A Correction Model Of Deaf Students’ Sequence Errors In Their Written Chinese

The sequences of written language of deaf people present different characteristics from those of hearing people.This study investigates whether the characteristics are universal and what the development trend is.Based on the observation of the qualitative and quantitative characteristics of a large amount of deaf written language,this dissertation employs the theory of cognitive psychology to explore the cognitive psychological causes of a series of errors of deaf written language sequence.Artificial neural network technology is used to simulate the corresponding cognitive psychological mechanisms leading to sequence errors,and further construct an automatic error correction model of the deaf students’ written Chinese.The results show that the model proposed in this study has the state of art performance in the same baseline models.It is proved that the cognitive psychological mechanisms which lead to the sequence errors of deaf students’ written Chinese can effectively guide the restoration of error texts,that is,the theoretical hypothesis is valid and plausible.Specifically,from the standpoint of usage-based research,this study collected the diaries,chat records and some compositions reflecting the daily use of language from the fourth grade of primary school to the third grade of senior high school from many special education schools or deaf-mute schools in several provinces.The total number of words is about 14.3 million,and the corpus has been tagged systematically by people.Due to the huge capacity of the corpus,it is enough to describe the written Chinese features of deaf students in North,Central and South China in various aspects of daily life.The whole corpus is tagged with artificial error correction markers for neural network model training.Based on the perspective of cognitive psychology,it is found that there are a lot of lexical deconstruction errors in the sequences of written Chinese of deaf people.Some structures and words are solidified and cannot be disassembled.This kind of widespread phenomenon shows that deaf students adopt chunking and meaning memorizing strategies when they memorize written language,just as they do when they are listening to L1 acquisition.But once something goes wrong,it is hard to correct it.This shows that the mechanisms of deaf and hearing people to memorize information are the same,being processed by working memory for both of them.In order to memorize speech materials,deaf people also use repetition and association.The difference is that,due to the lack of effective guidance and correction,the errors of the deaf will exist in the process of language acquisition for a long time after the occurrence of wrong chunks or related results.In the written Chinese of the deaf,there are many features,such as ellipsis,several predicates used together,and the thematization of the object.Sign language communication of deaf people is more time-consuming than that of hearing people.In order to express their intention quickly,the deaf tend to simplify their gestures.Through a long-term and extensive simplification process,the above rules have been formed.After the acquisition of written Chinese,due to the lack of syntactic knowledge in time,the rules of sign language are directly applied to the written language,which leads to the above-mentioned error characteristics.This process is consistent with the regulation of basal ganglia on speech skills.Speech units are combined in sequences and encapsulated after repetition for many times to achieve implicitness.Ultimately,language activities are highly automated and unconscious,and brain costs are reduced.Because there is no syntactic structure constraint,similar with hearing children in the early stage of L1 acquisition,the sequences of deaf students’ texts tend to put action words in the sentence final position,and the agent and the patient before action words,forming the declarative sentence order of SOV or OSV.However,the two nouns before the action words are not necessarily related to the action,that is,they are not necessarily the agent or the patient of the action.This study finds that such a sequential order stems from the strength of the subjective or objective stimuli that affects involuntary attention changes.This mechanism leads to the preferential expression of external factors that are more in line with the speaker’s subjective wishes,emotions,or with stronger stimuli.A large number of seemingly "unreasonable" metonyms have also been found in the written Chinese of the deaf.It is observed that the capacity of natural sign language and conventional sign language vocabulary is relatively small.In the process of communicative dialogue,attention will initiate a relevance search,pointing to and extracting a unit with contiguous semantics and pragmatics for expression.Metonymy of the deaf is a daily expression method based on insufficient vocabulary,which is different from the purpose of hearing people as a rhetoric.Therefore,it is more frequent and widespread.In order to verify the validity of the above inference,this study uses the long-term and short-term memory model to simulate the information repetition and association strategy processing mechanism of the working memory platform and uses the selfattention mechanism to realize the endogenous attention function to capture the syntactic features of sequences,including the word order.The above algorithm is integrated into the encoder-decoder architecture to learn the wrong & correct sentence pairs case by case.Transformation features are extracted to achieve the purpose of automatic error correction.Compared with other baseline models,the experimental results of our model indicates that the theoretical hypothesis of this study is plausible.