One major competitor (well known for anyone who's looked into this stuff) is Alchemy . I tried a New York Times link  on Aylien and Alchemy, and Alchemy performed much better -- in fact, Aylien didn't even successfully find the article body. I'm sure you guys will be iterating on improving the algorithms, but just wanted to flag that as a potential turnoff for anyone comparing your website demo with Alchemy.
Best of luck!
We don't rely on CoreNLP, or NLTK, we have our own sentence disambiguation, and our own part of speech tools. So we are a lot faster.
Our other api's let you piece together a lot of cool NLP projects with very little code.
It would be a nice to offer a library with a bootstrapped training set.
Machine Learning as a Service seems Hella Neat, tho.
It seems to contradict the paragraph before -- ML as a service seems a terrible idea for the reasons you just listed (among others). What's "Hella Neat" about that?
Servers are relatively fungible, given ops automation; it's painful but not the end of the world if you have to migrate away.
But the technology is still relatively immature in that building your own ML service in house - and having it scale, etc - is still a big pain.
I would immensely prefer it if we first brought ML libraries up to a higher level of maturity - as simple as apt-get install and adding `includes ActiveLearning::Bayes` to your models.
But if a client came to me tomorrow and said "there's this great Amazon API that we're thinking of using" I wouldn't consider that insane on first principles.
Is something broken? Maybe you could cache some recurring analyses.
Just tried a few links:
Am I missing something here? It seems like it's just parsing text, i'm not seeing any context(keywords, categories, summaries)
edit: It's giving fantastic results when pasting the raw text! :)
Are you guys using DBpedia? It's giving very similar results to a system I was working on in the past: http://www.zachvanness.com/nanobird_relevancy_engine.pdf
we do use DBPedia in our Concept Extraction. please have a look at the docs: http://aylien.com/text-api-doc
Sure thing(when running it on the urls, I don’t get any keywords: http://i.cubeupload.com/zubo4G.png
edit I see from another response that the server room is on meltdown, I'll wait for a bit.
we'd love to, but unfortunately some of our main competitors have restricting terms in their ToS (e.g. http://www.alchemyapi.com/company/terms.html) that prevent us from doing so. we will publish what we can though.
> Any information about what domains your training data is from?
they're mostly trained on general news and social media content (with lots of manual and automated cleanup). drop us an email if you need more details: firstname.lastname@example.org
Why can't you just run any of the standard NLP evaluations?
YOU MAY NOT ACCESS THE SERVICES FOR PURPOSES OF MONITORING THEIR AVAILABILITY, PERFORMANCE OR FUNCTIONALITY, OR FOR ANY OTHER BENCHMARKING OR COMPETITIVE PURPOSES
Suppose I am evaluating their service, before I decide to buy. I would be breaking these ToS, I guess.
What is special about your project ?
Edit: A quick glance at the API also shows that there doesn't appear to be much in the way of machine learning. Does this build models for you or is it just to dissect text?
This is a very interesting area... Good to see something new apart from Alchemy and opencalais !
Classification: arts, culture and entertainment - architecture .(WTF?)
Polarity: positive. (Nope)
Polarity confidence: 0.9994709276706056. (Well...)
Looks pretty rough to me.
There's no excuse for the polarity though. "Gone to shit" should be a pretty good indicator about the sentiment.
- classification is trained on longer texts (mainly news articles) so it won't perform well on shorter texts
- polarity: yeah that's bad - the sentiment is slightly ambiguous so maybe a lower confidence would've justified things
- corpuses: we update our indexes frequently + use higher quality / handpicked corpuses.
- features: our API provides Summarization and Hashtag Suggestion.
and future plans, obviously. hope that helps.