Show HN: A little knowledge portal I've been building | Hacker News Reader

Show HN: A little knowledge portal I've been building

tuvalie.com

72 pointstuvalie49 comments

Regarding the UI, personally, I would shrink things down a notch. It feels a bit "in your face" right now, if that makes sense. Besides that - love it!

Edit: even just changing font-size from 14pt to 11pt is much better imho.

Hi, eric. Thanks for the feedback. If you don't mind me asking, do you think it feels that way because of the color choices, spacing, order of content, or something else? The colors are tied to specific themes, which can be changed by registered users -- but I can still tweak the default theme if it would improve things.

(not eric, but had the same thought)

Everything just feels way oversized, in a Fisher-Price kind of way. The font, the colors and "chunky"-ness of the UI.

The full-width content-area doesn't help either. The whole page gets right up in your face and shouts the content at you.

Concrete suggestions:

- dial down the font size

- put some blank space in the left-right marigns of the page (responsive if possible).

- maybe choose a more passive shade of blue for the background.

EDIT: here's an example comparing your page with wikipedia http://recordit.co/Za7NuT8Hy7

> Everything just feels way oversized, in a Fisher-Price kind of way

I'd also recommend getting rid of the icons on the main menu and replacing them with text (with a small icon next to them if you must). I had to hover over the icons to figure out what they do.

This is one of the things I like about Bootstrap, as a developer I can now focus on building stuff without having to care about making it look nice (which I'm not good at). I just add Bootstrap and for 90% of things (I usually do business web apps) that's good enough.

Thank you! I'm going to work through those suggestions later. Do you mind if I ask for your feedback afterwards?

Sure, go for it :)

I made a couple of small tweaks. If you have a moment and are willing to take a look, I'd love to hear your thoughts! (You may need to perform a hard reload in your browser.)

Hi!

Here's just a quick comparison between your version and a version where I changed some small things. http://imgur.com/a/w1bUV

font-size 11pt body width 80%, margin: 0 auto; header height 40px and header > img 35x35px.

Hi, eric. I haven't touched the header yet, but after doing a hard reload in your browser, can you tell me if you like the body style more? (And thank you!)

Hey folks, I just wanted to apologize. I got an influx of several thousand views because of this post and exceeded some API usage limits, so the web and some of the image results were down for a bit. They're back up now, but if they exceed the limits again, they'll have to remain down for a bit (financial limitations :S).

I don't have a comment on the work itself, but like most sites, it has no awareness of line length, and it's therefore painful to parse the information without resizing the window.

Both HN, and the linked website, could benefit greatly in readability by simply limiting line-length to ~32em for textual information. It's strange that a browser, designed for human use, has no notion of humans in that regard.

Hi, xjay. Thank you for bringing that to my attention. I'll experiment with it later on today. If you don't mind me asking, do you have a reference site I can look at that you think really shines in terms of readability?

Sure! Matthew Butterick opened my eyes on the world of typography. He has a site with tips: http://practicaltypography.com/

He also has a very good talk called "Typography for Docs": https://www.youtube.com/watch?v=8J6HuvosP0s

Those are perfect, thanks! After I'm doing fiddling with things later, do you mind if I solicit your feedback?

I'd say that resizing windows to make text readable should be normal and using maximized windows on todays large displays makes little sense, but... I was looking at that site myself on a maximized window :-) Why? Because I had other tabs open with developer tools and I needed all the space. So yes, sites should set line length and line spacing appropriately. When we see lot of white space around a column of text we should know that our window is too large. By the way, there are CSS multi column properties to fix that

https://developer.mozilla.org/en-US/docs/Web/CSS/CSS_Columns...

https://www.w3.org/TR/css3-multicol/

It's supported by all browsers. Funnily IE leads the way here, the other browsers require vendor prefixes.

http://caniuse.com/multicolumn

To turn it around; maybe browsers don't utilize space very well?

Imagine if we could divide a tab into regions of sites instead?

Having a maximum line length is the humane thing to do, unless the person has phobias of empty spaces, of course. :>

The links you provided are not for the user, but for the designer. Text, however, is a global means of communication, so it should get better treatment, and care in general. Especially by browsers.

Hey, xjay. If you do a hard reload in your browser, you should see that the encyclopedic data is now bound to a 32em width (with a smaller font size). Does that look a bit better to you?

The idea is to have a maximum line length that makes sense to human beings, and that's what the site does now. It is very pleasant to read the text.

