Because of the growing use of social media across the globe, the amount of unsolicited huge quantity of e-mails has improved, necessitating the deployment of a truthful machine to filter out such troubles. Junk mail emails are the maximum common trouble at the net. It is straightforward for spammers to ship an e-mail containing spam messages. Spammers have the capability to steal critical records from our gadgets, including files and contacts. Numerous deep studying-based word embedding processes were evolved in recent years. Those tendencies inside the place of phrase illustration can be able to provide a solid strategy to such issues. On this studies, we will observe a method that employs herbal language processing (nlp) to come across junk mail and ham information the use of the junk mail electronic mail dataset. We've got used dense classifier sequential neural communit , lstm and bilstm and in comparison accuracies and consequences. The dataset's efficacy is decided the use of metrics including bear in mind, accuracy, and f1-score.The take a look at indicates that by using bi-lstm category, the dataset's typical accuracy improves. The general paintings is done in python and implemented in a jupyter pocket book