First Semantic Search for Persian based on Transformers
This project was made based on the Transformer models. The semantic search operation tested on three different scopes:
- Jobinja, is a well-known job search system in Iran in favor of job seekers and recruiters. The dataset consists of 9,952 job offers.
- Taaghche, consists of books' meta-information available on Taaghche as of 2019 (around 4,505 books).
- Universal, includes a huge range of topics from many sources, DigiMag, Chetor, Wikipedia, Ninisite, 1Pezeshk, and some others. For this particular example, we used only 44,000 records (out of 807,185 documents).
- You can see the results in the video. Also, it is important to mention that the whole dataset and code would publish soon
- For creating a real scenario some misspelling and grammatical errors happened in the demo.
