4 ms·

Show HN: More Like This Bookmarklet

20 points by tomw1808 12y ago

waxjar 12y ago
I've noticed many product pages these days do some "scroll enhancements" through javascript. This page seems to do that as well. It's very annoying! The page doesn't really explain what the bookmarklet does, only how I install it. I'd like to know what it does first :)
- rashkov 12y ago
  You can click this bookmarklet to get suggestions on similar pages when you're reading an article that you arrived at from hacker news or reddit. Seems like this service uses solr to index the articles as well as their content. Would love to hear more about the indexing strategy from the author.
  - tomw1808 12y ago
    Hi, thanks for the question and sorry for the late feedback :) Yes, its a solr index behind. I did a lot of stuff with LibLinear [1] for my masters, in combination with Apache OpenNLP [2]. With a mix of clustering and a lot of fun with lucene (3.x at that time) and sentiment analysis. Now, some years later, I re-thought my pretty buggy approach from the past and wanted to do something with that. A small side product is this bookmarklet. The crawling is a pretty basic web-scraping approach in Java. Not in-code (like Jsoup), but with a Firefox in a Framebuffer with a Webdriver, because I wanted to be able to have the full power of a modern browser rendering engine, instead of a in-code content-download. [1] http://www.csie.ntu.edu.tw/~cjlin/liblinear/ http://www.csie.ntu.edu.tw/~cjlin/liblinear/ [2] https://opennlp.apache.org/ https://opennlp.apache.org/
- tomw1808 12y ago
  When you drag&drop this bookmarklet to your bookmarks, nothing happens. The magic happens afterwards when you are visiting a site outside of nagrgtr: Inside the bookmark is a little javascript script that reads the URL you are just visiting and sends it to: http://nagrgtr.com/morelikethis/URL http://nagrgtr.com/morelikethis/URL thats it. Have a look at the href behind the link, its really simple as that. Greetings from Austria edit: The page has no scroll enhancement to my knowledge. Where do you see that, or on which device do you look at the page so I can reconstruct the problem on my end? Thanks.
  - waxjar 12y ago
    It does have scroll enhancement. It scrolls faster and keeps scrolling for a while after I've stopped scrolling, "easing out". I think this line is the culprit: <script data-rocketsrc="/static/js/nicescroll.js" type="text/rocketscript"></script>
    - tomw1808 12y ago
      my bad, removed it. Thanks for pointing me to it.
tomw1808 12y ago
Hi all, that little service came out of a larger project of mine, but I thought it could be useful to everyone. Unlike other sites, where you can search for something, the index here is based on crawling all and everything that is posted on HN, a lot of Reddits and many many more sites. So not only based on a headline or something, the comparison is really based on content behind the links. Happy to hear your opinion. Wish you all a great new year and good luck in 2015! Tom
- dang 12y ago
  > So not only based on a headline or something, the comparison is really based on content behind the links Could you describe your approach in more detail?
- ratsmack 12y ago
  The following sentence on your web page has three "is" words. It should have two "is" words and one "it" I believe. Is it safe to use? Is is finished?
  - tomw1808 12y ago
    thanks, fixed. What did I learn? Do not make landingpages when it's late :) Thanks.
- 97-109-107 12y ago
  Interesting stuff, best of luck to you as well. I imagine this is not in the intentions of your service as it is now, but I'm wondering if you could share some knowledge on the matter given your expertise. If you had access to a sizeable amount of links that a group of people read what sort of bite-sized knowledge could one extract about the interests/problems/expertise of the said group? Do you believe such results would be meaningful at all?
  - tomw1808 12y ago
    Great questions! Well... I'm wondering that myself a little bit too. I'm thinking of what you can do by feeding this information to machine learning algorithms :) But you will definitely see more of that soon... To give you a few buzzwords: machine learning, clustering, nlp, sentiment analysis etc. in a scalable realtime infrastructure. Basically throwing all of that into a huge data-bowl, turning on the mixer to 11 and wait and see. Also thinking to make my PhD based on such a project, who knows :) I just know: it must be easy to use and I must like to use it, it has to solve my own problems, or else it ends like so many other projects out there - I guess you know what I mean. Imagine you had access to this information, what would you want to see? Which problem would it solve? Would you eventually pay for it? How often would you use it? To how many friends would you recommend it? etc... :) Best Tom
DotSauce 12y ago
Nice bookmarklet! I have one quick suggestion. The font size and color on the results page make it very hard to read. Please increase size and switch from grey and light blue to black and darker blue.
- tomw1808 12y ago
  Hi, yeah, thanks for the feedback, much appreciated, as I'm definitely no web-designer. I will do that together with a pagination and will roll that out hopefully next week. As the holidays are over and I have a normal day-job and a family next to it (or vice-versa), I try to keep working on the core as much as I can and keep the design as simple as possible. I am personally a friend of infinite page scrolling, but it seems that not everyone likes it. Any thoughts on that? Thanks.
  - 97-109-107 12y ago
    I'm usually againts inifite scrolling, but I think it won't hurt in the results page as it is something that you visit once, get what you want and leave.

