CompoundCue Textual Question Answering Application For Compound Sentences

Abstract

"In my research, I aimed to improve the accuracy of question-answering models specifically on compound sentences. Existing question-answering models tend to perform well on simple questions that require short answers, but struggle with more complex questions that involve compound sentences. Compound sentences can be difficult to parse and understand, leading to inaccurate answers.

To solve this problem, I used a pre-trained question-answering model from the Hugging Face library, specifically the Bert-base-squad2 model. I fine-tuned this model on a dataset of squad v2 dataset that involved compound sentences. Fine-tuning involved updating the model's parameters on the new dataset to make it more accurate in answering questions that involved compound sentences."