There are a lot of AI narrator services you can use to create and record custom podcast. There is one called narrationbox which I specially use to create Audio Books from ebooks. You can use it to record custom podcasts as well.
That sounds awesome. I haven't produced content in so long because I'm super particular about technical explanations and have to start over if I don't get it exactly right.
A couple of feedback points (which you may already be aware of):
1. I did occasionally find that not enough info was shown for papers to determine if they were really in my areas of interest, so perhaps it's worth a look to see how to show a bit more (my work around was to view the abstract but that kind of defeats to point!)
2. The ability to keep playing whilst you search further pages is cool. However I noticed if you search for a topic that returns lots of papers and you go on to a later page (eg you've navigated to page 4 for example) if you then search for different papers that return fewer hits you are shown a blank page, not necessary because there are no hits but because they don't populate four pages (in the example that I'm on page 4). I'd expect most people would want to return to the start of the results with each search
If you end up in the above situation, you can jump back to populated pages but it's not what you'd typically expect as a user.
3. The abstracts often have "\n" sequences present as actual characters in the text, which don't match up with the actual line breaks - presumably they're left over from the source but it looks like they ought to be stripped out.
1. Yeah, there’s not much that the app itself can do about this. My intention is to make it dramatically easy to access the abstract so even if you’re the slightest bit interested in the title you’ll go check it out (on desktop you just have to hover over the abstract link to see it pop up).
2. This navigation + search issue is almost surely a bug. Not sure if I ever tested navigating ahead and doing a search, like you did.
3. Do you remember whether you saw it only on older papers (basically when navigated into further pages within some category)? Long story short, many of the older abstracts are littered with redundant \n characters because of a botched up Postgres data export + imports I did sometime back. Papers in the last week or so should be fine.
This is really neat. One piece of feedback is it begins and ends TeX expressions by saying “dollar” which is distracting. Probably best to strip the TeX syntax while retaining the expressions. Simple ones like O(1) should be understandable aurally, even if complex expressions may not be.
I'm the original poster. I felt encouraged to do a Show HN for Paper Time after seeing positive feedback in a different, side projects related thread a few days back [1].
I'll be around for a while so go ahead and ask me any questions, either around the idea or the technical parts.
A big shout out to the folks on HN who responded with very pertinent resources about UI design a few months back [2], when I'd just started building this.
1. A daily RSS feed - perhaps it could compile them into a 'single' podcast.
2. Human voice uploads - allow readers to upload their own voice (would require moderation - but would solve most pronunciation issues caused by TTS). You could randomly assign some number of articles to each moderator to read.
Daily RSS feed sounds like a good idea. RSS used to have the concept of media enclosures or something. Perhaps if I can embed the audio into that, people can listen right there? On the other hand, that will have fewer people visiting the website :D Heh.
2. My somewhat long term product idea is to let the paper submitters volunteer to read out their abstracts, if they choose. But some amount of automated TTS will always remain, given the sheer volume of content.
would really like a "curated playlist" option. The ability to filter based on "Topics" is a step in that direction and a further vetting down of options will make the app more usable.
nitpick: the TTS ends up pronouncing the markup as well. maybe the abstract can be cleaned before being passed through the TTS-generator.
eg. for the line: "This paper concerns designing distributed algorithms that are {\em singularly optimal}"
Haha, had never thought of it as an abstraction over abstracts! Thanks for the detailed feedback.
The TTS cleanup ought to become one of my top priorities. I’ve heard a lot of $k$ (dollar k dollar) is etc. I’m guessing it’s LaTeX markup bleeding over into the abstract text rendering somehow.
I am not able to figure out how to curate a list, since I’m just processing Arxiv papers as they come in! One option is to provide papers accepted at conferences.
Another is to let people upvote, but I fear voting dynamics can get messy. Also, without sufficient users, voting isn’t of much use.
I’ll see how I can implement permalinks. I’m not too familiar with doing that in single page apps so I need to work on it. Related question: would you like to be able to star/bookmark papers so that you can come back to them?
Nice, looks great! Could you give any thoughts on which of those resources in [2] were particularly useful for you? I have a similar backend background and interest in learning about UI design!
