How Mergeflow uses AI
By "AI" we mean software that somehow learns from data or other feedback. For example, we would count neural networks as AI. By contrast, a fixed string matching algorithm, even if it's complex, is not AI--as long as it doesn't learn from data.
Mergeflow uses AI across its platform, for various purposes. You can see and interact directly with some of these technologies, whereas others are part of the analytics "under the hood".
AI in your browser
More recently we've started using AI--specifically, large-scale generative language models--directly in our user interface. For example, Mergeflow's AI Mindmap Generator helps you map out your topic. It's like your digital thought partner for tech discovery.
For the "battery membranes" query above, the mindmap might look like this, for example:
AI "under the hood": Using AI for analytics
Mergeflow helps you discover things that are relevant, but that you perhaps didn't even know might exist. New companies, for example, or subject matter experts that you don't know yet. Because we want to help you discover such new things, we can't simply operate with lists of known companies, people, materials, etc.. And this is where AI comes in.
Discovering entities (companies, people, technologies, materials, ...)
Mergeflow detects various types of entities in the contents it collects. These entities include companies, people, emerging technologies, materials, and others. So that you can actually discover new companies etc. that are relevant but that aren't on some existing list, Mergeflow uses AI-based algorithms for several aspects of entity discovery.
Assigning patent classes to texts
Using AI-based algorithms, Mergeflow assigns patent classes to non-patent texts. This includes all Mergeflow data sets that are not patents (venture capital news, market estimates, scientific publications, news, blogs, etc.).
By assigning patent classes to non-patents, Mergeflow emulates what a patent analyst would do manually when they examine a technology for patenting.