There are really only tools available to search through shallow features from podcasts, like episode title or description. I wanted to solve the following objectives with my visualization techniques:
Podcasts are excellent for language learning. One tool that would be very useful is a "grading" tool. Something to evaluate the level/complexity of the language used, the clarity of speech, the speed of elocution.
These are great suggestions! I especially like the second; I was blind to the fact this was not apparent.
For 3: The intro-chart is definitely a bit opaque. I think it makes a little more sense to have with the JRE visualization. It's neat to see at a high level the views/like ratio over the entire catalog. What specifically irked you about having two y-axes?
Some of the charts could do with an explanation of what they are showing. For example, what is a ‘like ratio’ and what do the axis of that green line indicate?
I agree... Like-ratio is definitely an unnecessary way to say "percentage liked" (numLiked / (numLiked + numDisliked)).
The green line indicates the like ratio of episodes over time. The top of the line is green, indicating a high (relative to other episodes) like-ratio. The redder the line is, the (relatively) lower the like-ratio.
Again, perhaps my copy is a little too terse; I'm not the best communicator.
* Screen time is pretty interesting because of the variance. I think I like the episodes where Lex talks more, but with this I can check my idea of when Lex talks more vs what the actual screen time ratio is. (In the Anca Dragan episode (and maybe others), the colours used are very similar, makes things less clear.)
* I was confused about "views" and "likes" until I saw it was from YouTube (which is also a limitation, most podcasts don't have YouTube channel or the YT channel audience is much lower). (Btw, it says "I used youtube-dl to download all of the videos publicly available on the JRE channel", should be Lex Fridman's channel for this site.)
* I'm confused about the second graph: how should I be using the left and right axes? Things don't seem to be vertically distributed by views, and I find it unlikely things fall in the "under 20% likes" vs "over 80% likes" bins so sharply, so I must be misreading it.
* The blue banner below Screen Time looked like a footer to me, so I assumed that's where the page ended. It took some half a page of scrolling beyond that, for the rest of the stuff to load.
* The "Search for a different episode" under Episode Similarity doesn't seem to be working. I click it, search for Sean Carroll (and it lists a dozen things, a lot of which seem to be his own podcast episodes), I can't click on any of the results. I can't even get out of that Search screen, reloading seems the only way out.
These are really great points; I appreciate the feedback!
> In the Anca Dragan episode (and maybe others), the colours used are very similar, makes things less clear.
Very true, I need to come up with a better color scale for this visualization.
> I'm confused about the second graph: how should I be using the left and right axes?
This was definitely the most poorly executed visualization. It's meant to show you the catalog of episodes twice: once for the like ratio over time (at the top of the chart) and once for the views over time (bottom).
> The blue banner below Screen Time looked like a footer to me, so I assumed that's where the page ended
Really great point, updated the copy on this one - thanks!
> The "Search for a different episode" under Episode Similarity doesn't seem to be working
You're right, that was a pretty obvious bug I should've caught. It's fixed, thanks!
Visualizations look nice, though this site could use a bit of introduction and context. I’ve never heard of this podcast so I didn’t know if “Lex Fridman” was the author of the site or the host of the podcast. In fact, if it weren’t for the HN article title, I wouldn’t have immediately understood it was about a podcast at all - it’s not mentioned until half way down the page.
Just a brief sentence or two at the top to give context would help immensely.
Perhaps put a little description of what this page is at the top (An analysis of Lex Fridmans podcast popularity.) For those of us who didn't know who he is.
The top graphs has your drawing continuous lines between discrete points. You should pobably show the points on the top graph. (the lines are trend lines) its clear the like ratio stays pretty constant no matter what. You could put the podcast guest on the x axis too.
If you can gray the area in the bottom graph you are expanding in the above view that would be cool. Its moving forward in time automatically, but at some point it seems to run off the end (I"m not sure if I'm supposed to be able to manually set it)
In the topics/time the vertical scroll bar is a little strange but I'm not sure what could be better.