BTW a small issue but opening and closing your FAQ modal seems to be pushing URL history - which means you have to hit back repeatedly to exit the page if you interact with it. It doesn’t look like having the modal open and then going back closes it, so I think you could just drop the URL push without losing functionality.
Refactoring UI was a great resource that was recommended on that Ask HN. The authors have a book and some videos. Very actionable and systematic advice. That formed the basis of my understanding.
For more specific techniques, the Learning Web Design 5th Edition (by Jenna Something) is very good.
I skimmed Don’t Make Me Think, which validated some of my own thoughts and helps you avoid silly oversights. I also read Design of Everyday Things but found it very lengthy and somewhat pompous. Not sure that I got much out of it.
On that page someone has linked to a bunch of MIT material. I didn’t check that out in detail. Might be good... I don’t know.
Thanks for the bug report on the modal. Would have never noticed as I navigate to my site directly and never click the Back button there.
Hi LoveMortuus,
I have started off with CS papers only, and that too only a small subset of that (that’s why the website’s tag line says CS Research). May come around to adding more in due course. Is BioArxiv a good source of papers?
@pramodbiligiri, would you be willing to share a little bit more about how the thext-to-speech is utlized in your project?
What provider do you use, the costs related to it, is it a real-time text processing, or maybe you have a bulk processing routine, opinions on different TTS providers? You know, things like that for everyone who'd like to try TTS in their projects.
(We're talking about >3000 abstracts TTS-ed on the author's page, and that's not a trivial amount if one is for instance using external providers for the job).
Sure, it's not exactly rocket science. I'm using Google Cloud's Text-To-Speech - https://cloud.google.com/text-to-speech. You should be able to find the pricing details there. I'm guessing other cloud providers' TTS would be comparable as well? I haven't checked.
I am not a PhD but I like to read cognitive science and astro papers. I was wondering if there is an annotated version of papers available? Like a rapgenius for academic papers.
Do you mean like Fermats Library? People can share annotated papers there. There aren't a lot of papers though, but i found some good ones outside of comp-sci recently.
This is a really interesting idea! Audio definitely makes it easier to listen to papers on the go.
One thing that's missing in the paper curation space is understanding the knowledge state of the user. Once we have that we should be able to build generative models that synthesize text that can get people to read papers much faster.
I've been wanting to cultivate a habit of reading a relevant paper (or abstract) at some cadence, such as once a week. Can anyone recommend a way to find the important and/or influential papers in a given field? Maybe something similar to The Morning Paper.
2 minutes papers on YouTube is pretty good for physics simulation and AI papers. Plus you get a dude giving a quick explanation why the paper is relevant and important so you don't have to go in completely blind.
You can choose the topic you’re interested in and either go for “test of time” awarded papers in the field, or pick up a list from a graduate course in any University. That way the professor would have done the curation for you.
This is a good question, but I’m not sure if there’s a good answer.
One thing I liked about grad school is that there would literally be classes where students were assigned to read a foundational paper in the field and then discuss/write about it, so my personal go to has been to try and find “Intro to X” grad level courses and see their paper assignments (most professors tend to put most course material online).
It’s not something guaranteed to give you results but often I will stumble into interesting stuff I wouldn’t have otherwise.
> This is a good question, but I’m not sure if there’s a good answer.
I think you've actually given a great answer here. This is essentially what I did in grad school as well and found it really helpful. I never thought to continue doing this in fields that piqued my interest. Thanks for the suggestion!
While we're on the topic, here's the advanced distributed systems reading list from UIUC:
That is a known bug. Actually I haven’t even come around to fixing such issues yet.
Another thing you should probably (not!) try: Click Play on a paper and click on Next/Prev :) The only way you can recover from that is by refreshing the page. Sorry about these rough edges as of now.
Papers are obtained from Arxiv.org on a regular basis. Their license for metadata allows one to build these kind of things (I am not a lawyer).
Thanks for the bug report. Have to double check once I get to my laptop, but that’s probably intended behaviour. Select All is a short for... Selecting all the topics. Once you uncheck it, you can select and deselect individual topics at will. Let’s say I clear all the topics as you suggested, what should be displayed in the list of papers then? Anyway, wouldn’t defend that dropdown too much, at this point!
