All Microsoft Microsoft Certified: Azure AI Engineer Associate certification exam dumps, study guide, training courses are prepared by industry experts. Microsoft Microsoft Certified: Azure AI Engineer Associate certification practice test questions and answers, exam dumps, study guide and training courses help candidates to study and pass hassle-free!

Implement Natural Language Processing Solutions

1. Overview of Natural Language Processing Services

So we're into the next big section of this exam that says "implement natural language processing solutions." It's worth 20% to 25% of the overall score. Now, you may be familiar with the term "natural language processing." This is how a computer can understand the spoken or written language of humans. Now, in this case, we're going to be dealing in English. So we want to be able to speak to the computer in natural language, just like a human would to another human, and have the computer understand what we're saying. So we're going to break this out into text analytics in this section. There is a speech service in the next section.

And Azure can even do things like translation. Translation. Computer translation has been around for a while, but now that we can understand the intention, we can understand things like sentiment. We can basically translate stuff in a much better way to computers in 2021 and beyond than they previously could. Microsoft Azure has this Lewis language understanding service that we can use to add natural language to our own applications. We're going to create a model. We're going to use a model in our own code, so that's the end of this section. So the first section has to do with text analytics. As you would expect, processing text is actually easier than processing speech. When someone types a phrase, in this case in English, you have to try to understand what they're saying. And so if somebody says, "What is the weather going to be like tomorrow?" Then the computer is going to have to pick out the word "weather" as a key phrase and "tomorrow" as the key phrase.

And so in even complex paragraphs of text, if you can extract the words that actually mean something as opposed to the filler words, the adjectives, and the in-between words, then those are the ones you can ignore understanding in any particular text. Who are the people's names, which are names of places, and are there any kind of URLs or links in there? Sentiment is the concept of whether something is positive or negative. And so, if you can read a text and understand, oh, they're seeing some good things about this or they're saying some bad things about this, and even something as simple as understanding what language the written text is in, you'll be able to communicate effectively. So all of these things are effectively available within Microsoft Azure Cognitive Services. So in this section of the course, we're going to start to dive into text analytics, followed by speech.

2. Extract Key Phrases using Text Analytics

So let's go ahead and examine the code for processing key phrases. If you go to the GitHub repository for AI 102 files, under the text analytic subdirectory, there's a Python script called Key phrases. Now in this script, we're using the Azure Cognitive Services SDK. There is a language module and, under that, a text analytics module. And so we're going to import the text analytics client to do this key phrase analysis. Now, similar to passing scripts, you need your own particular endpoint and key from cognitive services. We can then use that to authenticate and create a text analytics client for our use.

Now, in this code, we are passing in four different sample texts in four different languages. So we can see here that it's not just English that this supports; it supports a number of languages. In fact, we can quickly switch over to the documentation and look at key phrase extraction to see which languages are supported. So there are a number of languages—it looks like at least two dozen languages here that are supported for key phrase extraction. So we have four samples. Japanese, German, English, and Spanish And very simply, we're just going to ask for the key phrases here, and then we're going to run the key phrase method against that text analytics client to retrieve that.

So we have a formatted object here that we're passing in, and then we can basically iterate over the responses that come back. So we've passed in all of the phrases at once, and it's going to process them all at once and then output that very simple example, simply showing you how easy it is to call the cognitive services. So let's switch over to PyCharm and see how that is going to run and what kind of output we're going to see from this text. So we switch to PyCharm, run the script, and it returns very quickly, and we can see that it returned the key phrases from the four phrases we passed in. Now, I don't speak Japanese, German, or Spanish. Maybe you do, but I only speak English. And we can see here that the key phrases are "cat" and "veterinary," which seems correct to me, so you can sort of see the basis of natural language processing within Azure and that the computer can understand what is relevant and what is not. And this can help us with translation and with our chat bots. In the rest of this section, we'll see future cognitive services that we can basically drive attention from, including sentiments.

3. Extract Entity Information using Text Analytics

In this video, we're going to use the TextAnalytics service to extract the entities contained within text. Now what I did was go to the Wikipedia article for Microsoft Azure and take the first paragraph from it. And we can see here that, even within Wikipedia, certain things are already clickable. And so these can be considered to be entities. And what I want to do is make Microsoft Azure recognise the entities within this text. So I started a new NetCore application. You can see the console. application for Hello World. What we're going to do is basically start building upon this by calling the Microsoft TextAnalytics API and getting it to recognise whatever objects and entities are within this text. So the first step will be to import the text analytics namespaces.

So I'm going to go up here to the top and add the text analytics namespace. Now we can see with the red squiggly line that the package itself is not yet added to the code. So I'm going to go into the NuGet package manager, and I'm going to go into Browse, and I'm going to look for Text, Azure AI, and text analytics. And I'm going to install the latest version, which is 5.1%. I do have to accept the requirements of any kind of licence agreement. So this is the licence acceptance. And so now we've got the package installed. Now using C, we're in B-flat here. We're going to have to create the Text Analytics client.

