Comments on: More data usually beats better algorithms?

By: RIGHTSHOPPING

RIGHTSHOPPING — Thu, 24 Nov 2011 08:57:28 +0000

Kitchen works are synergistic with the proper gadgets. Proper gadgets not just assure the proper utilization of the ingredients, but also serve the safety concern of user as well.

By: Siddhartha Reddy

Siddhartha Reddy — Mon, 21 Apr 2008 15:19:09 +0000

Thanks for the comment, Harish. You are absolutely right about the importance of good data in Speech Engineering. In fact I think this is a common theme in many if not most fields of Computer Science. Specifically in Speech Engineering, Google provides some very good support for the idea: Google runs a free 411 (inquiry) service in US that works using speech recognition and they have gone on record to say that the primary purpose of this service is to provide better data for their Speech Recognition algorithms.

By: Harish

Harish — Sat, 19 Apr 2008 09:11:33 +0000

hey sid,
stumbled upon your blog.. interesting post.
the importance of data over the algorithm is a useful concept even in the realms of speech engineering (coding, recognition and synthesis) in order to capture more variability. Ofcourse one might feel that one can use a parametric model (like the HMMs) to circumvent a lot of data but the naturalness (which is very important in recognition/synthesis paradigms) is lost. But, I suppose the resources you’re dealing with has a significant say in what model you adopt. So yes it depends on the application to choose between statistical or parametric models.