Mira el vídeo gratis de Accuracy, Semantic Understanding, & Intelligent

Gain an understanding of semantic understanding and learn how intelligent assistants will increase their accuracy over the next few years.
TRANSCRIPT:
The other area where intelligent assistants are going to get much more accurate, is semantic understanding. When you say semantic understanding what that refers to is, having a computer system be able to understand what the meaning of a given sentence is, or the meaning of specific words in that sentence mean. And today how that's primarily done is by relying on something called a knowledge graph. What the knowledge graph is, is essentially a database of all of the known, or existing terms, concepts, or entities that are in the system. So a knowledge graph will typically contain names of people, names of places, names of known concepts, names of companies, names of songs, etc. Not only does it have the names of these concepts, but it has information, or metadata, about them that indicate what other information it might be connected to. So for example, there might be a knowledge graph entry for Albert Einstein, and that entry would allow you to infer that Albert Einstein is a scientist, Albert Einstein was a professor at Princeton, he was born in Germany, he was a physicist, he won the Nobel Prize, etc.
Right now the largest knowledge graphs that exist are on the order of hundreds of millions of concepts. Probably the really good concepts or nodes in those knowledge graphs, there's far fewer than that. But what's going to happen over the coming years, is those knowledge graphs are going to grow. They're going to get bigger, and as they get bigger, these intelligent assistants are going to get much smarter. Specifically, what's going to happen is these knowledge graphs are going to get much more personalized to your own personal knowledge graphs. So for example, if you're a physician and you say the term, spinal meningitis, that means something very specific to you. That's a disease, that's probably related to other diseases. Right now, my guess is that there's not very many knowledge graphs that have good information about that. But you can imagine in five years from now, every medical condition, every medical term, every medication, etc., will be part of a knowledge graph. If you're a physician and using one of these intelligent systems, it will be able to understand what you're talking about.
So that's, I think, the way that these knowledge graphs are going to get better, which is going to drive better accuracy in the semantic understanding that these intelligent systems will be able to have. The combination of having better speech recognition through better language models and acoustic models, plus better semantic understanding through broader, more personalized knowledge graphs, is ultimately what's going to make intelligent systems a few years from now, much much more accurate and much better able to understand what you're talking about.