DEV Community

Discussion on: Spoken Language Understanding (SLU) vs. Natural Language Understanding (NLU)

Collapse
 
leonardpuettmann profile image
Leonard Püttmann

Great article! What do you think about IBM watson? I used their NLU services for my bachelors thesis.

Collapse
 
dilek profile image
Dilek Karasoy Picovoice

Glad you enjoyed it.

You're right! Given the limited space, I didn't include IBM Watson in the article. The answer to this question also is "it depends", unfortunately.

Accuracy, ease of use and total cost of ownership are common criteria people use generally. [picovoice.ai/blog/selecting-natura...]

Watson NLU might be preferable if you're familiar with the Watson ecosystem - ease of use. However, if you care about privacy, IBM keeps the voice data for training purposes. So, it can be a deal-breaker.

At Picovoice, we also publish open-source benchmarks. For accuracy, you can reproduce it and evaluate the accuracy across various SnR. Target environment, noise, users' accents, distance from the microphone are important factors for accuracy. Even just for accuracy, I'd say it depends :)

picovoice.ai/docs/benchmark/nlu/#r...

Collapse
 
leonardpuettmann profile image
Leonard Püttmann

Wow, I didn't know that! Is there a place where you can see which providers keeps your data and which don't? Or do I need to look that up in their terms of service? Thanks for the attached articles, I will check out the benchmarks, too

Thread Thread
 
dilek profile image
Dilek Karasoy Picovoice

Not sure but fine prints, T&Cs, FAQs do the job and also explain how to opt-out when possible.

Will my data be stored by NLU?
By default, all Watson services log requests and their results.

ibm.com/cloud/watson-natural-langu...