Looking for experience!
Often we come across documents, or text, in general which contain words and phrases that might be difficult for a normal guy to understand. For example - legal documents (docs published by Public Administration, contracts), and medical reports. We have developed an end to end web application, that takes such texts and tries to provide a lexically and syntactically simpler version of the provided text. These changes can be of the form - replacing complex words and phrases, breaking sentences, etc. We finetuned the small and medium models upon publically available datasets - Asset Data, MedWiki, and SimPA Corpus. We made use of the 'generate' API to get the simplified text. Also we used the 'embed' API to filter out noisy examples from the data, and also to choose the best possible response from the mulitple generations.