Why was I not surprised to see that David Fravor (Navy Fighter Pilot, 'credible' UFO sighting) garnered the most views by a wide margin in the dataset? Sigh...
19 comments
[ 48.7 ms ] story [ 701 ms ] threadhttps://joerogan.faith/ http://lexfridman.faith/
There are really only tools available to search through shallow features from podcasts, like episode title or description. I wanted to solve the following objectives with my visualization techniques:
- Topics Over Time: How does the term frequency of a topic change over time? https://i.imgur.com/f0BX7mx.png
- Screen Time: How much speaking/screen time does each participant take up during a podcast? https://imgur.com/a/P76ewPh
- If I like episode X, what other episodes are similar in content?
Looking for feedback and suggestions, thanks!
1) add y-axis labels 2) indicate to the user that they can scroll for more content 3) charts with two y-axes!? really!?!
For 3: The intro-chart is definitely a bit opaque. I think it makes a little more sense to have with the JRE visualization. It's neat to see at a high level the views/like ratio over the entire catalog. What specifically irked you about having two y-axes?
The green line indicates the like ratio of episodes over time. The top of the line is green, indicating a high (relative to other episodes) like-ratio. The redder the line is, the (relatively) lower the like-ratio.
Again, perhaps my copy is a little too terse; I'm not the best communicator.
* Screen time is pretty interesting because of the variance. I think I like the episodes where Lex talks more, but with this I can check my idea of when Lex talks more vs what the actual screen time ratio is. (In the Anca Dragan episode (and maybe others), the colours used are very similar, makes things less clear.)
* I was confused about "views" and "likes" until I saw it was from YouTube (which is also a limitation, most podcasts don't have YouTube channel or the YT channel audience is much lower). (Btw, it says "I used youtube-dl to download all of the videos publicly available on the JRE channel", should be Lex Fridman's channel for this site.)
* I'm confused about the second graph: how should I be using the left and right axes? Things don't seem to be vertically distributed by views, and I find it unlikely things fall in the "under 20% likes" vs "over 80% likes" bins so sharply, so I must be misreading it.
* The blue banner below Screen Time looked like a footer to me, so I assumed that's where the page ended. It took some half a page of scrolling beyond that, for the rest of the stuff to load.
* The "Search for a different episode" under Episode Similarity doesn't seem to be working. I click it, search for Sean Carroll (and it lists a dozen things, a lot of which seem to be his own podcast episodes), I can't click on any of the results. I can't even get out of that Search screen, reloading seems the only way out.
> In the Anca Dragan episode (and maybe others), the colours used are very similar, makes things less clear.
Very true, I need to come up with a better color scale for this visualization.
> I'm confused about the second graph: how should I be using the left and right axes?
This was definitely the most poorly executed visualization. It's meant to show you the catalog of episodes twice: once for the like ratio over time (at the top of the chart) and once for the views over time (bottom).
> The blue banner below Screen Time looked like a footer to me, so I assumed that's where the page ended
Really great point, updated the copy on this one - thanks!
> The "Search for a different episode" under Episode Similarity doesn't seem to be working
You're right, that was a pretty obvious bug I should've caught. It's fixed, thanks!
Edit: I b dumb
Just a brief sentence or two at the top to give context would help immensely.
Perhaps put a little description of what this page is at the top (An analysis of Lex Fridmans podcast popularity.) For those of us who didn't know who he is.
The top graphs has your drawing continuous lines between discrete points. You should pobably show the points on the top graph. (the lines are trend lines) its clear the like ratio stays pretty constant no matter what. You could put the podcast guest on the x axis too. If you can gray the area in the bottom graph you are expanding in the above view that would be cool. Its moving forward in time automatically, but at some point it seems to run off the end (I"m not sure if I'm supposed to be able to manually set it)
In the topics/time the vertical scroll bar is a little strange but I'm not sure what could be better.
Despite these very minor points, I like the page.