TheSpoon is a tool for searching trough restaurant reviews, built by food enjoyers for food enjoyers.
TheSpoon utilizes Sentiment Analysis and Information Retrieval techniques to allow users to search trough a selection of reviews of restaurants throughout the USA, accessible trough a Next.js front-end.
Requirements are listed in the /REQUIREMENTS/requirements.txt file and can be installed by running make install_requirements
To run TheSpoon Internet access is required, both to download the Sentiment Analysis module and to access the Vercel JavaScript front-end.
In the DATASET folder there is already a (relatively speaking) small selection of Yelp restaurant reviews retrieved from Kaggle.
For the sake of ease of sharing (the original dataset was around 6 GB) and because Kaggle needs a user key to retrieve data from their API, the data has been already downloaded and trimmed.
By running make index or python ./INDEXER/indexer.py those reviews will be indexed and analyzed by using Sam Lowe's model.
The index can be deleted, alongside the model's cache, by running make delete_index.
To start benchmarking run make benchmark or python ./runBenchmark.py. The results will be saved to ./GENERATED_BENCHMARK/
The benchmarking data can be deleted by running make delete benchmark.
The benchmarking suite has both automatic and manual tools. The user will be asked to perform the manual evaluation of the queries by which the NDCG will be calculated.
To start up TheSpoon it is sufficient to run make start, then head to our Vercel deployment.
University of Modena and Reggio Emilia undergraduates in Computer Science @riccardomassi and @filipczuba.
This project is licensed under the MIT License.