 So briefly, what is an LTK? Before I answer that question, I'll give you a small overview of what natural language understanding is, because giving you an overview at an LTK does without telling you what the problem is is like giving a solution without a problem. So briefly, natural language understanding is trying to parse plain unstructured text, basically the wind that's coming out of my mouth and reaching your ears, and makes sense out of it. It's a difficult problem. It's far from being solved. There are many steps to it, and there are somewhat two approaches to it. There's a more computational linguistic approach to it, which tries to understand how language works. And then there's all the people who actually just want to use the results of natural language understanding. And for example, parse Wikipedia and build databases of information which is structured and which can be used for reasoning, automated inferencing. One of these main users is DBpedia. Right now, they're limited to using content from the info boxes and from links, which is a very small part of what information is stored in Wikipedia. And they'd like to use natural language processing to gather information out of the text and make it directly into triples, which contain logical information. So how do we go about trying to understand text? Well, let's look at the question, what is an LTK? We can break it down into components. So very briefly, we have a word, which is what. It's a word that starts with wh, and many of those words in English, what, why, when. How is an exception, because it doesn't start with wh, but there are words that introduce questions. And then it's followed by a verb, is. And a noun phrase, an LTK, which is a named entity, typically, is what we call it. So it's a word that represents something specific. So how do we go about breaking text down into its components? Well, typically, we'll run a sort of pipeline. First, we split text into sentences. It seems easy. It's not as easy as you'd think. Trying to cut the sentences in terms of punctuation is generally how you do it. But then there are some words like US, for example, which end with a dot and can confuse a system. Once a text is split into sentences, you also need to split it into words or tokens. And then you can analyze those. The next step will depend a bit on the pipeline. But typically, we'll either involve determining what type of word it is. So breaking it down into tags, such as WP on this example. The what and then VBZ for verb nouns and so on. Another step is looking up named entities. So things like an LTK usually have some kind of a database. And you try and find words that match in there. There are other approaches which use machine learning instead. So they try and use features like, is this word capitalized? Is this a sequence of nouns, many in a row? And try and guess based on that. So I'll give you a very quick demo of a short NLTK approach to breaking part of this down. I'll be looking. So I'll be showing you briefly how to parse a sentence in terms of obtaining the parse tree. So first, import an LTK. Don't actually do that in real life. An LTK does an import star in itself. It makes looking at examples very easy. On the other hand, it's somewhat of a nasty trick. And then I'm using a very simple sentence where there's no need to break apart punctuation or anything. It's just a series of words that can be tokenized really quickly. And then loading a grammar. A grammar is basically just a definition of the language. And then I get a parse tree out of it. So you see the different parts, the different components of the sentence. And quickly, and then the named entity lookup. You have NLTK groups. Actually, I don't think you can see the whole thing. But basically, NLTK uses chunking. It finds a bunch of capitalized words and tags them as an organization, which natural language toolkit actually isn't. This is looking at the whole thing. But basically, it's looking at natural language toolkit and thinking it's an organization because it's not better informed. So quite briefly, here are some resources to learn more about NLTK and natural language processing. And do you have any questions?