Data to Words

Generated by author using Midjourney

One of the more promising aspects of the natural language generation capabilities is the ability to interpret data and be able to relay it back to people in plain language.

The opportunity isn’t just in the interpretation of data and speaking to someone and speaking to someone about what it means, it’s also in understanding the tone that should be conveyed and what is the best way for the individual receiving the information to receive it.

e.g. predicted bus delay > 5 minutes “looks like the bus is running late” or “Uggh… delays suck don’t they?”

The challenge can be broken down into a few parts:

• Understanding what the entities are in a dataset. What does time mean?What is the system actually checking?
• What things should be interpreted and what things should be communicated?
• What channels should be used to communicate this information?
• How much evidence needs to be provided in the communication?
• What is the tone of the communication that balances the needs of the sender and those of the recipient?
• How can the system learn whether it’s being effective?

Cracking this nut can yield game-changing advances in service and support.