Recommender Systems are one of the greatest inventions of Machine Learning, providing your users with personalized content based on their behavior. Making sure, however, that the content is at its highest quality, can be cumbersome. You need to know as much as possible about each candidate item.
Next to Vector Similarity Search approaches using pre-trained embeddings, another very promising method is to apply Named Entity Recognition to the descriptions you have. This way, you can scrape machine-readable information out of your texts, helping you to find the most similar or interesting item.
As with any Machine Learning model, such methods require large amounts of training data. kern offers the development environment to build such datasets in short time, with all tools at your fingertips you need. From regular expressions to active transfer learning, 3rd party applications or crowd labeling.
Are you interested in giving kern a shot? Try it now with our free version. Reach out to us if you have any questions along the way, we’re happy to help you anytime.