Friends, I have a major update on the blog. Integrating Microsoft Bot Framework with hybris. Overall, the main part of the article is not about integration, but about Microsoft’s bot solution. It includes two products: Bot Framework and Language Understanding Intelligent Service (LUIS). An interesting solution for creating an IVR-like experience in chats. Welcome, as such materials are either not yet available in Russian, or very few (but I could not find any).
Month: September 2016
September 18 2016, 20:24
I managed to make some progress with the recommendation system (bigdata analysis, creating recommendations on what to buy). I’m preparing an article for hybrismart.
I really need some kind of demo database. Specifically, I need a list of products and statistics on the sales of these products grouped together in one basket tied to some abstract user (ID). If I can get such a database, then I can give you the following recommendations: (a) what else to suggest to the customer ID from the products that they have not yet bought and (b) which products to recommend on the product card, using the sales data.
Can anyone share? Ideally, of course, if the product database were in English, but that’s probably an unattainable ideal 🙂
September 18 2016, 15:29
Found an interesting resource for learning English. I was looking to rewatch “Scent of a Woman” in its original version with subtitles, and voila, found something useful. It allows you to watch videos in the original version with subtitles, and when you hover over a word, it immediately translates the word right there in the video. You can turn on subtitles in both languages at once, which is very convenient. The Russian subtitles are small. And also, if you pause the video, the most recent subtitles are displayed – so you can reread what was missed not being understood in time.
I love such startups. Probably everyone around was saying, “Stop. What’s the point in investing effort, you’ll never have enough money to buy movie rights.” But the guys just ignored all these opinions and made something good for themselves. Probably, they don’t buy rights after all, and someday the range of videos will greatly diminish, but right now, there’s plenty to watch there.
You can click on a word, and it goes into a dictionary. Also, this site has one “how not to do things” pattern – at the bottom of the page there are various links, which are unreachable because a list of movies loads dynamically.
September 17 2016, 18:40
September 17 2016, 11:46
Ran 5km in 26:44, 5:34/km. Got upset. She says she was saving energy for the finish, but it suddenly appeared, and she could have gone faster. I think when you have a stream and a huge hill with a 26m elevation change before the finish, that’s cool 😉 well done, Lisa!
September 17 2016, 11:11
Turns out, last weekend was the biggest competition of the year, yet it’s considered just a minor school race. In the photo – the start. Somewhere in the middle – Lisa. Five km
September 16 2016, 20:46
The variety of tastes and preferences in Virginia is at least indicated by a Ford Mustang parked by the house with eyelashes over the headlights.
September 16 2016, 16:31
The music I am working to today
September 16 2016, 16:13
Hello everyone. I’ve come up with an idea for a rapid personalization engine for e-commerce, for which I need live traffic, big data. I would be very grateful if someone could send me the logs.
Any e-commerce logs that can be used to learn something and based on which some demonstrable conclusions can be made are suitable (for example, if a customer only browses a certain section or visits products from one group, they should be categorized under “favorite customers of this group”).
The only complexity is that most logs do not have a customer identifier. However, they do have IPs, which will suffice for now. But if anyone has customer IDs, that would be just perfect.
Ideally, I would also like to get data on orders. I don’t need surnames, first names, emails, or addresses. Even the names of the products are not necessary.
As a result, I plan to create a prototype system, which will receive messages from the e-commerce site, simplistically in the form of logs, and immediately, in real time, rules will be applied (I plan to use the Drools engine) such as: “[when] the customer has visited 10 pages from the sports section, [then] they should be shown a sports banner.” Information that customer X should be shown such a banner is sent back to the site, and the next time a page with banners loads, it will be displayed.
So, there is no such system yet, but I am sure I can build a prototype over the weekend or a little more. I am concerned about performance. I need to test all this with large volumes, a high frequency of events. That’s why I need the logs.
I will publish the results on a blog about hybris/e-commerce. I can either disclose or not disclose your data – it depends on your preference. Among the nearest topics related to this on the site, there is a recommendation system and three articles about Drools in e-commerce.
Hi, I'm Rauf Aliev. 