is_dataset_new, upload) use is_dataset_new if you want to create a dataset and upload if you want to uplad a version

in the dataset folder a file called dataset-metadata.json should be added as follows:

{
  "title": "Airbnb cities reviews", 
  "id": "moghazy/airbnb-cities-reviews", 
  "licenses": [{"name": "CC0-1.0"}]
}

More examples can be found in the notebook model_trial.ipynb

Models & Analysis:

1. GRU: gru_model.py results and analysis in model_trial.ipynb
2. REMBert: airbnb_sota_sentiment_RemBert.ipynb
3. DistilBert: DistBERTandSpatial.ipynb

DistilBert Model Training

Train the Model

The model is trained using the DistilBERT architecture with the following configuration:

• Epochs: 1
• Batch Size: 128
• Learning Rate: Optimized with weight decay

Save the Model

After training, the model and tokenizer are saved to the directory ./fine-tuned-distilbert.

Topic Modeling:

Models and README.md for this specific part can be found in topics directory

Libraries Used:

• pandas==2.2.2
• datasets==2.20.0
• transformers==4.42.3
• scikit-learn==1.2.2
• matplotlib==3.7.5
• seaborn==0.12.2
• geopandas==0.14.4
• folium==0.17.0
• nltk==3.8.1
• gensim==4.3.3
• vaderSentiment==3.3.2
• wget==3.2
• wordcloud==1.9.3
• kaggle==1.6.17
• kagglehub==0.2.8
• keras==3.4.1
• textblob==0.17.1
• tqdm==4.66.4
• pip install pandas==2.1.4
• scikit-learn==1.2.2
• torch==2.3.1+cu12
• transformers==4.42.4
• datasets==2.20.0
• beautifulsoup4
• nltk==3.8.1
• langdetect==1.0.9
• matplotlib==3.8.0
• seaborn==0.12.2
• tqdm==4.66.4
• pip install pandas==2.1.4
• numpy==1.26.4
• matplotlib==3.8.0
• seaborn==0.12.2
• scikit-learn==1.2.2
• wordcloud==1.9.3`