And so we're going to use the Cognitive Services key and endpoint to basically create ourselves a Text Analytics client. Now in the Azure Portal, I have created a Text Analytics SDK client. If we go in here, I'm going to copy a key and then copy the end point. We should be able to build the application. At this point, it doesn't do anything, but we don't have any errors. All right, so the first thing we're going to do in this particular case is look at identifying the entities. So instead of having this "hello world," I'm going to paste in some code. Now what it's going to do is call the Cognitive Services client, which is the Text Analytics client, and it's going to call the recognised entities. The method passing in this paragraph, which was taken right from Wikipedia, is going to return a collection of entities.

And as long as the number of entities is greater than zero, which it should be, it's just going to loop through every entity that has been returned and include the text and the category. So I'm going to run this in debug mode, and let's see how it goes. Here we go. It has taken this paragraph and extracted what looks like about two dozen entities. It recognised the organization, Microsoft, at least four times.

The skill of cloud computing, including building, testing, deploying, and managing all skills and infrastructure as a service (software as a service, platform as a service), is also tagged as a skill. So we can use this type of entity recognition if we want to do some clever linking, if we want to try to understand what the person is saying, classifying text into certain categories, creating an automated index type service, it wasn't that difficult. If we close this out, we can see that all we did was basically set up the client, and then we just called recognised entities with the text. Now, just recognising the entities is only one part of what Azure can do.

There's another thing called linking entities. Now, linking entities is essentially returning HTML links to various sources, such as Wikipedia for some of these. So we saw that already. As a result, I was able to identify Microsoft and Infrastructure as services. Now let's see if it can link those to the appropriate URL. So now we'll come up with the entity and the URL. We're using the recognised linked entities method. Now in this case, some of those were duplicates in the recognised entity's call. In this case, it's linking the Wikipedia links from Microsoft Azure, Microsoft software as a service, platform as a service, and infrastructure as a service. So if you find some value in attaching useful links to content, you can have that automatically done using the linked entities command within the Text Services API.

4. Extract Sentiment using Text Analytics

So in this video, we're going to talk about retrieving and processing sentiment from written text. Now, first off, what is sentiment? Well, "sentiment" is simply the concept of whether a piece of text is a positive statement or a negative statement. If you look at the reviews for a restaurant, for instance, typically they come along with a star rating, and you can tell from looking at the stars whether it's a one-star review or a five-star review.

But if you didn't have the star rating and you just had to read the text, could you determine whether the person enjoyed going to the restaurant or did not enjoy it? Well, on the surface, you might think it's fairly simple because they might use the word "great," and great is typically a positive sentiment. But what if they say, you know, this was not great? Well, they're using the positive word "great," but then they're qualifying it with "not." As a result, a sentiment can be difficult for computers to understand. You can't just recognise single words and say they're positive or negative. So let's look at how Microsoft Azure Cognitive Services processes sentiment from phrases. So we're going to go to GitHub, and under AI102 files: text analytics, there's a sentiment script.

Now, this script actually contains different samples, not only for sentiment, but we're looking at keyphrases again, as well as entities. But there is a sentiment analysis right here at the top. And so in this example, we aren't using the RestAPI again; we're not importing the SDK at all. We are using the endpoint, and we are simply building the URL to the sentiment call using the endpoint. Now, in this case, we're processing it in a couple of different languages. Once again, the API supports multiple languages for analysis. We can see that I enjoy the new Xbox One now that we are clearly speaking English. And so clearly, this is starting off as a positive review. We're expecting that to come back extremely positive. So sentiment returns a number from zero to one. So when we call the sentiment URL using a JSON post and get a response back, we're expecting this first one to have a really high score. The second one is in Spanish, which I am not great at, but I am going to recognise this as being a terrible day, um, dia terrible. So I'm expecting this to be a negative sentiment.

So, let's run this through Pie Charm to see if it returns a high score for the first one and a low score for the second. So we switch over to PYC, we run the script, and we can zoom in here. We have to scroll to the first example, and we can see here that the first statement, as we expected, is 100% positive, 0% neutral, and 0% negative, and the sentiment comes back as being positive. So that's not unexpected. We expected the first one to be positive. The second example in Spanish is 93% negative, 5% neutral, and 2% positive. There's a little, tiny bit less confidence in the negativity of it. It still comes back overall as negative. And so again, that's what we expected. So we can see here that Microsoft Azure CocktailServices can analyse text and understand more complicated examples of whether a statement is positive or negative. And it comes back with confidence scores. So there are some statements that might actually be neutral, and it's not going to assign positive or negative sentiment to them.

Microsoft Certified: Azure AI Engineer Associate certification practice test questions and answers, training course, study guide are uploaded in ETE files format by real users. Study and pass Microsoft Microsoft Certified: Azure AI Engineer Associate certification exam dumps & practice test questions and answers are the best available resource to help students pass at the first attempt.