自然语言处理研究综述  

A Review of Studies on Natural Language Processing

作　　者：赵铁军[1] 许木璠 陈安东 ZHAO Tie-jun;XU Mu-fan;CHEN An-dong(Faculty of Computing,Harbin Institute of Technology,Harbin Heilongjiang 150001)

机构地区：[1]哈尔滨工业大学计算学部,黑龙江哈尔滨150001

出　　处：《新疆师范大学学报（哲学社会科学版）》2025年第2期89-111,F0002,共24页Journal of Xinjiang Normal University(Philosophy and Social Sciences)

基　　金：国家自然科学基金联合基金项目“人机协同翻译方法及关键技术研究”(U1908216)的阶段性成果。

摘　　要：近年来,自然语言处理因在分析与建模人类语言任务领域取得诸多成果而备受关注。当前,大规模预训练语言模型展现出强大的对话问答和文本生成能力,带来自然语言处理研究的新一轮热潮。自然语言处理在机器翻译、文本摘要、信息抽取等领域应用广泛。文本首先讨论自然语言处理针对语言学四个不同层次文本信息的分析手段,对自然语言处理的基本任务组成进行概述;其次,讨论自然语言处理在具体下游任务中的应用现状,包括自然语言处理在具体任务中的应用历史、当前的研究趋势以及面临的挑战;最后,在大规模预训练语言模型研究对数据集提出更高要求的背景下,对自然语言处理领域已有的数据集及评测基准集等进行讨论。In recent years,natural language processing(NLP)has attracted much attention due to its numerous achievements in analyzing and modeling human language tasks.Currently,large-scale pre-trained language models have demonstrated powerful question&answer and text generation capabilities,bringing about a new wave of enthusiasm in natural language processing research.Natural language processing is widely used in various fields such as machine translation,text summarizing texts,and retrieving information.This paperfirst discusses the analysis methods of natural language processing for linguistic texts at four different levels,and provides an overview of the basic tasks of natural language processing.It goes on to discuss the current applications of natural language processing in specific downstream tasks,including its application history,current research trends,and challenges it is faced with.Finally,in the context of higher requirements for datasets in large-scale language model research,it discusses the existing datasets and evaluation benchmark sets in thefield of natural language processing.

关 键 词：自然语言处理 句法分析 语义分析 机器翻译 问答系统 信息抽取 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]