基于被引片段识别的科技文摘综述

目的/意义]基于被引片段识别的科技文摘生成是文献计量学、信息检索和自然语言处理等领域共同关注的研究问题。通过梳理相关成果，可为后续研究提供借鉴。方法/过程]本文首先介绍被引片段概念，进而从被引片段识别与分类、文摘生成与评价等步骤对相关研究进行综述。结果/结论]当前被引片段识别总体上可以分为机器学习和检索两类，分面判定还存在标准不一致的问题，摘要生成与评估方法相关研究较欠缺。

Review on Scientific Summary Based on the Recognition of Cited Spans

Purpose/significance] Scientific summarization based on reference spans recognition is the common research question in study felds such as bibliometrics, information retrieval and natural language processing. By reviewing the related research achievements, we hope to provide some references for the further research.Method/process]This paper introduced reference spans concept and presented and analyzed the latest development of areas such as reference spans recognition, summary facets classifcation and summary generation and evaluation.Result/conclusion] At present, the methods of reference spans recognition could be put into two types, including methods based on machine learning and methods on retrieval. Tere were no uniform standards of discourse facets of abstract. Te researches on the produce of summary and evaluation methodology were still relatively lacking.