Entity Linking Intelligence Service

Microsoft Cognitive Services is a set of APIs, SDKs, and services available to developers, exposing a suite of machine learning APIs that enable developers to easily add intelligent features- such as facial, speech and vision recognition, emotion and video detection, and speech and language understanding into their applications.

At last count, the Microsoft Cognitive Services suite of services include over 30 APIs, SDKs, and services typically grouped into one of six categories: Vision, Language, Search, Labs(Experimental services), Knowledge and Speech. 

Entity Linking Intelligence Service is a part of the Cognitive Service Suite and comes under the Knowledge category of the same.

So, on the surface, the concept of Entity Linking Intelligence may seem straightforward. This idea that a word may appear in the content and we simply identify it and then link to it. This would typically be a proper noun but it could really be any noun. And the process is simple. We are provided with some kind of content, news article, Twitter feed, a blog post. We submit that content to the service and the service returns the list of entities. But here's where the process gets tricky. Let's say we are reading an article that says Apple Corporation stock fell to historic lows. Apple would be an important entity or Apple Corporation might be an important entity in that sentence. But we then have a sentence that says- An apple a day keeps the doctor away. Obviously, the word apple means two different things in these two different contexts. 

Entity Linking Intelligence Service provides the concepts of recognition and disambiguation based on the context of the content.

Entity Linking Intelligence Terminologies:

1. Entities: These are simply human readable Wikipedia entry labels or titles. Every extraction returns a list of related entities, and every entity includes a list of matches. So, an entity might be Microsoft, or Chief Executive Officer or even AWS. 

2. Scores: A level of confidence ranging from 0.0 to 1.0.

3. Matches: A recognized entity extracted from the content containing a collection of entries.

4.  Offsets: Starting character position of a match collection entry. 

This can be used in any area which needs to have Artificial Intelligence.