The next step in data abstraction is to identify what we call Attribute Semantics. What does Semantics means? Well, it means, basically, the meaning of some attributes. So there are a number of predefined Semantics that are useful, very useful to identify. The first one is Spatial and Temporal Semantics. So, what do we mean by Spatial and Temporal? Well, it means that it's very important to identify when an attribute describes some spatial characteristics, often geographical characteristics, and Temporal, whether it describes something related to time. Let me give you more examples coming from the same dataset that we used in the previous video. Again, in the sales dataset, we can identify one attribute called region, which is the region of where this product has been sent. So, there are many different regions, in this case we have, for instance, Atlantic, West, Ontario, and so on. So every single item here represents Category, right? So that's a categorical attribute. But other than being Categorical, it's also Spatial because it describes a spatial location. That's very important. Here is another example of an attribute that is actually a pair of attributes that are actually Quantitative, but they also describe something Spatial. So we have Latitude and Longitude of different States, to identify the location of the States typically on a map. So Latitude and Longitude are quantities, they are numbers, as you can see from this image, but they describe something Spatial. As a side note, in this example, you can already see that for a given attribute semantics, you can have attributes of different types, in this case, we have one example of Quantitative, and one example of Categorical, and both have Spatial semantics. Now, let's move on to Temporal. So, in the same dataset, we have one attribute called Order Date, and Order Date is, again, Quantitative, but it describes something related to time, it's Temporal, okay? Another important characteristic for attribute types is also to identify these three specific cases, whether the attribute is Sequential, Diverging, or Cyclic. Let me show you what I mean by Diverging and Cyclic. So Diverging means that for a given quantity, it is possible to identify a zero value and above this value, all the elements are positive, and below this value, all the elements are negative. In this case, we have a Quantitative attribute that is also Diverging. Diverging because we can identify a middle value, and we can identify values that go up and values that go down. As we will see in a moment, that's very important to figure out what is the best way to visualize this type of information. The last one is Cyclic. What does Cyclic mean? So let me give you an example, again, from the same dataset. If, from the date, the Order Date, we extract information about in which month the order has been collected. We have 12 months, okay? But these data covers a span of multiple years. So, a given data point can fall into, actually every data point falls into one month, but we can have multiple years covered by the same month. So these data is actually Cyclic because we go as we progress over time. If you are looking at the information of which month the order is, this information is arranged cyclically across the months. One last useful characteristic to identify is whether an attribute is Hierarchical. What does Hierarchical mean? Let me give you an example. Again, from the same dataset, we have information about Product Category, so the product that has been ordered, what kind of category of product it falls in. But we also have sub-categories. A sub-category is a Sub-Category of a primary category. When you put these two attributes together, you can actually create a hierarchy. For every category, you have a number of sub-categories. Then the next category, a number of sub-categories, and so on. So this structure is Hierarchical. So let me give you a summary of all the elements that I presented that belong to the general activity of performing data abstraction. So the first one we have seen is identifying Dataset Type. And we introduced two main types, Tables, and Networks with Trees. The second concept we introduced is Attribute Type. We saw that we have three main Attribute Types, Categorical, Ordinal, and Quantitative. Then we introduced a number of additional attribute characteristics that can be specified on top of Attribute Types and are important to understand their meaning. So the first pair that I've shown is Spatial and Temporal. Then I've been talking about Sequential, Diverging, and Cyclic, so the order. The last one is Hierarchical, whether it's possible to organize this information in a hierarchical fashion.
