Write-up

https://www.overleaf.com/project/61d0a7713f346c0e13093699

Aim

Build a formal framework that estimates the authorship probability for a given pair (user, photo).
To illustrate the proposal, we use data gathered from TripAdvisor containing the reviews (with photos) of restaurants in six cities of different sizes.

Setup

Enviroment

• python3.6
• Install requirements.txt

Run pre-trained models

In order to run pre-trained models you have to:

1. Download the pretrained models from: https://www.aic.uniovi.es/downloadables/ELVis/models.zip
2. Create a folder called "models" (if not exists).
3. Unzip the compressed files under models/ path.
4. Run the Main.py file with stage='test' or stage='stats' and city='<city>' (view parameters section).

Train the model with pre-generated data

In this case, you need to follow these steps:

1. Download a city data from: https://dx.doi.org/10.34740/kaggle/dsv/944945
2. Create a new folder called "data" (if not exists) and iside, another with the name of the city.
3. Unzip the files under data/<city>/ path.
4. Run the Main.py file with stage='grid' or stage='train' and city='<city>' (view parameters section).

Parameters

You can configure:

• stage

  • "stats": If you want to obtain stats about the dataset.
  • "grid": To train a model testing different configuration values.
  • "train": If you want to train the final model.
  • "test": To evaluate the model.

• city: City to work with

• lrates: List of leaning rate values (if you want to try different values)

• dpouts: List of dropout values (if you want to try different values)

• epochs: Epoch number

• seed: Random state

Citation

Please cite the following paper:

Jorge Díez, Pablo Pérez-Núñez, Oscar Luaces, Beatriz Remeseiro and Antonio Bahamonde: Towards Explainable Personalized Recommendations by Learning from Users’ Photos. Information Sciences, in press. 2020. https://doi.org/10.1016/j.ins.2020.02.018