The automatic and linguistic processing of web data such as comments, blogs, SMS, sentiments, recommendations, opinions, etc., are complicated because of their many forms of writings, styles, spellings, linguistic structures, etc., on one side, and the lack of pre-processed and standardized data, on the other. Implementing solutions that can help address these issues is a real need and challenge for the textual wrinting normalisation process. It’s in this context that we propose an unsupervised approach based on deep learning that implements an aided system for the writing normalisation in a given and specific context of use.