 Hello, my name is Zala. In the previous video, we have loaded a collection of text documents into orange. The documents were proposals to the Slovenian government. Now, I will show you how to cluster these documents. We start by importing the documents from the server with the import documents widget. We simply copy-paste the URL to the URL line. We then use the corpus widget to set the title variable for the documents. We also declare that the text we would like to analyze will consist both of the title and the content of the proposal. We continue with text preprocessing. Like in our previous video, we turn the text into lowercase, split it into words, normalize the words by using only their lemmas and exclude stop words and numbers. Now for the magic. We will use document embedder. The widget sends our pre-processed corpus to the pre-trained deep network and profiles each document with its characteristic vector. With document embedding, we first embed the words. We then represent the document with the mean vector of word embeddings. While not very informative, it is still instructive to see these embeddings in the data table. Embedder profiles the documents with vectors of size 300, that is, with 300 columns. We can now compare these vectors to find similar documents. We would like to visualize the documents in the map so that the documents that are close to each other have similar embeddings and are, hopefully, semantically related. The dimensionality reduction from 300 to two dimensions is called tisny. Here is a tisny map of our documents. We can select the documents from the tisny map and explore their content in the word cloud. For example, on the top left is the cluster related to traffic. And here, on the top right, are the documents related to property tax and income. The documents at the bottom of our semantic map deal with working hours, jobs and student employment. Nice. Notice that we can also retrieve the documents we selected in the tisny map in the corpus viewer. Here are the documents that deal with work and workers' rights. We can construct document maps with interesting semantic content using only a few widgets. Just like for documents, we can also find clusters of words. I will show you how to do this in my next video.