20 points by tomw1808 12y ago

I've noticed many product pages these days do some "scroll enhancements" through javascript. This page seems to do that as well. It's very annoying! The page doesn't really explain what the bookmarklet does, only how I install it. I'd like to know what it does first :)

rashkov 12y ago
You can click this bookmarklet to get suggestions on similar pages when you're reading an article that you arrived at from hacker news or reddit. Seems like this service uses solr to index the articles as well as their content. Would love to hear more about the indexing strategy from the author.
- tomw1808 12y ago
  Hi, thanks for the question and sorry for the late feedback :) Yes, its a solr index behind. I did a lot of stuff with LibLinear [1] for my masters, in combination with Apache OpenNLP [2]. With a mix of clustering and a lot of fun with lucene (3.x at that time) and sentiment analysis. Now, some years later, I re-thought my pretty buggy approach from the past and wanted to do something with that. A small side product is this bookmarklet. The crawling is a pretty basic web-scraping approach in Java. Not in-code (like Jsoup), but with a Firefox in a Framebuffer with a Webdriver, because I wanted to be able to have the full power of a modern browser rendering engine, instead of a in-code content-download. [1] http://www.csie.ntu.edu.tw/~cjlin/liblinear/ http://www.csie.ntu.edu.tw/~cjlin/liblinear/ [2] https://opennlp.apache.org/ https://opennlp.apache.org/
tomw1808 12y ago
When you drag&drop this bookmarklet to your bookmarks, nothing happens. The magic happens afterwards when you are visiting a site outside of nagrgtr: Inside the bookmark is a little javascript script that reads the URL you are just visiting and sends it to: http://nagrgtr.com/morelikethis/URL http://nagrgtr.com/morelikethis/URL thats it. Have a look at the href behind the link, its really simple as that. Greetings from Austria edit: The page has no scroll enhancement to my knowledge. Where do you see that, or on which device do you look at the page so I can reconstruct the problem on my end? Thanks.
- waxjar 12y ago
  It does have scroll enhancement. It scrolls faster and keeps scrolling for a while after I've stopped scrolling, "easing out". I think this line is the culprit: <script data-rocketsrc="/static/js/nicescroll.js" type="text/rocketscript"></script>
  - tomw1808 12y ago
    my bad, removed it. Thanks for pointing me to it.

tomw1808 12y ago

Hi all, that little service came out of a larger project of mine, but I thought it could be useful to everyone. Unlike other sites, where you can search for something, the index here is based on crawling all and everything that is posted on HN, a lot of Reddits and many many more sites. So not only based on a headline or something, the comparison is really based on content behind the links. Happy to hear your opinion. Wish you all a great new year and good luck in 2015! Tom

dang 12y ago
> So not only based on a headline or something, the comparison is really based on content behind the links Could you describe your approach in more detail?
ratsmack 12y ago
The following sentence on your web page has three "is" words. It should have two "is" words and one "it" I believe. Is it safe to use? Is is finished?
- tomw1808 12y ago
  thanks, fixed. What did I learn? Do not make landingpages when it's late :) Thanks.
97-109-107 12y ago
Interesting stuff, best of luck to you as well. I imagine this is not in the intentions of your service as it is now, but I'm wondering if you could share some knowledge on the matter given your expertise. If you had access to a sizeable amount of links that a group of people read what sort of bite-sized knowledge could one extract about the interests/problems/expertise of the said group? Do you believe such results would be meaningful at all?
- tomw1808 12y ago
  Great questions! Well... I'm wondering that myself a little bit too. I'm thinking of what you can do by feeding this information to machine learning algorithms :) But you will definitely see more of that soon... To give you a few buzzwords: machine learning, clustering, nlp, sentiment analysis etc. in a scalable realtime infrastructure. Basically throwing all of that into a huge data-bowl, turning on the mixer to 11 and wait and see. Also thinking to make my PhD based on such a project, who knows :) I just know: it must be easy to use and I must like to use it, it has to solve my own problems, or else it ends like so many other projects out there - I guess you know what I mean. Imagine you had access to this information, what would you want to see? Which problem would it solve? Would you eventually pay for it? How often would you use it? To how many friends would you recommend it? etc... :) Best Tom

DotSauce 12y ago

Nice bookmarklet! I have one quick suggestion. The font size and color on the results page make it very hard to read. Please increase size and switch from grey and light blue to black and darker blue.

tomw1808 12y ago
Hi, yeah, thanks for the feedback, much appreciated, as I'm definitely no web-designer. I will do that together with a pagination and will roll that out hopefully next week. As the holidays are over and I have a normal day-job and a family next to it (or vice-versa), I try to keep working on the core as much as I can and keep the design as simple as possible. I am personally a friend of infinite page scrolling, but it seems that not everyone likes it. Any thoughts on that? Thanks.
- 97-109-107 12y ago
  I'm usually againts inifite scrolling, but I think it won't hurt in the results page as it is something that you visit once, get what you want and leave.