Research On Error Detection Of Chinese Students’ English Compositions

Language is the major tool of human communication, and English is theinternational communicative language. With theEntering into the21st century, China has successfully joined the WTO, andBeijing2008Olympic Games and Shanghai World Expo2010have been successfulbid. English plays an increasingly important role in global political, economic,cultural, science and technology exchage. Chinese students began to learn Englishfrom primary school, and have been learning English for decades. Chinese studentsmust have English exams when looking for a job or studying aboard. Chinese studentsare increasingly focusing on the English learning.For a long time, English writing is an important means to measure the Englishlevel of English Learners. However, English writing of Chinese students is bad. Thereare many differences between Chinese and Western culture. Chinese students oftenmake many kinds of errors in their English compositions. Correcting Englishcompositions is time-consuming and laborious. Teachers often have heavy teachingtasks. Teachers’ workload will be more arduous if they correct each student’s Englishcomposition. Take advantage of computers to detect errors in English compositions ofChinese students and provide correction advice, we can reduce teachers’ burden, anddirectly contributed to the students more involved in the practice of writing toimprove students’ English writing.This paper focuses on theories and key technologies of natural languageprocessing, spelling correction and grammar checking. The research work is asfollows:（1）English composition preprocessing based on NLP tools. This paperanalyses each step of natural language processing in detail. We use Stanford CoreNLPand MorphAdorner tools package to preprocess the students’ English compositions.The pretreatments, which include tokenization, lemmatization, Part-of-Speech taggingand named entity recognition, are foundation of spelling correction and grammarchecking.（2）Spelling checking and correction. This paper lists common spelling errortypes in English composition of Chinese students. Based on some theories andalgorithms of spelling checking, this paper uses the dictionary method to check non-word errors in students’ English compositions, and use the minimum edit distancealgorithm to correct non-word errors. N-gram model theory is applied on thereal-word error checking.（3）Grammar checking. This paper proposes the POS bi-gram model insteadof the word bi-gram model in the statistical method. This paper put a priority onstudying matching rules. Firstly, the types and causes of common grammar errors inChinese students’ English compositions are analyzed in detail. Secondly, somespecific grammar errors types can be detected by some determination algorithms.Lastly, the paper analyzes those grammar errors types which can’t be detected bysome determination algorithms, and detects them by mating rules.