Natural Language Processing and Naval: The Art of the Podcast

Podcasts are a harmonic creation.

They are fascinating tools, connecting us to the speakers in a very personal way, almost like we are a part of the conversation too. I laugh out loud at some things, or nod along, reacting emphatically, despite being in a different time and place than both the interviewer and interviewee.

Podcasts are tools of connection. Almost everyone either wants to start a podcast or knows someone who has started a podcast (for better or worse).

Podcast creation is the perfect storm of an open consumer base, an ever-evolving Internet, and accessible recording software and hardware. The barrier to entry is low. In the most basic approach, all you need to have is a cellphone and Internet access.

We also like to hear from online ‘celebrities’ or people that we greatly admire, and use the content for learning and entertainment. It’s easy to go onto Spotify or Apple to get access to a wide variety of podcasts, on an unfathomably wide variety of topics.

What’s the best podcast?

When Austen, the CEO of Lambda School and someone with 130k+ followers on Twitter, tweeted this, I was curious to see what people would say. I will caveat this with the fact that most of Austen’s followers are tech / tech-adjacent people (most responses had the word ‘software’ or ‘building’ in their bio) so the responses are likely biased into the tech sphere.

Image for post
Source: Austen Allred

I mined their responses in a spreadsheet.

Image for post

The most recommended podcast was The Tim Ferris Show (16% of all responses, n = 208) followed closely by Joe Rogan, as shown above.

Specific combinations were recommended as follows:

Image for post

Out of the 208 responses, 8% of people specifically recommended the Joe Rogan + Naval combination. 7% specifically recommended Naval and Shane Parrish of The Knowledge Project. Another 2% was Naval’s own Podcast, How to Get Rich.

Naval represented a total of 24% of all recommendations that specified an interviewee (n = 148, 35 responses for Naval). Peter Thiel and Derek Sivers followed, making up 5% and 4% of all responses, respectively.

Image for post

The Knowledge of Naval

Naval is fascinating, and it makes sense that he is a quarter of all interviewee recommendations because he is in his words, “a hero among young male geeks”.

Naval represents a next-level sort of thinking, and voices what a lot of us feel but don’t always vocalize (the future is entrepreneurship, power over the monkey mind, the debugging, the duality of thought, etc. etc.).

Eric Jorgenson recently wrote The Almanack of Naval Ravikant which captures a lot of his wisdom. There are several hundred threads on Twitter with key takeaways from the book, as well as from Naval’s interviews.

Even just running the frequency metrics for words from the transcript from The Knowledge Project, you can see why Naval is so admired in this space.

Image for post

The word ‘think’ is mentioned 161 times in his conversation with Shane Parrish. The word ‘can’ is mentioned > 100 times, tied with ‘people’ and ‘read’. From what I can gather from Naval, those four words encapsulate a large portion of what he seems to value and care about.

But let’s do a deeper dive.

Text Mining of Naval

I pulled the Naval’s interview transcript from The Knowledge Project Podcast, The Tim Ferris Podcast, and Joe Rogan’s podcast. I used the R packages for Natural Language Processing to get a sense of sense of sentiment, emotion, and word association.

I used the NRC Word-Emotion Association Lexicon (aka EmoLex) which was developed by Saif Mohammad, a researcher at NRC. EmoLex divides the words into eight emotions: anger, fear, anticipation, trust, surprise, sadness, joy, and disgust. The words are segmented via crowdsourcing (you can read more here if you’re interested).

Image for post
Source: Safi Mohammad

Emotion Association and Sentiment Analysis

The Knowledge Project

On the Knowledge Project, a lot of the conversation fell into a ‘trust’ emotion. Almost 400 words. Words like wisdom, wealth, thoughtful, etc.

Image for post
Plot of Shane Parrish and Naval: Distinct Emotions

On a percentage basis, both trust and anticipation outweigh any of the other emotions.

Image for post

The sentiment scatterplot is below, with most of the conversation remaining positive over time.

Image for post

The average sentiment of the conversation was 0.077, the most ‘negative’ out of all of the conversations.

The most positive sentence:

If you can be more right, more rational, and that’s one of the reasons why I love your blog because it really focuses on helping you be more right, better decision-making, more rational, then you’re going to get nonlinear returns in your life

The Joe Rogan Podcast

For the Joe Rogan podcast, there was much more anticipation and joy, relative to trust. Trust still dominated.

Image for post

There was a more even dispersion across the emotional spectrum too, with a decent amount of fear (~10%) interjected into the conversation.

Image for post

An equal amount of sadness and anger was represented here too, but the conversation was relatively positive.

Image for post

The conversation was very positive in terms of sentiment, scoring 0.0985.

This was the most positive sentence: Very professional, very quick, very thorough but he did more diligence on me than I did on him

The Tim Ferris Show

Tim Ferriss and Naval had a very trustful conversation.

Image for post

There was more trust, more anticipation here — this podcast was very similar to the conversation Naval had with Shane Parrish, according to this distribution.

Image for post

There was an average sentiment of 0.0901.

The most positive sentence: Most of the ways we try to get peace from mind are indirect, whereas if you understand things if you see things properly you will naturally slowly develop peace from mind

Most negative: “So you have to ruthlessly, ruthlessly disappoint everybody”

Image for post

The Podcasts

The Knowledge Project

The Knowledge Project has an interesting word frequency distribution, with the most common words in the podcast descriptions being ‘author’, ‘making’, ‘can’, ‘learning’, ‘life’ and ‘world’.

Image for post

Econ Talks

Econ Talks was another top podcast, with the word University the most prominent (likely due to the interviewee’s job title) as well as argues, author, book, and policy.

Image for post

The Tim Ferris Show

The Tim Ferris Show was lessons, life, master, and building.

Image for post

Joe Rogan didn’t have descriptive podcast names (just the names of the interviewees) so no word cloud for JRE.


There is power in trust. There is a thin line between a good podcast and a great podcast, and the interviewee is 85% of that. A good interviewer is just as important — someone who lets the interviewee talk, but also contributes to the conversation meaningfully.

Conveying trust + anticipation (hope) seems to be a strong combination for a successful podcast (at least for Naval).

I will also highlight that in this dataset, there was a lack of diversity, both in terms of race, gender, age, etc. I’ve included some excellent podcasts below for even more thought and perspective. Please comment any below!

3 responses to “Natural Language Processing and Naval: The Art of the Podcast”

  1. I love your statistical analysis of human behavior.


  2. […] piece of writing you’ve seen last week was an article by Kyla Scannon on natural language processing of podcasts’ content. The article specifically […]


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: