Natural Language Processing will supercharge your business (Pt. 1)
Natural Language Processing (NLP) is integral to the future of your business, and I want to help you be prepared to make the most of it.
Before we get to all of the incredible use cases and talk about what the future will look like when it’s powered by NLP, we’ve got to talk about what exactly it is.
NLP includes multiple different pieces:
Speech to Text
- The ability for an algorithm to take .mp3, .wav, and other sound files and write out everything that is said.
Understanding of Language
- The ability for an algorithm to understand what is written or said. This comes in many different forms and the area that still has yet to be fully cracked. It is the HOLY GRAIL of NLP end-to-end functionality. If you’ve seen ChatGPT taking the internet by storm, they’ve done an incredible job with this component of NLP!
Emotional Context
- Some algorithms are able to actually understand the inflection of a voice within the sound file. For example, you could say, “Hey” and be either greeting a friend or yelling when somebody cuts you off on the freeway.
Text to Speech
- The ability for an algorithm to take a written set of words and create speech with natural human inflection.
Let’s do a deeper dive into each pillar of NLP so you can more fully see what’s possible.
Speech to Text
I actually used a basic NLP service on a free app on my phone to write the original draft of this article! I dictated an initial draft on my phone, and then finished editing on my computer.
Clearly, Speech to Text still has its limitations. I was using a free service that was not connected to wifi. For a non-wifi small application, my phone was able to take the words I said and write them down. However, I couldn’t do any formatting, punctuation, or capitalization. Essentially, I had one very long sentence. I then went back and added periods, commas, capital letters, etc.
Also, I did this on a plane with plenty of background noise! The application was able to convert most of the words I said and worked quite well. That is… until the overhead speaker came on and took over!
As I was formatting, I was also removing “cabin crew please” and “seatbelt sign is on” every few lines! But for all of that interference, I was very impressed with its performance!
Understanding of Language
There is a very distinct difference between “understanding” and “understanding”… if you understand what I mean!
While it’s true that the Speech to Text app I was using could “understand” my words, it didn’t “understand” their meaning or my intent behind them.
Let’s do a live example right now:
shut down the app to open a new page
Well, clearly nothing happened and I’m still recording this article.
If this app could “understand” my language, it would have responded to my request. If I was on my HoloLens 2 using an app with NLP, it would actually shut down the entire device.
This distinction actually reminds me of my 1-year-old! He can repeat back to me exactly what I say, but still has no clue what he is actually saying to me (and normally has no intention of responding to my requests!).
Emotional Context
Some applications can actually attempt to understand the meaning behind a user’s language as well as the emotion behind it. As humans, we can’t help but do this. Think about a simple statement like…
hey man sit down
How’d you read it at first pass?
Was it a pleasant greeting to a friend at your house?
Was it an annoyed plane passenger yelling at the guy holding up takeoff?
Was it a kind father consoling his son warmly after losing a baseball game?
Same words, different emotional context. And even as humans, we’re constantly getting this wrong (or at least not-quite-right).
Hopefully NLP algorithms can learn from the original experts on emotional context recognition: dogs.
“You are so weird, yes you are, yes you are…” with a sweet, excited tone will have a dog jumping with excitement! But “I love you, you’re a good dog” with a rude and abrasive tone will send it away with its tail between its legs.
Text to Speech
This is a very straightforward pillar of NLP. Write something down, and it becomes a sound file. Around 15 years ago, we had this tech — but it sounded very robotic still. The tech was just still very new! Nowadays it can sound like anything from a robot (on purpose this time) to a pirate to a 25-year-old English female with a northern dialect!
You can take it even one step further and use a service like MURF.AI and sound exactly like anyone you want! It uses multiple Machine Learning algorithms all combined together. The dialect/accent can mimic anyone and any voice, assuming there is a big enough database.
There are lots of providers on the scene with NLP services to business and end users.
- MSFT — Azure Cognitive services: Cloud-based. Easy to deploy in any region of the world.
- NVIDIA — Riva: Needs to run on a NVIDIA card. This can be run in the cloud on a NVIDIA card, on prem server, or on local device.
- AWS: Cloud-based. Similar to Azure.
- And many more niche companies, apps, and softwares.
Each has its own pros and cons. Many of these companies offer simple services “right out of the box,” and others enable companies to use their tech as a baseline and then build on top of it to specialize for their unique use case.
So now you know what NLP is! But now it’s time to figure out why it matters for your business.
Do you need further guidance on how to navigate this for your business specifically? DM me on LinkedIn and we can chat!
