 Social media are full of interesting data on human behavior and sentiment is one such thing. Sentiment analysis gives us a quick glance into emotions in any kind of text and can be used for brand monitoring, review analysis, observing story arcs and for a commander systems. In this video, we'll learn how to extract sentiment from text with orange. First, let us get the data. We will use Twitter widget to retrieve tweets with hashtag machine learning to see how people feel about this field. You can use any other keyword or even browse by ad mention. We will keep the analysis manageable by retrieving only 100 tweets in the English language. Let's go! I always like to check my data in a corpus viewer. We seem to have some podcasts, articles, retweets and so on. Now it's time to compute sentiment scores. Sentiment analysis can be lexicon-based, semi-supervised or supervised. In orange, we're using lexicon methods, which means we store lists of positive and negative words, then compute how many occurrences of each there are in the text. The approach we'll be using, Vader, is a little smarter and can work with phrases, negations and punctuation. So, three exclamation marks will count more than a single one. Also, it doesn't report just on a single score, but will report on a positive, negative and neutral score and on the final compound score. We have computed the scores for our corpus. The easy way would be to use data table, then find the scores and sort by frequency. But this is just no fun. Let us plot them instead. I will use Hitmap, which shows all four attributes and where the blue represents the lower scores and the yellow and the white, the higher scores. Still, our data is unorganized. Let us use clustering to put rows with similar scores together. Now my negative scores are at one end and the positive ones at the other. Select, say, the negative cluster and observe the texts in the corpus viewer. It seems like some are indeed negative, but others not so much. Certain words, such as problem, are considered negative. Yet, in the language of machine learning, they're totally normal. We deal with problems every day. To finish on a positive note, let us select the positive cluster and inspect it. Just as before, we have some truly positive tweets and some that are ambiguous. Today we have learned how to extract sentiment from text in orange, how to plot it in a Hitmap and how to explore the results. Stay tuned for more text mining videos.