Hi, this is me mccray with the third and last part of the taxonomy basics videos. Today we're gonna be talking about taxonomy for retrieval. We tend to think of taxonomies as something physical, something that we can see, we saw the taxonomies come in different shades and that it is possible to create combinations or mappings using faceted and hierarchical structures. But even when we see representations of hierarchies and the interface allow us to click and move back and forth in the tree or filter out our results using the facets, the system. See taxonomies in a very different way. Systems see hierarchies as vocabularies and not just any vocabulary, but the standard vocabularies that have been designed to model information in different levels.
This is the case of the simple knowledge organization system or OS, which is one of the vocabularies created by the W three C. The organization behind the semantic web standard vocabularies are a topic on its own. But here I want to highlight two things about them. One is that in addition of the semantic properties that allow us to declare which term might be our top level, which is not our Children and which ones are brother, we have options to include other metadata useful to manage and create a taxonomy. Moreover, as we see in the left vocabularies can be combined here we have for example, cost being combined with DC terms and this C terms specifying who is the author of the taxonomy when it was created and when it was modified, vocabularies are also important because they allow interoperability that is the ability to exchange information and create a communication process with users in these ways.
Taxonomies become the main language, a language that can be rich and that can be refined even more. When we take the perspective of the system, it is hard to understand what is the difference between a navigational and a vacant taxonomy. And it is hard to find a distinction because we don't do anything different in a taxonomy editor. When we do an educational taxonomy or a backend taxonomy, it is not the case that for one we have instruction and for the other one, we don't have it. But again, the way we implement the control vocabulary is what makes the difference and vitamins can add more or less restrictions. But let's take a closer look to some of the walls of backing taxonomies in the taxonomy domain. You'll hear about precision and recall. Precision refers to the ability of matching queries, seeing exa exactly what we're looking for when we are typing something Rle On the other hand, is more about providing a variety of results.
But for any of them, we need as prerequisites a consistent logic that extends the test of time and that captures nonpreferred terms or susceptive terms. That would be useful for both goals. We would like to have as well additional metadata to create those mappings and enhance the search experience. You can think of vacuum taxonomies as ingredients for multiple recipes where we are not interested on a dish only but have a diverse group of ingredients to combine. Let's see some examples. This is a broad query for milk that is returning 993 million results. On the top of our results, we have the Google info box and the link to Wikipedia to break down even more. This 993 million Google is using a couple of different techniques, query refinement with questions that may have been inserted by other users and popular rankings of results that are also displaying a map with some of the things that we have around that include the term milk.
When we scroll down, we have the taxonomy of products displayed in parts, top stories with recipes as well as a movie that is also related to the results. Not all of these results are the same. Obviously, we have a combination of recipes, products, info boxes, websites, restaurants, but all of them have presented drink milk, which is the main goal when we are doing rico precision, we experience in a very different way. Usually by typing the first letters and seeing that the result we are thinking is the first that is being shown, even without writing the whole thing, Google knows how to identify what we want because they include misspellings and synonyms and they add all of these things into their taxonomies and algorithms and they use them for retrieval backend. Taxonomies are implemented more for the goals of precision in Rico.
When we are displaying results, we want to provide variety and accuracy, but also serendipity those things that you wouldn't even expect. But that are nice to find. And this is obviously considering what could be an ambiguous query or a broad query or even a front query to provide the user with the things that he expects. And also the things that would be surprising for him as a way of summarizing this series. I would like to stress that control vocabularies organize information objects independently of what is our implementation, navigational or backend taxonomies.
Control vocabularies need to be distinguished the limit and defined to create taxonomies. As we have learned today, we need to apply standard vocabularies and scores or simple knowledge organization system is the vocabulary that we use to model taxonomies. And the sora the implementation of a taxonomy can influence the way we model it. And in some cases, we can be brought and in some others, there is not a possibility. I hope that you find this series useful and if you have any questions, don't hesitate to reach out.
Thank you so much.
