Title Generation of Talk Shows

Abstract

An extraordinary video title describes the most notable occasion compactly and catches the watcher’s consideration. The Talk shows have manually generated titles and may be the title of video is not matched with the content of video. Due to incorrect titles or the content mismatch mostly the views are less. This research study proposes a technique for title generation. This research study propose a technique for generating the title using NLP, and Deep learning techniques. The lda2vec and Long Short term memory are used to generate the titles of talk shows. We have compared the performance of LSTM model with lda2vec model.The proposed technique is validated on Custom dataset and all the news dataset and reported good results. The LSTM model generate good title as compare to Lda2vec model. The LSTM model acheives higher accuracy as compare to lda2vec model if we increase the number of epochs to train the model.