 It's the holiday season and every holiday there's so many cookies to eat and just thinking about it I'm imagining the ultimate jam cookie. It's rich and sweet, full of homemade jelly and dusted with a nice layer of powdered sugar and served with nice warm tea. But then again there's so many other cookies to choose from. So this year I've made an Excel spreadsheet to help me decide. I have chocolate chip cookies, oatmeal raisin, peanut butter, macaroons and linser. There's just too many to choose from. So let's do some cookie mining. Now I already showed you my collection of cookies but I can also try to load it into orange and then I can ask chat GPT to make some descriptions. Now I'll convert this data into a text corpus so that each cookie has its own description. Then embedding will convert these descriptions into a set of numbers which we can use for clustering for example. We find the coconut cookies are clustered together and so are some of the more traditional European cookies, the spiculas and the vanilla ketfer. Now I don't even know how to pronounce that let alone what that cookie is. I'm going to need some pictures. Now luckily I actually have a bunch of named images of each of my cookies. So let's try to add all of these into our workflow as well. Okay I'll merge the descriptions with the images and then feed this data into my clustering again. So here we go, the European cookies, the pecan cookies, all the funny looking cookies. Now I can even ask Orange to describe the cluster so using chat GPT I can type in something like what is specific to the following list of cookies in the start and answer with up to five words in the end to make sure chat GPT doesn't get too wordy. Now here's what it says about the European cluster and here's what it says about the peanut cluster and what it says about the chocolate cluster. Now let's instead try to describe what kind of cookie I want to eat. For example right now I feel like eating a soft cookie with pecans. So I'll use text embedding again and try to find some matches. There they are, all the pecan cookies from my database. Maybe instead I feel like eating cookies filled with jam. Wow, they really do look great. Now I could really do this all day. I could even maybe use t-sne to map all the cookies or try to use a bag of words and then keyword extraction. Or maybe I could extract some features using GPT and use them for predictions. But I won't do that right now. Instead I'm going to go eat some cookies. Happy holidays everyone and I hope you have a great new year.