Micro Citation Importance Identification and Its Application to Literature Evaluation

We present our approach for identifying the importance of citing sentences, where the importance of citing sentences is termed “micro citation importance” in our research. This approach characterizes a regression method based on the pretrained language model SciBERT, where the citation function is incorporated into the citing sentence as its input. Remarkably, our approach demonstrates superior performance on the 3C Citation Context Classification Shared Task corpus, suggesting that both regarding micro citation importance identification as a regression problem and integrating the citation function contribute to enhanced performance. Furthermore, we extend our investigation to literature evaluation, introducing a novel metric coined “micro citation frequency” derived from micro citation importance acquired by our proposed regression method. Notably, it is observed that micro citation frequency outperforms the established metric of citation frequency in terms of evaluating high-quality papers, further validating our proposed regression method. Our work not only enriches citation content analysis, but also holds implications for optimizing literature evaluation.