So now we're going to invoke the pre-trained ML APIs from within AI Platform Notebooks. The first thing we need to do is ensure that the APIs are enabled. So we can do that by clicking the navigation menu, go into APIs and services and clicking library. I'll search the Vision API which we're using first. Click it. We see the API is enabled. We can do the same with the other APIs. In this case, we're using the Translate API, also enabled, the Speech API, Speech-to-Text API, and the Natural Language API. So those are all enabled. Now I need to get an API key so I'm going to go to APIs and services and click Credentials. Then I'm going to grab my API key. I'm going to click create credentials here and create a new API key. In this case, I'm going to go ahead and copy this API key. Then you want to close this, and I'm going to go ahead, and I'm going go back to AI Platform Notebooks. Now, I've already created a Notebook server here, so I'm just going to go ahead and open Jupiter lab. Next, what I'm going to do is I'm going to pull from GitHub a repository that we're going to use in this course. The way you do this within Jupiter lab is that you click the Git image, this Git picture here and if you do it, you do a mouse over itself and say git Clone. I'm going to type in the URL for this Lab and you should paste this from the Lab. This is GitHub/Google Cloud Platform/Training Data Analyst, and I'm going to click Clone. After a second what it should give you is that new repository is directory right here. So in this case I have training data analyst. I'm going to go into training data analyst, and then I'm going to navigate to where the slide is. So it's in CPB 100 lab4c, and then ML APIs. When I double-click that, I'm going to see this editable notebook that we can run right on the VM. The first thing we're going to do is run a change our API key and paste what we already pasted from GCP Console. And then I'm going to run. I can do this by pressing Shift and then Enter. I can also do this by pressing the Play button here and I'll run the cell as well. We're going to ensure that we have the Translate, Cloud Vision API, Natural Language API and Cloud Speech API enabled, which we've already done. Because we're calling the APIs from Python, we're going to install this. We're going to install the Python package. So the first thing we're going to do is we're going to use the Translate API. Have you used Google Translate before? This is the same thing but within API form. So what we're going to do is we're going to take a set of input sentences. It says, "Is it really this easy?" "Amazing technology." "Wow." Then translate from one language, from English into our target language, which is French. Then we're going to print out each of those translations. So when I run this, I have the first thing. Is it really this easy? I'm not going to read that because my French pronunciation, isn't very great, amazing technology translated in French, and then wow. Next, we're going to invoke the Vision API. The Vision API can work off an image from cloud storage. And it can extract the characters to the language in a sign or an image, using Optical Character Recognition. In this case, we have the sign that's in Chinese. I don't read Chinese, so I want to see what it says. So the first thing we can do is we can make an API call to the Vision API and get the text of this image. So when I run this, I'm going to get this large bunch of output. This is in JavaScript Object Notation in which I'll see each of these characters, where it's located and also what language it's detected here. In this case, it is detecting that the language is Chinese. We can extract that locale and description by running this code block, and when we do so we see that Zh, that is the language code for Chinese. Next, we can translate that by using the Translation Service, the source being in Chinese. So in this case, Zh is stored in this foreign text variable, and then the target language being English, and then our inputs, the inputs that we extracted using the Vision API. We run this, we get the source text and then the translation, which reads, "Please protect and protect health and create a beautiful water environment." Next, what we're going to do, is we're going to do sentiment analysis with the language API. So that's going to give us a sense of whether a particular quote or a particular piece of text is positive or negative. So in calling this API, we're taking each of these quotes, we're feeding them to the language API, and we're generating both polarity score and the magnitudes score. So when I run this, I get a polarity score, in this case, 1, for positive, and then a strength magnitude. So to succeed you must have tremendous perseverance, tremendous will. Very strong statement. Polarity here as negative 1 and the magnitude of 0.5. It's not that I'm so smart, it's just I stay problems longer and etc. You can play with this on your own to see how the sentiment analysis scores for a particular block of text. Lastly, we're going to invoke the Speech API. The Speech API can work on streaming data or an audio content, and, or it can work on a file on Cloud storage. So here we're going to pass an audio file from Cloud storage, and we're going to take it and try to transliterate that from the audio into text that we can read. So when we run this, on this file, we get a transcript here and a confidence score, the transcript being how old is the Brooklyn Bridge, with a pretty high confidence score, and we can reduce that down just to these values. So that's how you invoke the various ML APIs. Go ahead and play around with it. If you want some more of a challenge, go down to the challenge exercises, in which there's some portraits from the Metropolitan Museum of Art in New York, and you can use the Vision API trying to find which of these images to depict happy people and which of them depict unhappy people, and you can dig around to see what part of the response could use some of those values.
