中文文本拼寫糾錯研究綜述

打開文本圖片集

中圖分類號：TP391.1；TP301.1 文獻標識碼：A 文章編號：2096-4706（2025）08-0138-08

Abstract：Chinese Spelling Correction（CSC）isacrucial foundational task inNaturalLanguage Processing （NLP），and providessupport forthedownstreamtasks andresearch.Theresearch in the fieldofCSCtaskscontinues to develop，mainly divided into eror corrction methods based onN-Gram language models，Deep Leaming，andLarge Language Models （LLMs）. Firstly，techaracteristicsoftheN-GamlnguagemodelanditsapplicationinCSCareanalyzed，rvealingitsadvatagesin capturing contextual information.Secondly，methodsbasedonDepLearning improve theaccuracyof error coectionthrough deep neural networksand are widelyused in Chinese text procesing.Atthesame time，theriseofLLMs provides new ideas for speling correction，demonstrating their enormous potentialindealing withcomplex languagephenomena.Thisreviewprovides adetailedoverviewofthecurrentresearchstatusintheCSCfeld，providingareferenceforscholars engaged inrelatedresearch.

Keywords： Chinese text; spelling correction; N-Gram language model; Deep Learning; Large Language Model

0 引言

中文文本拼寫錯誤（CSC）是自然語言處理（NLP）領(lǐng)域的一個重要的基礎(chǔ)研究方向，其目的是檢測和糾正文本中出現(xiàn)的拼寫錯誤，為后續(xù)的文本分析、信息檢索、文本生成等任務提供了干凈、準確的輸入數(shù)據(jù)。（剩余13506字）