This license is in no way Open Source. Yes, you can play with the source, but you cannot build something useful with it and redistribute under the same license.
You are right that HTK is not OpenSource and nowhere on the web site does it claim to be. However, your second claim is totaly wrong. Many groups use HTK to train acoustic models and language models that they then ship in their products (with their own recognition software).
What does "at will employment" have to do with non-competes? "at will" just means that both sides can terminate the employment at any time for any reason (or no reason at all).
> State-of-the-art recognizers have an error rate of ~10% on that test
Well, it is more like 15%, especially if you want to do it fast (this typically means 10 times slower than realtime). One can't really afford to run at 300xRT for large-scale transcription. When we had to recognise 500 hours of broadcast news we ran our system at around 15xRT.
Slashdot Mirror
You are right that HTK is not OpenSource and nowhere on the web site does it claim to be. However, your second claim is totaly wrong. Many groups use HTK to train acoustic models and language models that they then ship in their products (with their own recognition software).
Gunnar (maintainer of HTK)
What does "at will employment" have to do with non-competes? "at will" just means that both sides can terminate the employment at any time for any reason (or no reason at all).
most importantly the 3.4GHz one on the quoted SPEC page has 2MB of L3 cache, but the 3.6GHz one has none.
You might want to re-check your textbook for the definition of the median.
Redhat just released an advisory with links to updated RPMS: RHSA-2003-279
Your explanation of the cross product is sligtly wrong. I think you mean
| A x B | = |A| |B| sin[theta]
i.e. you forgot the magnitde on the left hand side and it's sin instead of cos
> State-of-the-art recognizers have an error rate of ~10% on that test
Well, it is more like 15%, especially if you want to do it fast (this typically means 10 times slower than realtime). One can't really afford to run at 300xRT for large-scale transcription. When we had to recognise 500 hours of broadcast news we ran our system at around 15xRT.
Gunnar