Automatic recognition and classification of future work sentences from academic articles in a specific domain

Future work sentences (FWS) are the particular in academic papers that contain author's description of their proposed follow-up research direction. This paper presents methods to automatically extract FWS from and classify them according different future directions embodied paper's content. recognition will enable subsequent researchers locate more accurately quickly reduce time cost acquiring corpus. At same time, changes content be illuminated, a foundation laid for in-depth semantic analysis sentences. The current on automatic identification is relatively small, existing cannot identify papers, thus conduct data mining large scale. Furthermore, there many aspects work, subdivision conducive specific development directions. In this paper, Nature Language Processing (NLP) used as case study, extracted classified into types. We manually build an annotated corpus with six types FWS. Then, classification implemented using machine learning models, performance these models compared based evaluation metrics. results show Bernoulli Bayesian model has best task, Macro F1 reaching 90.73%, SCIBERT weighted average 72.63%. Finally, we keywords gain deep understanding key described FWS, also demonstrate determination reflected by measuring similarity between abstracts.