Thanks for considering the feedback.

PS: The typographical rule I based 32em on comes from the general rule that, there should at least be room for two alphabets (a-z) on one line. I found 32em to be a sweet spot. The em unit scales in proportion to whatever pixel size you choose for the font, or whatever font size I tell my browser to enforce, so it's highly consistent.

PPS: Another bonus for having a sensible maximum line length, is something I'd never thought about unless it happened to me. It's the nuisance called eye floaters. It's basically an obstruction that floats around in your vision. So whenever the eye moves, the obstruction will dash across your field of view to follow, before coming to rest again. The less the eye has to move in general, the better.

PPPS: For HN (and wikipedia) I eventually ended up making a CSS rule that would override the width of various sections, to comply with the human inside. It's just that there is no line-length attribute, so it must be hacked together specifically for each trouble site.

Thanks for the info, and again, I really do appreciate the feedback! I don't want people to have to make up their own rules for things simply because the defaults aren't sensible. For this particular project, certain UI considerations were deferred until later. I never expected this post to garner more than a handful of views, so needless to say, it was an abrupt call to action to start fixing those things once it broke 10,000+ views.

Nice work, compliments. I also like the UI, no clutter or heavy JS. But I suggest you add a little noscript message saying this only functions with it enabled.

That's a great idea, thank you. A little ashamed I overlooked it, actually :/ It'll be thrown in with my other to-do's for the day.

No problem, even the no-js users overlook it, including myself...

Hi, gkya. The noscript element has been added. Based on some other feedback I've received here, I also reduced the font size and line width for encyclopedic information display. (It may require a hard reload to notice the difference.) Since you had a positive impression of the original design, it'd be good to hear that it hasn't changed with the update. And thanks again!

Checked it, yes, I see the noscript message. I rather like a more plain style for web pages that is based on text rather than icons, but it is personal preference. The font size is better, and it zooms in and out nicely (I zoom out on it for easier reading). If I would change one thing, that'd be to use a serif for some longer text. You're welcome, and kudos again.

Nicely done! I especially like the timeline. Most of the data for the search results I tried are available via DBPedia and other sources and you did a great job tying it all together.

BTW, re: UI: I found it easy to use, and it looked nice, on the iPad Pro I am using right now. It is kind of people here to offer suggestions for improving the UI, but my suggestion is to leave the UI as is and keep working on the back end stuff

Thanks for the kind words! You might have seen the UI after I just pushed a small improvement through (it was right around the time of your comment), so it may have been less offensive for your particular viewing! (If it doesn't change when you do a hard reload, you saw the updated version.) Hopefully that addresses some of the issues, but I agree that focusing on the core service is more important in the long run.

Great work! The UI is a bit old school -- reminds me of Windows 3.1 crossed with some wireframe creation program, but it is not hard to use. But I especially like the timeline and encyclopedic information. Since I run a search engine myself (solveforall.com), I would like to ask you the sources of the encylopedic information, timeline, and quotes. The result that doesn't seem to add much value is the NELL project one, which lists the type hierarchy. Maybe that could just be summarized to "Albert Einstien is a male scientist" ...

Of course autocomplete would be nice as well as more tolerant input acceptance.

I don't get the finder ... do you plan that people will register, then you can find other users that way? This is not a general search for someone by email, right? That would be cool...

I am looking forward to using your API as well! Do the endpoints on https://tuvalie.com/?do=developers work?

Anyway, cool stuff. If you want to discuss further with me, contact me at jctsay AT aim.com.

Thanks for the feedback! What's being demonstrated here is one part of one feature of a larger project I've been working on over the last several years. I've built a lot of custom parsers that continuously mine/merge facts from various sources. (I think it's sitting at around 215 million atomic facts, right now, outside of what it can dynamically calculate.) The APIs work (theoretically), but they're very rough. Showing this project off is very new to me!

215 million facts, that's impressive! Anyway, congratulations on shipping. If you ever want to collaborate on search, please let me know.

I'd love to find ways to collaborate with anyone who's interested :) I also forgot to mention that you can download an export of (most) of the timeline events from http://endlessorigins.com/ -- I update it from time to time with new events, too. To my knowledge, it's the largest freely available structured data set of historic human events. Could that be useful to you?

That looks great, I am going to look into this at some point. Thanks a lot for the link and all your hard work.

