For my Programming A to Z class final I wanted to keep experimenting with language. I decided that the best way to do it would be through TED Talks since so much of my experience before ITP was based around them.

For my Final I created a Sound Like A TEDx Speaker generator that takes user story, meshes it together with the content of one of three most popular TED speakers and generates new text for user to use.

Why I chose to do it? It roughly takes two to three months to prepare a TEDx Talk: from initial conversation, to drafting proposal, to countless numbers of Google hangouts until a speaker can hold a final up to 18min talk text in her hands. Such intense preparation is necessary not to only have a text to say on the stage, but to have a story where place of every word makes sense.

Which TED speakers? For a test run I chose 3 most viewed TED speakers, whose texts I was working on: Ken Robinson, Amy Cuddy and Simon Sinek.

I was having a hard time to decide how exactly the project should look like, so I decided to make two prototypes.

Version1: is done using RiTa.js library and RiTa.getPosTags function, which means that every time user inputs her story, RiTa.js library will divide it by parts of speech and then will replace them accordingly with parts of speech from one of the speakers’ talk.

Version2: is done using Markov chain generator. After user submits her story, it will be meshed together with a selected TED Talk text and using n-grams and Markov chain new talk will be generated.

Here are examples of both:

screen-shot-2016-12-14-at-18-40-48 screen-shot-2016-12-14-at-18-40-58

Below you can find:

  • github repo of RiTa.js prototype and code running for you to try
  • github repo of Markov chain prototype and code running for you to try

Leave a Reply

Your email address will not be published. Required fields are marked *