Good morning everybody and welcome to a new video nugget. Today we're starting a fully

new part of the course. It goes about communication, communication between humans and humans or

humans and agents. And as such, we will talk mostly about natural language. So if you think

about it, natural language was long considered as a skill that only humans have. That's no

longer strictly true. We know that some animals have language skills. We know that the higher

primates, for instance, gorillas, can actually learn languages. They can't pronounce human

language because they don't have the physical language track for that. But they can be taught

sign language and they've even been recorded once having learned sign language to teach

it to their offspring and others. We know that even some birds like crows have relatively

complex communication skills that not only have word-like tokens but also higher grammar

functions. So that's a language skill. And we're just basically beginning to understand

what whales and dolphins and so on are saying. So even though it's not something that's exclusively

human, language is a sign of high levels of intelligence. And if you think about it, it's

really amazing that we can express complex thoughts. For instance, the concept of a pom

dippy in a sentence, which is just basically a sequence of words and do that in a matter

of seconds and do that without ever breaking a sweat. And on the other side, on the side

of the hearers, we can decode a sequence of words into a complex thought, establishing

a line of communication. And finally, one thing I really find amazing is if you watch

children then in a matter of two, three, four years, essentially toddler age, a child learns

tens of thousands of words and relatively complex grammar relatively easily. If you

think about the grammar a child of four uses is essentially the same grammar it uses for

life. Later it learns more concepts, understands more concepts and learns more words in the

process. But in the grammar and complexity, we see almost all the complexity that language

has to offer. So it's reasonable to expect that natural language should be part of AI.

Of course, there's other things that play into the picture. We, for instance, have linguistics

that kind of look at the structure of language and communication. But it's a central part

of AI and that's what I want to cover in this part. So we assume that humans kind of learned

how to speak about 100,000 years ago and about 7,000 years ago we see the first written languages

starting to appear. Another thing is that Alan Turing based his namesake test on natural

language because it's one of the main communication forms that is needed and communication is

needed in the text. So we want AI agents to be able to communicate with humans. We also

want AI to be able to acquire knowledge from reading written documents. After all, that's

the main way humans actually accumulate knowledge. You're reading books, you're reading these

slides. So if we want AI agents to have access to this kind of knowledge, we need to teach

them how to read. So we're going to look at in this part of the AI2 lecture, we're going

to look at specific information seeking tasks that agents might have. You should be warned

that again, like always, we're just basically skimming the surface here with the intention

to show you how certain things work so that you get an insight of what research or the

methods in this area are like. We're not actually going to be in any form exhaustive. So we're

going to talk about language models. We're going to look at text classification, information

retrieval and information extraction, how to get at semantics problems. Again, just

touching the surface but showing you what salient methods are so that can go on from

here. There are full of course, lectures on this topic. There is a chair of corpus linguistics

and computational linguistics. All of those actually will give you, if you want to branch

out in that direction, a much deeper insight on things. So the first chapter here is on

what we call natural language processing, which is kind of the technological end of

things. I would like to briefly introduce that. So natural language processing is an

engineering field at the intersection of computer science on the one hand, AI and linguistics,

which is really concerned with the interaction between computers and natural human languages.

Just to fix some terminology, we have the input pipeline into the computer, which we