The inference problem needs a triple solution

Posted on: September 03, 2015, by: Ann Kelly

Back to News and Blog

A triple is a relationship construct that consists of 3 parts; a subject, a predicate and an object. In semantics, triples are an atomic form of intelligence that can be used to describe basically everything in the world around us. In our everyday world we think of subjects and objects as nouns and predicates as a way to express the relationship between subjects and objects.

For example:

Bob (subject) lives in (predicate) Minneapolis (object)

In this fact Bob is the subject, Minneapolis is the object and “lives in” is the predicate.

You could also say:

Minneapolis (subject) is a (predicate) city (object)

Here Minneapolis is the subject, city is the object and “is a” – is the predicate.

To take it one step further you could say:

Minneapolis (subject) is located in (predicate) Minnesota (object)

Here Minneapolis is the subject, Minnesota is the object and “is located in” is the predicate this fact identifies a geographic relationship between Minneapolis and Minnesota.

One of the problems in computing is that machines can’t draw inference in the way humans do; in other words take facts, examine the information within the facts and then draw conclusions. For example, from the individual facts above we can infer that Bob lives in a city named Minneapolis which is located in Minnesota. While this fact is not explicitly stated, we draw the conclusion by connecting the dots (in our head) between the 3 facts to discover new information; something a computer can’t manage on its own.

So why all this talk about inference and what do triples have to do with it?

Up until now drawing inference was thought of as a human process. In the real world humans engage in inference every day; when they read documents, emails, news reports or other items. Using mental processes they’re able to reach conclusions based on specific evidence. Now with semantics and advanced technology like Smartlogic’s Semaphore, computers can determine facts and draw inference too.

So you might be wondering where these all important triples come from. That’s where Semaphore comes in.

Semaphore’s semantic platform lets you extract more value from your content. Semaphore Ontology Manager lets you create and manage your own proprietary ontology as well as leverage linked open data to import public domain specific vocabularies. Once a semantic model is in place, auto classification occurs. Here rulesets are published from the model and combined with sophisticated Natural Language processing and entity and fact extraction to result in a set of precise and consistent metadata tags or you can extract “triples” for each information asset – that’s where the triples come from.

By, expressing information within documents as triples, organizations can automate records management based on document content, leverage gigabytes of scientific research to discover new drugs, improve their drug development pipeline and analyze thousands of contracts in order to determine areas of risk.

Triples are the solution to the inference problem.