Document Similarity

Recommendation systems are powerful discovery machines. They can help your customers find similar content or products. They are also particularly useful in the legal and compliance industries, where not having the full picture can have serious real-world consequences. To provide a state-of-the-art, intuitive user experience, these systems rely on semantic similarity rather than keyword matching.

By transforming documents into high-density, semantically rich text embeddings, document similarity can be computed based on their content. This is akin to how we as humans would recommend similar documents. Instead of checking for lexical matches, we think about the meaning contained in a document and then try to select the texts that are semantically closest to it.

To leverage document similarity in deepset Cloud, you can choose any suitable embedding model from a commercial provider or open source platform. deepset Cloud makes it easy to try out different models and see them perform side by side. You can even send your prototype systems to stakeholders and testers, so that you can gauge how they would perform in the real world.

Our LLM platform for AI teams lets you swap, evaluate, and monitor models, so that you can take immediate action if you notice a drop in performance. By choosing to work in deepset Cloud, you provide your users with the most intuitive and comprehensive information discovery experience, and your developers with the smoothest and most stable project workflow — from prototyping to production and beyond.

If you want to have more detailed information on the above, please contact us here.