Generation of Research Paper Titles

Can NLP accurately and effectively generate research paper titles? In this research paper, an effective and accurate artificial intelligence NLP model is tried to be determined by evaluating various models and methods for title generation.

Can NLP accurately and effectively generate research paper titles? In this research paper, an effective and accurate artificial intelligence NLP model is tried to be determined by evaluating various models and methods for title generation. Titles are a vital part of any work. People tend to struggle to find good titles for their work, including researchers. A title must be representative of the whole paper and its topic, so it must be carefully chosen. Titles are the main reason people read a paper, so it’s very important to attract the audience of the researcher. Apart from this, normalizing research paper titles with generated ones can eliminate bias and exaggeration from them. To determine an accurate and effective model, various methods for title generation were tested given the abstract of a research paper. The methods tested vary from very simple, like “bag of words”, to complex with GPT-2. The best result was getting a sentence similarity of 0.9032, which was 66% higher than the initial model. Through this project it is confirmed that generating good titles with NLP is a hard task, but the results from the models can facilitate the process of creating a good and compact research paper title.

Explore More!

Research Paper

Source Code