Annotate "How to Get Startup Ideas" | Hacker News Reader

Annotate "How to Get Startup Ideas"

annotatepg.com

41 pointsjulien_c40 comments

I recommend ditching the pseudo-book UX with a page that scrolls vertically (like the rest of the internet). Clicking to navigate through several pages reminds me of this tweet I saw the other day.

"I like this thing that I'm reading on this website, but I wish it was spread across more pages." –Nobody, Ever

Otherwise, looks great

Definitely needs tweaking (use more of the blank space), but I disagree. Scrolling is much more difficult on the eyes, since you need to track the last line you read as it moves vertically. Do this 100 times in a row and it gets pretty annoying.

There was a really cool project a few months back of a variation of vertical scrolling, where the current page is "peeled" back revealing the next page from top to bottom.

The effect was you can read as you normally would but lines are stationary so you don't have to track them. You'd reset your eyes to the top of the page after finishing one. Not sure how I'd track it down sadly.

I believe this is the project you're referring to, though it's now integrated into an ebook reader web app: http://www.magicscroll.net/

You're thinking of magicScroll : http://www.magicscroll.net/ScrollTheWeb.html

I somewhat disagree. I don't like having to wait on page loads, but I much prefer the Kindle-style book scrolling (with no page loads) to vertical. If it doesn't have to load pages, book-style works better for my eyes. I believe there was a Show HN involving a plugin that did this that had some research on why it was preferable.

Thanks for the feedback! Updated: http://www.annotatepg.com/

First off, awesome. I've always wanted something like this (especially for pg's essays) and kicked around ideas for how to build something like it. A few bits of feedback, mostly on the design/interaction side: I don't think only offering twitter sign in is ideal, the pagination was not very intuitive/easily discover-able (horizontal might be ok in mobile context with swipes, but on desktop people want to scroll), the function of the bottom progress bar is also not intuitive to me (seems more like a zoom at first glance), the content itself is a little too "boxed in"- don't think there is a need to contain it like that, and lastly the mobile experience (Gnex) is lacking (everything is quite small, ui nearly impossible to interact with).

Totally understand that this is just a demo and you've probably thought of this stuff already but thought I'd chime in just in case it is useful. Great work!

Are the documents hosted on here intended to be public or privately shared? If the former, I'd worry a bit about clashing with what RapGenius are doing.

Web document annotation isn't a new idea - I worked on a product called Reframe It 5 years ago doing something similar: https://www.youtube.com/watch?v=U1swp-2QeMI , and we weren't even close to the first to market: http://en.wikipedia.org/wiki/Web_annotation .

Just like everything else, there's room for more than one player in a market - why should they be worried about competing with RapGenius?

Interesting - I was working on DotSpots around the same time. ReframeIt was one of the people we figured were in the same space.

I've also hoped that web annotation will take off one day, but it seems fated to sputter every time.

We sure were! Fun to see someone we considered a competitor all those years ago on HN also.

I think Rap Genius are on the right track: hosted specific-document annotation seems like a good method to develop a web-annotation technology while still making money at the same time. We had a lot of interest from specific-document research organizations like lawyers and journalists, but the "annotate the whole web" pitch didn't seem to go over too well.

Sorry, didn't mean to imply 'RapGenius is doing this, so you should not do it'–I just meant that 'RapGenius is doing this, FYI you're competing with them'.

products can be technically similar without really competing if they're targeting different use cases

The problem is RapGenius is trying to cover basically all of the "annotate X" usecases. For example, they already have some of pg's articles (Though they need some love! http://rapgenius.com/Paul-graham-cities-and-ambition-lyrics and rapgenius.com/Paul-graham-startup-growth-lyrics if you're feeling helpful), Douglas Adam's speech "Is there an Artificial God", a strong and expanding base in poetry, and are not shy about their plans to annotate everything.

Cool. Presents very well. Just to reenforce points others have made:

* Why can't I scroll? ... scrolling is how I read

* That "x" to close the comments pane is very small... I probably clicked around 3 or 4 times trying to get the thing to disappear.

This is a great product idea, and a very simple yet powerful take on its execution. Others have pointed many ways it could be improved (scrolling, the "x" on the comment panel, etc.), but I cannot emphasize either how interesting this could be.

