NER-SENT-arabic is a project to a A virtual tourism company interested in knowing the places that people mention in their conversations and their opinions towards them.
- the project in two parts:
- first part:
- The NER system is fine-tuned with 3 languages and 2 sources of data. It takes a body from an LLM and adds a special head, then performs some tricks and fine-tunes it on all datasets. After that, it conducts error analysis to understand the strengths and weaknesses of our model.
- second part:
- Sentiment analysis system: In this part, I used an LLM model that is ready for inference. This part is a proof of concept with MVP.
- first part: