The OpenAmplify Web Service is the first and only of its kind. It exposes, via an open API, 250 man-years of development effort in a web service based upon more than a dozen granted patents. OpenAmplify simply does a better job of surfacing the meaning of web content, at massive scale and speed. Here's an overview of the thinking and technology that makes it possible.

The Real Deal in NLP

Natural Language Processing, as a term, is used very loosely in the marketplace. At OpenAmplify, we have always taken a rigorous and academically-led approach to NLP. The objective is to permit a computer to understand human language with as much accuracy and utility as possible. So, what's the best way to do that?

It's not about sorting words by frequency. It's not about looking at what word sits next to another word. In short, it's not about overcoming the fact that a computer doesn't understand language. Rather, it's about making the computer as literate as possible. And that's hard to do.

Learning Like a Human

When humans learn their first language, the process looks something like this:

1. First, some basic vocabulary is absorbed, along with the implications and value of using that vocabulary. "Mama" attracts mother more directly than just crying; "eat", a little later on, produces a meal faster. This simple reward mechanism starts the ball rolling.
2. As the vocabulary and experiential reinforcement proceeds, more complex combinations of words come into play. The beginnings of a grammar form. Words begin to have different so-called syntactic roles in that grammar: the instinctive concepts of subject (I do something) and object (something gets done to me) develop.
3. Over time, vocabulary and grammar develop to the point where the child competently interacts in his or her first language, although subtleties (and their implications) remain a challenge. Humor, irony, sarcasm, obfuscation, etc., all develop over time. Context is hugely important to such subtleties (which, by the way, makes them hard for computers to handle).
4. These skills are refined by taking note of, and selectively incorporating, things that are heard in the speaker's environment. We learn by observation and judgment.
5. Those who master all of the above emerge as those we most consider to be "well spoken". 

If we aspire to bring NLP to its highest level of accomplishment, we do well to bear the aforementioned processes in mind, and to try to mimic them in computer code. Those who cannot do so find lesser methods and call them "NLP". At OpenAmplify, we have successfully pursued the very path outlined above, and that's why our NLP is the real deal.

Why OpenAmplify is Better

It is precisely because OpenAmplify truly mimics human understanding of language that it is better positioned to deliver what the marketplace needs: a real understanding of web content.

OpenAmplify is, first and foremost, based upon correctly handling number (2) above. Vocabulary (1) is largely clerical: compile lists of words and somehow "memorize" them, as a human would. But understanding the syntactic roles, the subjects, objects and relationships between words, is the real key, the necessary foundation for discerning the meaning of the text.

OpenAmplify's text analysis core detects syntactic roles with a very high level of accuracy, using techniques that are patent protected. The foundation of sentiment analysis, detection of intentions and all other components of "meaning" rests upon accurate identification of syntactic roles. And that's why OpenAmplify is better. The corollary is also true: without syntactic roles, whatever you build will be inferior.

Unique Decomposition of Meaning

Building upon a superior foundation, OpenAmplify then applies a unique view of what "meaning" means to deliver value to the customer. For computers to use the meaning data, the data must be provided with a definition and structure that makes sense and is useful. OpenAmplify's inventory of "signals" is a completely unique representation of "meaning".

The concept of presenting the meaning of a web text in terms of things like "Guidance" (is the author asking for advice? or offering advice?) and "Decisiveness" (how certain is the author about taking a given action?) is completely ours. It gives you the meaning of the text in a way you can immediately understand and use. We asked ourselves, "What is it that people really need to know about the meaning of a text?" And then we answered it. Nobody delivers meaning like OpenAmplify, because nobody has thought it through in the way we have.

That's why we can attack the types of nuanced, subtle aspects of understanding web content, as discussed in (3) above. We're working on sarcasm, humor and many other advanced signals right now. Expect important announcements on a regular basis.

Designed for Scale

Accessing the meaning of web content is less than useful if it cannot be done at scale. Billions of daily tweets, many millions of new web pages and blogs...any service that claims to add value must be able to handle volume, quickly and dependably.

1. OpenAmplify has been architected as a homogeneous web service that runs on the Amazon EC2 compute cloud.
2. It is able to scale itself up and down according to the load it encounters. There is no practical limit to the amount of traffic the service can handle.
3. It has no single point of failure. All requests can failover to other machines without intervention or loss.
4. All database functions are replicated.
5. Large requests are divided across numerous machines to optimize response time.
6. Recent requests are cached to speed up repeated requests for the same content.

In short, OpenAmplify is designed from the bottom up to scale massively. That same architecture ensures service dependability, too. OpenAmplify customers expect the service to be up, and fast, 24/7 and regardless of load. That's how we've implemented it.

Getting it Out There

Our mission is to surface every last shred of meaning from content. We won't get there overnight, though. We'll always be chasing that extra one percent of accuracy, that millisecond of response time.

The perfect is not, however, the enemy of the good at OpenAmplify. We have, over the course of ten years, always taken a "get it out there" approach to delivering value. As soon as we're sure something new is good enough and stable enough to help you, we release it. OpenAmplify is now in its fourth major release, with new functionality delivered at least once every quarter.

Open, Documented and Available

OpenAmplify is accessed via an open API, with solid backward compatability, numerous legacy versions still supported, and frequent releases offered with plenty of advance notice. The service is fully documented on the community web site, with full descriptions of all formats, API parameters, code snippets, interpretation guidelines, etc. The API itself is wonderfully easy to implement.

Use of OpenAmplify is free to all at development/research levels of usage. There are no restrictions on how you use the service, as long as it's legal and ethical. Joining the community automatically provides you with a private API key, which is all you need to use OpenAmplify. Should your needs exceed the daily free transactional allowance, you may contact us for very reasonable commercial terms or, in certain pro bono cases, extra free transactional allowances for further research.

Try the Real Deal

We invite you to get an API key and discover the power of OpenAmplify. Technical support is available through the community, and we'll work with you to make your application succeed.

The best semantic analysis of web content, delivered quickly and dependably at any scale: that's OpenAmplify.