EXPLORATORY DATAANALYSIS AND TOPIC MODELLING ON TED TALKS

Abstract

TED (Technology, Entertainment, Design) is a non-profit organization that influences the audience across the globe to deep dive into thinking .The short, powerful talks in more than 100 languages, from great inspired achievers engage the curious people and change their way of perception about issues on science, entertainment, business, technology, global concerns and various other topics. Why do some TED Talks get more views, go viral? What makes a TED talk the change maker in outlook, attitude and behaviour? What intrigues and influences people? This paper aims to analyze the various drivers behind maximum view count of certain TED talks from the start of 2006 till the end of June 2020. The analysis takes into consideration various parameters such as the speaker’s profession, chosen topic, his/her transcript, number of views, comments, tags to name a few. This analysis will help an aspiring TED Talker identify the drivers and plan a video that will attract more view counts. This research paper uses Exploratory Data Analysis with a special emphasis on Natural Language Processing using the native Latent Dirichlet Allocation model from Gensim and the LDA Mallet. Analysis of the TED Talk data suggests that the content is a prime driving factor rather than the public speaking abilities thereby making the view count less predictable.