Man, this is almost exactly a project I had envisioned years ago. I was going over a list of terms I needed to know for a midterm in school, and thought it would be neat if I could just plunk in a list of comma-separated terms and have it fetch definitions, wiki blurbs, images, and whatever else it could find in to cards for each term.

Funny you should mention that! For registered users, comparisons across concepts are an actual feature. The service tries to summarize on a per attribute, cross-entity basis what makes each item different or similar. It's pretty computationally-expensive, though, so I didn't want to slow down this demo by showing that off. I can post a couple of screenshots of it in action if there's any interest?

This is so cool, it shows you're a real developer's developer, if you get what I mean!

One thing that confused me a little at first, UI wise, is the scroll bars on the page Web Results and Image Results panels. I was just scanning pages and thought the page was hanging for some reason while using my mouse wheel.

Thank you! I can honestly say that comment put a smile on my face. As far as the scroll bars go, I just didn't want to take up too much screen real estate for those individual sections. (So that people who want to spend time with them can, but those who don't can skip over them a bit quicker.) Do you think there's a better way to handle that? I'm very open to suggestions for just about anything!

Sorry this is a little bit delayed! It's hard to see when you get replies here :)

My feeling is to either make liberal use of "Click to expand" or to just let users scroll the whole page vertically. I think we've become used to long, vertical content elements.

Thanks for anyone who checks this out. I'd like to release a free API soon with most of the data the portal contains. That's one of my big objectives since hearing about Freebase shutting down. At any rate, all thoughts and feedback are greatly appreciated!

What would be really awesome (and probably insanely difficult to make) is an API which does something similar to what Google does when you define: stuff. It usually comes from a dictionary, but sometimes it has more interesting sources. I'd love to put something like that in an IRC bot.

This is even more experimental than the main service, itself, but does something like this meet your needs (genuinely curious)? https://tuvalie.com/fae/?q=Albert%20Einstein&api

While that is useful, it wouldn't meet these particular needs. What I'd want is a single piece of text, with maybe a link, giving a best-effort description/explanation of the subject.

So when specifying "Albert Einstein" Google gives:

> Albert Einstein. BrE. (1879-1955) a physicist, born in Germany, who was possibly the greatest scientist of the 20th century. In 1905 he published his theory of relativity. This led to the equation giving the relationship between mass and energy, E=mc2, which is the basis of atomic energy.

But when specifying "hubbub" Google gives:

> a chaotic din caused by a crowd of people. "a hubbub of laughter and shouting"

Stuff like that, where it (somehow) automatically determines which piece of information is most relevant. Again, I don't think this is really doable without having as much search data as Google does.

Completely subjective, but I'm very drawn to the timeline. I would be interested in a site that was organized entirely around that, with the videos and pictures displayed chronologically and by importance.

I'm a big fan of the timeline, myself! I actually regularly export some of the timeline data to a TSV file for others to be able to use in their own projects. http://endlessorigins.com/ -- I'd love to know what kind of uses you can find for it!

Information looks good. Design looks a bit old doesnt affect anyway but I would recommend updating it. SEO of the site is worse.

What is the plan for revenue? contextual advertisement? Donation? Or something else?

I would love to collaborate and help in UI/UX and SEO. Ping me if you wish to connect: vivek@eyuva.com

Recent work: oizom.com, metavideos.com

Hi. Thanks for the feedback. If I can help it, I don't want any advertising to touch the knowledge functions of the site. It's part of a larger project that offers subscription services and sponsorships for interest communities, so all of this is just one piece to that puzzle. Donations would be wonderful, but I don't really see getting a lot of them. If nothing else, I still love building this stuff. I've been at it for several years and my interest level hasn't diminished at all, so I have that going for me!

Registration rejects email addresses with certain characters ("+" at least)

Also, I think it's a bad idea to offer common financial codes as recovery codes for your site (SSN, mother's maiden name, etc.)

Thank you, DrScump. Both of these have been fixed. Users will be warned to change their security question if they're using an old, insecure question type, and no one will be able to select those questions beginning immediately. I was using a standard list of security questions, but I absolutely see the value of avoiding such info.

Invitation code? Not liking my email address for some reason.

Hey, if you're willing, could you shoot me an e-mail from the address it doesn't like? You can reach me at hello @ the domain of the site. That's the first I've heard of the issue, and I'd like to make sure it gets fixed. (And thanks!)

Invitation code?

Reply on news.ycombinator.com