Neural Error Correction for MLA In-Text Citations

Although using in-text citations is an integral part of the academic writing process, it is oftentimes a head-scratcher for many students. Currently, popular AI writing assistants do not make recommendations on specific formats like MLA. This AI-powered prototype aims to help fill the gap by detecting and correcting common mistakes with MLA in-text citations - such as the ones illustrated below.

Under the Hood

Training data was created using pattern-based error generation (3 million observations). Machine translation from erroneous to correct observations was built using Sequence-to-Sequence (Seq2Seq) modeling with Keras. A pre-trained BERT NER model is leveraged during text pre-processing steps to identify author names.

Demo

Click here to try out the web app demo.