Let's take a closer look at the actual APIs Google offers to make sense of your unstructured data. We know that in a typical business scenario a company ends up using data that is being generated from multiple sources. Typical examples of widely used data are RDBMS inventory, SAp and Spreadsheets. These data sources use strict data formatting rules and host data accordingly. Hence, the data is known as structure data. Apart from these a lot of data is donated from variety of other sources in the form of email, audio, video, images, text, social media likes and comments. All these types of data are typically free from strict formatting has known as unstructured data. As a business a prominent and important question is how to process unstructured data, which typically constitutes 90% of entire business data? This type of data has tremendous potential to provide detailed insights, which can benefit the business. In this section we will see which AI technologies can be used to process this unstructured data to have business impact. In this module, we are going to concentrate on the cloud natural language API for processing unstructured data in the form of text. However, keep in mind that there are equivalent APIs for image, video and audio data. Again, when we say that we are enriching unstructured data, we are saying that we are applying labels to it. We are providing labels for questions like what is the subject of this email? And does this comment have positive or negative sentiment? The NLP API provides many features with which text analytics can be performed. The first feature is Syntactic Analysis. Syntactic Analysis first breaks up text into a series of tokens, which are generally words and sentences, and provides information about the tokens internal structure and its role in the sentence. It can label a token as a noun or a verb, singular or plural, first person or second person, masculine, feminine or neutral. And provides grammatical information such as case, tense, mood and voice. Here is an example of the entity and this is output for a small piece of text. By the way, you are free to try this out of navigating to the cloud NLP API homepage yourself and inputting your own text. You can see how Lawrence of Arabia is correctly tagged as a proper noun. In production the NLP API's output is returned as a JSON object. At the time of this writing the API supports ten languages. See the outline documentation for current language support. The Cloud NLP API also offers entity analysis for recognizing people, locations, organizations, events, artwork, consumer products, phone numbers, addresses, dates and numbers. For our text, notice how TE Lawrence is tagged as a person and Lawrence of Arabia is tagged as a work of art. The second feature provided by the NLP API is Sentiment Analysis. Sentiment Analysis identifies the emotional opinion of our writers attitude. It's presented as a numerical score and a magnitude for the intensity of the feeling. It does not identify specific emotions, but groups them into generally positive, negative or neutral. For example, sad and angry are both negative, while funny and happy are considered positive. Note the sentiment analysis output for our short piece of text here. The score of a documents sentiment indicates the overall emotion of the document. Our score of 0.8 indicates a very positive document as expected. The magnitude of a document sentiment is proportional to the length of the document and tells us how much emotional content is present within the document. Entity sentiment analysis combines both entity analysis and sentiment analysis, and attempts to determine the sentiment positive or negative expressed about entities within the text. Entity sentiment is represented by a numerical score and magnitude and is determined for each mention of the entity. Those scores are then aggregated into an overall sentiment score and magnitude for an entity. In our example the sentiment score for pizza of 0.9 is very high, as expected. Finally, the NLP API classifies what your text is all about. At the time of this writing content is classified into 620 categories. In our Lawrence of Arabia example, the API correctly predicts that the text is about a movie with 91% confidence.
