Context-aware Citation Recognition and Recommendation

Abstract

With the increase in the publication volume of the paper it is difficult for today's researchers to review all relevant literature in a limited time and find the most appropriate reference when writing a paper Therefore some researchers have begun to try to turn the process of finding citations into an automated procedure and focus on the domain area of citation recommendation The purpose of citation recommendation is to automatically recommend the appropriate citations to a sentence that needs to be cited which will be of great help to the researcher's thesis writing or reviewing the literature However previous studies assume that the input itself is the sentence that needs to be cited In addition previous studies usually calculate the similarity score by simply using the full-text of paper which is not accurate The content of the sentence usually only represents a specific part of the paper not just a summary of the paper This study proposed a citation recommendation framework (CiRec) which contains a citation contexts detection process to find the sentences that need to be cited in the input manuscript; and for the process of comparison this study is divided into abstracts full-text and the In-link contexts of the papers compared to a sentence which is to find the most similar part between the sentence and the cited document The experimental results showed that the comparison method proposed in this study reflects the behavior of citation search of the researchers which is better than the current method

Date of Award	2019
Original language	English
Supervisor	Hei-Chia Wang (Supervisor)