A Corpus-based Study Of The Content Validity Assessment Of Proofreading And Error Correction In TEM-8

This paper was carried out to investigate the content validity of proofreading and error correction in TEM8 with the assistance of CLEC and BNC. Errors designed in TEM8 were classified according to error types in CLEC. Then error types in TEM8 were compared with error types in the sub-corpus of CLEC: ST6 in order to test the content validity of the proofreading and error correction part in TEM8 on the statistics analysis basis. Content validity touches upon two aspects: content relevance and content coverage. This paper assesses the content validity of proofreading and error correction from these two aspects.This study first explored characteristic of input by analyzing the total tokens, average sentence length and readability of proofreading materials to check whether proofreading materials were consistent or not. All proofreading and error correction materials of TEM8 were put into Readability Analyzer 1.0. The characteristic of input of proofreading concerning total tokens are consistent based on the test specification of TEM8 varying from 178 to 290, with tokens of the year 1998,2008 and 2009 a little out of the rang. The layout of proofreading in TEM-8 concerning average sentence length in these 13 years is under a comparatively balanced control except for the ASL of the year 2001, 2004 and 2009, which can guarantee a certain degree of difficulty level in a certain sense. The readability of the materials is within the scale from 60 to 30 which corresponds to"fairly difficult"to"difficult", which can prove that the difficult level of the proofreading in these 13 years is consistent in general, except the year 2004 with the readability below 30 corresponding to very difficult. From the perspective of content relevance, the content validity of proofreading and error correction in TEM8 is relative high.From statistic results, it can be found that designed error types in TEM-8 covered about 67.13% of the total error types in ST6, which can indicate that error types in TEM-8 are relatively even distributed concerning the content coverage. The percentages of 28 error type in TEM8 and in ST6 are put into SPSS13.0 to see the correlation relationship with the purpose of studying whether error types designed in TEM8 can represent errors English seniors often made in their writing. Statistic results showed that designed error types in TEM8 are significantly correlated with these in ST6 on the level of 0.374.The correlation is significant on the level of 0.05, which indicates that errors designed in TEM8 are representative to a certain degree. From statistic results of the study of answers to collocation error type, it can be found that: some collocations are representative with the Z-score above 2.0, while some collocations are not representative with the Z-score below 2.0.This thesis ends with some implications of the present study and suggestions for future study.