What it does
It includes a head-to-head benchmark rather than calling TypeSafe's hosted Jev.
Benchmarks & research
An independent text-scoring model that tries to improve on the jevlike starter project.
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
It includes a head-to-head benchmark rather than calling TypeSafe's hosted Jev.
You can use this tool to build a system that sorts customer messages. Your developer gives the program a message, like a refund request, alongside a list of departments. The tool reads the message and picks the best department. It can also help control game characters by picking their next action.
To start, your developer trains the software using a text file of past messages and correct choices. Before the program makes a decision, it must always receive the complete list of possible answers. It cannot invent a new choice on its own.