jayantaadhikary / dataset-error-reduction

Reducing Error in NLP Datasets using LLMs

Dataset Error Reduction

To reduce error in NLP Datasets using LLMs.

The datasets used currently are SQuAD and RACE which are Extractive Question Answering Datasets.

Use in your system

Fork and Clone this Repository.

1. If you are using a OpenAI API key, create a .env file in the same directory and add your OpenAI API key.

OPENAI_API_KEY = YOUR_API_KEY

2. If you are using a LLM model locally using a local server, then use the files inside of the LocalLLM folder instead of the default files. The folder /ollama consists of my notebooks which were using Ollama to run LLMs locally.

PS - I use LM Studio and Ollamato run the LLMs locally.

References:

1. SQuAD: 100,000+ Questions for Machine Comprehension of Text (Rajpurkar et al., EMNLP 2016)
2. RACE: Large-scale ReAding Comprehension Dataset From Examinations (Lai et al., EMNLP 2017)