Question of the Week (05/03/21)

From Stojancho Tudjarski

Q: “there are so many obvious things that can be done with only 20 lines of Python code (thinking Huggingface). However, it is quite hard to explain to non NLP folks. How do we optimally approach potential customers and explain NLP capabilities?”

A: 3 perspectives from matched professionals

“In my experience, I’ve found that it’s best to show customers what our product can do, and anchor our conversation around the product demo as early as possible. NLP is more abstract than the other ML disciplines, so talking in terms of concrete examples and analogies is typically more effective than deconstructing a question theoretically.”

  • Anonymous, Founder & CEO tech startup that uses NLP-

“Potential customers don’t care about (or even know about) NLP. The best way to approach them is by talking about their own pain points. If selling a sourcing tool, you can say “right now sourcing is extremely time consuming and repetitive for you; With our tool, you simply enter high level criteria as if you are talking to a human sourcer. Our software leverages cutting edge NLP to produce a shortlist of candidates with the same accuracy and nuance that a human sourcer has”. Customers then might ask how your AI/NLP works, you can explain at a very high level a few ways your NLP works. Tie each explanation back to how it helps the customer. E.g. “we trained our NLP models from millions of profiles from scratch to understand synonyms, so you won’t miss a smart engineer just because s/he didn’t explicitly mention django in addition to python”. Avoid too many technical terms like “word embeddings”.

“People might not know what a transformer model does or how it works, but if you tell them it’s good at labeling what certain things are in a sentence (NER), and that you can then automate processes based on that, people can start to get ideas. I think one of the ways to optimally approach new customers would be to show them what you’ve already built”.

Have your own Q? Ask here:

Interested in sharing expertise? Answer here:





Leave a Reply