This paper researches a relatively common problem among learners of second languages mistakes of subject and verb agreement. Human beings have attempted to handle this problem in previous times, generally through the utilization of manually made rules or statistical models. These methods may have okay effectiveness for relatively simple sentences, but once matters become a little more complex or the environment changes, they therefore cannot always maintain very good effects. In order to circumvent that problem, we have constructed an alternative type of framework which introduces several concepts simultaneously instead of adhering to only a single one. It is composed of three component parts. There exists a component which is based on constraint, it marks the positions where the agreement could possibly have problems. there exists a kind of agent which travels inside the sentence structure, therefore it endeavors to comprehend how every part combines together in the syntactic aspect. In addition to this, a probability based filtering step examines possible revisions and evaluates which of these appear comparatively more probable. This system does not completely rely on rules, nor does it purely rely on data, therefore it mixes these two aspects with a certain method. In the actual operation, this therefore makes it a little more stable in different situations. When we carried out some experiments, the obtained outcomes were obviously superior to that which earlier methods have reached correct rate increased by over 15% on standard benchmark data collections. Certainly, this is not a perfect solution, but it does give indication that the bringing together of structural restrictions and probabilistic inference can provide help. At the very least up to current moment, this can be regarded as a feasible method that can promote the promotion of grammatical error correction work, particularly for language learners. It also can possibly provide certain guidance for the future research about automatic language studying tools.
