A feature ive been missing on hackernews, that perhaps you'd be willing to add, is a community written tldr for each link.
I.e apart from title and link, a short (200chars or so) description anf tldr.
51–60 of 69 posts
A feature ive been missing on hackernews, that perhaps you'd be willing to add, is a community written tldr for each link.
I.e apart from title and link, a short (200chars or so) description anf tldr.
Two bug reports:
1. Do you have a way to receive bug reports other than HN? :)
2. After doing a search, the left category menu disappears, and stays disappeared even after clicking the HH "home" link at top left. This is true for FF and Chromium latest-ish on LinuxMint.
There are two possible bugs here: a) do you actually want the left menu to disappear, and b) what your intent is for clicking the top left "HH".
- Go to HH.
- Search for something. Results appear as you type, nice. No indication from browser that a new page is loading; guessing no load by design. But left menu disappears.
- Manually erase search bar. Menu back.
- Type out a search again, menu disappears.
- Click "HH" at top left. Browser indicates a page is loading", but the search is not erased and (therefore?) the menu is still missing.
- Re-enter HH either by typing the URL into the browser location and clicking "make it so", or by clicking in from another site (like HN). Search field is empty, therefore the left menu is available.
EDIT: This was going to be a separate bug, but I think it's related to above.
Scrollbar behavior is buggy.
- Clear site cookies. ("It's the only way to be sure.")
- Don't click anything, just move the mouse around and scroll, with mousewheel or dragging scrollbar. Scrollbar intact, entire page scrolls.
- Click in search field, don't type anything. Scrollbar disappears, mousewheel scrolling has no effect, regardless of where the mouse hovers. Entire page jumps slightly to right, appearing to "chase" the disappeared scrollbar.
- Type something in search field that gets results. Scrollbar returns, top of scrollbar is even with bottom of search field, page does not jump back; I'm guessing this is "your" scrollbar rather than the browser's scrollbar. Mousewheel only has effect if mouse is hovered below the search field, in the area region where the scrollbar exists.
- Click on any non-active area outside the search field. Search field jumps left very slightly. Scrollbar is back to full length (browser's scrollbar?), but there are now two separate scrolling areas:
- Hover mouse at or above search field level. The entire original front page, including the missing menu and the default "Today" list of sites, scrolls up into the area from viewport top to bottom level of search field (which also scrolls up and away with the rest of the page). Search results do not scroll.
- Hover mouse below the search field, mousewheel scrolls the search results, phantom page at top of viewport does not scroll.
- Drag the scrollbar, the "top" scroll area scrolls.
Looks great. Can you tell us how you built it? I'm most interested in automatic categorization of submitted articles.
The vector of an article can be obtained by summing the vectors of its words (minus stop words). For a topic you just sum up 5-10 of the topic keywords. You don't need to exhaustively list all the topic keywords because word2vec automatically maps them in close vicinity.
This system has the advantage that you don't need a training dataset. It's unsupervised learning coupled with a small amount of supervised topic pointers.
Earlier quoted context omitted.
Yes, we're analyzing the titles. They're vectorized and fed into a LSTM net which was trained on a manually tagged training set. The set is not very big yet but yields good enough results for the initial proof of concept.
How did you go about selecting a sample for manual tagging, and how did you decide what tags to end up using?
We initially tried training the classifier only with GitHub based samples and using the user-given tags from there. Although we grouped the tag base into a reasonable number of distinct categories, the way how GitHub users tag their projects turned out to be just too inconsistent and often unrelated to the titles, so manual tagging was seen as a better option for getting decent results fast enough.
If you have any more specific questions feel free to drop me a mail to arturs@finch.io
Earlier quoted context omitted.
Actually, you can only be fined by the EU if you have a dominant position in the market you are using to promote your service. The goal is to avoid monopolies jumping from one market to the other (which is particularly easy to do in tech). "Good" monopolistic companies may benefit the customer in the short term, but proper competition benefits everyone in the long term. I know it was just a joke, but it really doesn'…
I know that and i'm happy that there are still entities in world that can still influence google and other mega corps.
I'm now feeling the need to properly contextualize hyperbolic jokes, so that people reading don't take it literally.
How are the topics assigned? Two Show HN posts I have posted recently (that are Open-Source and Javascript) don't have the relevant tags attached and I can't see a way of attaching them.
The tags are added fully automatic and there is no option to add tags manually for now. That being said, we really want to make the tagging algorithm better, so hit me at human@finch.io with your case.