Grammatical Error Correction(GEC) refers to detecting poor grammar and correcting it accordingly. These errors include spelling, incorrect use of articles, prepositions, pronouns, nouns, and poor sentence correction. In recent years, NLP has made significant advancements due to the integration of various ML and DL models. This helps propel the field of Grammatical Error Correction to new heights and opens doors to various innovative applications in language processing. This work shows the application of the T5 model to correcting grammatical errors and compare it to pre-existing models and architectures.
The dataset utilized for this task is the NAIST Lang-8 Corpus of Learner English, refined for the 14th BEA Shared Task. It contains texts written by English learners, which may include non-native speakers of English. As soon as the learners write a passage in a language they are learning, the native speakers correct it. It is a data set used for research and evaluation in NLP and the acquisition of a second language or SLA. The corpus is designed to support research on language learning and error detection in learners of English. The corpus is often used as part of the BEA Shared Task, where participants develop and evaluate NLP systems for various language-related tasks, such as error correction and grammatical error detection. The corpus consists of texts written by language learners who used the Lang-8 platform to practice their writing skills. The NARA Institute of Science and Technology (NAIST) makes the corpus publicly available for the 14th BEA Shared Task The dataset is accessible through the NAIST website.