parth126
commented
2 months ago - Outcomes: A new dataset for NER on solar patents + understanding and implementing a NER model
- Data can be generated using a mix of human labels and LLMs
- Recommended doing all experiments on an additional known dataset. This will help in a fair evaluation.
- Atleast one NER needs to be implemented from scratch. Only training the models within exisitng tools like spacy is not sufficient.
Title
NER on patent data
Team Name
Octane
Email
202318005@daiict.ac.in
Team Member 1 Name
Aditi singh
Team Member 1 Id
202318005
Team Member 2 Name
Hani Soni
Team Member 2 Id
202318044
Team Member 3 Name
Kruti patel
Team Member 3 Id
202318032
Team Member 4 Name
Garvika singh rajawat
Team Member 4 Id
202318061
Category
Optimizing an existing system
Problem Statement
We are going to do Named Entity Recognition on the solar patent data which we are going to scrape from Google patents based on different solar modules, technology, material and innovation. We are then going to train a patentBert based on the initial few annotations we make on the collected data for NER task.
Evaluation Strategy
We are yet to decide on the evaluation strategy.
Dataset
Not yet
Resources
https://arxiv.org/abs/1901.08746