Thanks for the question, whitten! You raise a very interesting point.
We definitely plan to contribute information back to Wikidata and, in fact, we already did as part of our experiment. We contributed more than a thousand edits, mainly on company data like industry of operation. Of course, we need to strike a balance between what we contribute back and what we keep as competitive advantage.
Besides that, we always comply with licenses and give credit to the original source when this is required. For example, when displaying Wikipedia content there's always a prominent link to the page where the content has been extracted from.
Thanks a lot, freediver! This is Francesco, co-founder.
We are using Entity Linking to detect entities in text. The algorithm takes into account context to disambiguate words, i.e., in "Mercury is a chemical element with the symbol Hg and atomic number 80" the word Mercury should be linked to the chemical element rather than the planet. When documents are indexed this way, the user is able to perform queries for specific concepts rather than simple keywords.
We had a demo open until a week ago, but we decided to close it to focus on finishing up some infrastructure work that will help us handle more efficiently maintenance of data sets. On the other hand we didn't want to drag it for too long and decided to just launch :)
We'll re-open to the public once we do that. If you're curious I can share a login for you to try it out.
> Techstars contributes $20,000, which is commonly used as a stipend to support living expenses during the program, and in return receives 6% common stock from each company.
I was considering Techstars, but the deal looks kind of unappealing, especially compared to YC.
I mean, yes, they give you access to their network and "300k$ of cash equivalent hosting, accounting and, legal support", but I suppose you also get that at YC plus 120k more in cash.
It's just one of the examples that came to my mind and that I thought was good. From what I know people working in pharma routinely have to search thousands of documents (e.g., research articles) looking for information.
Well, you can obtain these results with Google, but not really in most scenarios. For example, if you search "cities in Palestine with more than 1000 inhabitants" you will get lists of cities in Palestine and things like that, but not that New York Times article talking about Hebron (and not even a complete list of cities that match your criteria). This happens with lots of queries, where in the best case scenario Google will just return you a list of items (and not actual content talking about those items) and in the worst case scenario you will just get random results because it failed to parse the query.
It is an enterprise search engine which departs from traditional keyword-based search in order to provide an easier way to run complex, semantic queries on huge collections of text documents.
Why it's cool: image a pharmaceutical research task, where you need to find all documents mentioning drugs that interact with a specific class of diseases. In a normal setting, you would need to first research which drugs satisfy your condition and then either build a boolean OR query or probably query them one by one. Doqume saves you this hassle, because it allows to express conditions like "drugs that interact with infectious disease" with a simple user interface. As a result, you can get both the items that match your conditions (i.e., in the example, all the drugs that we know interact with the class of diseases that you specified) and the documents that match the query (e.g., recent research articles mentioning those items). The approach is not specific to pharma and you can easily build queries that span across several domains (e.g., "cities with more than 1M inhabitants", "USA companies with more than X employees", "singers who are born in Chicago", etc...).
If you want to give it a try you can see a demo with this query building capability at http://doqume.com/search.html
This also happens very frequently when I'm connected through a VPN. Must be that the IP is blacklisted or something, but basically there's no way to go past the captcha because it will keep showing images to validate forever.
I see many people here complaining about this feature, and while I completely agree with them I also find it difficult to imagine an effective way to clearly explain these concerns to a random average user with limited to no knowledge of computers. How would you go in explaining this stuff if you wanted to convince somebody to switch to another browser?
Correct me if I'm wrong, but I don't think that's relevant: if you open your service to EU users you have to comply to GDPR as far as their data is concerned. Am I missing something here?
Doqume (http://doqume.com)
It's a way to search text documents by specifying complex semantic conditions on items that have to be matched instead of simple keywords.
I think it's interesting because 1) it is challenging and it involves a lot of NLP (which I like) 2) I believe it's novel or at least I have not seen the same concept around.
Whether or not it is actually useful to somebody remains to be seen...
Can you use WolframAlpha to search news articles mentioning "members of al-qaeda" or scientific papers mentioning any "drug for the treatment of schizophrenia"?
Sure, the company is Babelscape (http://babelscape.com). For the translation tasks you can find massive parallel dataset with several language pairs at http://opus.nlpl.eu/, the other two things that you mentioned are not really in my area of expertise so nothing comes to my mind at the moment.
Please consider putting some code snippets in the home page, I needed I don't know how many clicks to find a random code snippet to look at in the docs.
We're looking for feedback on this new project that we managed to develop in a three-day rush.
The concept is simple: users can compete in our daily leaderboards of arcade games (as of now we just have one but we plan on scaling up) by paying a small entry fee; the top 10% of players is paid everyday at 00:00 GMT and the leaderboards reset.
We're just supporting XRP because of fast transactions and low fees but we plan on scaling this aspect too.
We put the login with Facebook as the only option because we use the information about movies the user liked to provide better customized results. We realized that this was potentially going to lose us some users, and we kind of put it here on HN as a test, to see if somebody was going to complain.
Anyway, we are going to have a "normal" registration/login feature very soon and both options are going to be available.
We definitely plan to contribute information back to Wikidata and, in fact, we already did as part of our experiment. We contributed more than a thousand edits, mainly on company data like industry of operation. Of course, we need to strike a balance between what we contribute back and what we keep as competitive advantage.
Besides that, we always comply with licenses and give credit to the original source when this is required. For example, when displaying Wikipedia content there's always a prominent link to the page where the content has been extracted from.