HSK Composition Corpus-oriented Automatic Recognition Of Wrong Conjunction Usages And Application

The function words undertake double tasks of grammar performance andsemantic analysis, it plays a crucial role in the grammar analysis and theunderstanding of Chinese. The wrong function words can make sentences’ meaningexactly opposite or mistaken. So it has a great significance on researching theconjunction wrong usage.This paper researched automatic recognition of conjunction wrong usages whichstarted from the modern Chinese conjunction knowledge. The paper extracted theconjunction error corpus from the HSK dynamic composition corpus to construct thewrong conjunction usage corpus. It attempted to realize the automatic recognition ofwrong conjunction usages based on rules and statistics. In the rule-based approach, wereferenced the rules in modern generalized conjunction Knowledge Base to write theformal rules for errors, formulated wrong usage rules for conjunctions, and expandedthe Modern Chinese Conjunction Rule Base. Automatic recognition for wrong usageswere based on the precise conjunction usage rules. Since it was complex when writingthe rules and also influenced by the knowledge and experience of the rule-makers, weused the model of CRF for conjunction errors’ automatic recognition, which was theprevalent statistical method. The statistical methods were available to learn thecontext knowledge automatically or semi-automatically, but couldn’t perform well forrelatively sparse distribution of conjunction wrong usages. We combined theadvantages and disadvantages of the two methods, which took the rule and CRFmethod together. The experiment results showed that the rule method couldautomatically recognize conjunction errors. Further horizontal and verticalcomparison of the experiment results showed that the accuracy of the rule method washigh, but there were relatively a high rate of recall and a high F-value whencombining the rule and CRF together.On the basis of automatic recognition for usage of conjunction errors, we constructedan auxiliary teaching system, which could be used for teaching Chinese as a foreignlanguage teachers and students. The system relied on the rules and corpus ofconjunctions and could assist students and teachers to learn and teach. In the assistedlearning module, when you wanted to search certain conjunctions, the system would show the correct and wrong examples of different usages. In the conjunction wrongusages identification module, the user could give candidate sentences which may bebiased, after the analysis of the system, it would give such references about the biasedtype and the proposed right sentences.