文本作者身份识别研究综述

目的/意义]鉴于传统的作者身份识别方法不适用于当前大量涌现的网络文本。综述近年文本作者身份识别的典型方法和关键问题,并进行客观分析和评价,以期为进一步研究提供新的思路。方法/过程]分别从应用领域、文体特征选取、作者身份建模和性能评价指标等方面对国内外作者身份识别相关研究现状进行客观分析,梳理相关领域研究发展脉络和趋势。结果/结论]作者身份识别需要适应短文本、不规范文本、海量、高维和多语种环境,需更具表现和刻画能力的多层面特征和相应的作者身份建模方法,并借助信息检索、机器学习和自然语言处理领域的最新研究成果提高效率和准确率。

Review on Text Authorship Identification

Purpose/significance]The traditional authorship identification methods are not applicable to web text.In this paper some typical methods and the key problems in recent years are reviewed in order to provide new ideas for further research.Method/process]We objectively analyzed the authorship stylistic features selection,the authorship modeling and the performance evaluation indexes respectively,presenting the latest development of the related areas and trends.Result/conclusion]Authorship identification should adapt to short,non-standard,mass,high-dimensional,sparse and multilingual text.More efficient multidimensional features models and corresponding authorship identification methods are required.The latest achievements in information retrieval, machine learning and natural language processing are the promising solutions to improve the efficiency and accuracy of authorship identification.