Our Technology
What we do
Seeker Solutions unites two technologies: Natural Language Processing (NLP) and Machine Learning. We apply NLP technology to unstructured textual data and provide computers with the reasoning skills to use that data in a meaningful way that reduces the need for human involvement, easing the human workload. This means you can take repetitive and boring tasks ranging from spellchecking 10,000 pages to document classification for a library and delegate these tasks to a computer, saving people time and effort and making life easier. Computers are meant to assist humans with their day-to-day lives, and what better place to start than to understand and be able to use human communication? However, to do so, we need to have a greater understanding of all aspects of human communication, which is no simple task.
Relationships between words and documents
As an example, what are document relationships? If you want to look up published articles related to the one you’re reading, how is that relationship determined? That’s one thing we’re working on defining. Anything can be related to anything else, and it’s not always as obvious as “Sam has a girlfriend.” Words can be related without sharing etymology or even meaning. A stove is inherently different from a plate, but both are related by being found in a kitchen. These are connections that a computer cannot draw, but for a human, these connections come naturally. Our work revolves around inferring conceptual relationships: finding relationships between documents that are not explicit and obvious, and then showing a computer how to find and interpret them.
The Search for Meaning
When looking at making a sentence, there are two things to consider: syntax and semantics. First, syntax gives you the structure of the sentence; the framework you should fit for it to make sense. A system with semantic understanding knows parts of speech and even roles that words can play in a sentence - so in “Sam has a car”, we know that “Sam” and “car” are nouns, and “has” is the verb, and even that “Sam” is the possessor and “a car” is the possessed. We can consistently get good results when analyzing a sentence’s syntax; but that’s not enough.
Once you understand the structure, you need to deal with semantics: what the sentence means. This refers to anything that can’t be counted, qualified, or often even seen, but that still adds understanding. Innuendos, puns, sarcasm and implications all fall into this category. This sort of reading between the lines is something most people do without having to identify and process but it is nearly impossible to get a computer to understand what makes a comeback witty.
We can help sort, manage, and access a huge base of information - but not the way a computer would. Our goal is to make the information intuitively accessible for a human.
Getting to answers
The system gains knowledge and ‘understanding’ by building relationships and meaning – this means we can put all the information together to make it easier for users to access. Also, because we’re looking at semantic meaning, we get additional insights that would otherwise get overlooked. For example, by using techniques like sentiment analysis, we can determine personal bias, author’s point of view, or overarching positioning of an issue. This is providing information that wasn’t explicitly given – and more, relevant information is always better when trying to make informed decisions.