Hi
I am trying to incorporate the NL capability in my voice application. I have written SLM and trained it using sgc utility. Training was successful. From the Nuance documentation (Advanced NLU Developer's Guide) I came to know that we need to tune for perplexity such that we get the lowest possible value to increase the accuracy.
When I did this for test sentences I got some results with some values for each word.
Please help me in understanding the perplexity results.
How to achieve the lowest value for perplexity?
Thanks & Regards
Swapna