This makes me think of "Project Syndicate", at http://www.project-syndicate.org/, the public debate venue for economics/public policy thought leaders from across the board. They have an interesting discussion system based on comments users post about paragraphs. Maybe there's an avenue for future cooperation there?

One way the product could improve is if the social component of the annotation concept was made more obvious. An explanation of what happens when someone comments would be welcome.

Also, will the be available as a private/semi-private product? I could use this for my Kindle books, but would want to share only select highlights among the many I make.

Overall, great job! Good luck with the YC application!

Nice!

Curious to know what you're using to parse the essays into sentences. Isn't that quite difficult in the general case?

Thought I'd jump in here. Yes, parsing sentences are actually quite difficult.

Disclaimer: I am currently working on a text summarization startup, http://Summary.io

There are a few different approaches to parsing sentences, and only a few giant NLP libraries. What I have found the best in 95% of use cases to to write custom (RegEx) rules.

Attempting a sentence such as: And then Mr. Bean (http://www.mrbean.com/?index) said to Col. Sanders, "Holy moly sentence extraction is hard!"

You have lots of little things like making sure you take the full quote and disregard periods in surnames, http links, etc. Parsing just by periods, question marks, and exclamtion points are going to lead to a lot of problems.

Looks to be mostly just splitting by the period -- see article "Startup == Growth", page 17 near the bottom (splits on 1.7x at the period, although it shouldn't)

Better algorithm, just as simple, would be splitting at ". ". Making something smart enough to not split at "fig. 1" and the likes would be more difficult.

Sentence detection is a hard problem, and I don't know that anybody has solved it for the general case. But pragmatically speaking, it is solved for many - if not most - common cases, and there are Open Source NLP libraries out there that do a pretty good job. For Java, you have OpenNLP and Mallet, the C/C++ world has Ellogon and Freeling, Python has NLTK, etc.

Some thoughts:

1) When I click the "list" button at the top next to the essay name, it shows me the same list that is on the left side of the screen. Why? Is there something I can do there that I can't do somewhere else?

2) Why make your auth twitter? I see no value in tying it to twitter unless you are expecting people to tweet all their annotations, which seems unlikely.

3) There's a ton of unused white space on the right, which makes the app feel unbalanced.

4) The site is not very mobile friendly -- not a knock against you, but since it looks like you are already doing responsive css, I thought the mobile site would be much tighter.

5) Whoa, why are there dozens of media queries in there -- are you really changing things every 10px?

Looks cool! I'd like PG annotations to be integrated as annotations from users, so you can read them and follow your reading easily (his comments can be huge but quite informative).

After you click on a comment, I found closing the comment pane less simple than it could be. The x to close it is fairly small and hidden, and I think it would be nice if clicking on the comment bubble again just toggled it away and/or clicking on the text still visible on the left slid the comments away.

Seems quite similar to rapgenius

Looks very nice.

I think allowing the article list to hide would make the app feel more balanced. Right now the main part of your app takes up a very small real estate on my screen.

Good luck on getting into YC!

Is there a way to create a link directly to a specific annotation? I was looking for a quick way to share a quote/annotation with someone. I'm not interested in using or signing in with Twitter.

Great idea, i just pushed a modification to have urls pointing to specific quotes. Now you can copy the url when the comment panel is opened and share that url with anybody.

Nice work! I look forward to the royalty check in the mail. :)

Original poster here. You want to link to a quote by pg, not an annotation you've written, right?

Yep. I don't really have anything to add to the quotes but I'd like to share them and point back to the source easily.

I think your site needs to make the original source more apparent/clickable as well.

It took me a minute to figure out that I needed to click the margins to turn the page. Maybe add a small arrow or other indicator other than cursor:pointer. Other than that, very nice.

That's really cool. But oh-my-god let my mouse wheel scroll.

This is an awesome demo. Way to go and best of luck. I have a feeling you will go far.

Ahh, rapgenius for normal docs!

Awesome product as usual!

How is this supposed to be monetized?

Can't scroll?

good idea, terrible interface :(

I like it, simple and elegant.

Reply on news.ycombinator.com