1
00:00:00,480 --> 00:00:05,754
So in this next segment of the class, 
we're going to look at log-linear models. 

2
00:00:05,754 --> 00:00:09,978
Log-linear models offer a very powerful 
set of techniques as an alternative to 

3
00:00:09,978 --> 00:00:14,266
the smooth estimation techniques we've 
seen earlier in this class which are, in 

4
00:00:14,266 --> 00:00:21,260
some sense or other, simple way of 
estimating and, parameters of models. 

5
00:00:21,260 --> 00:00:24,539
And defining models. 
So we'll the see that log-linear models 

6
00:00:24,539 --> 00:00:28,193
allow us to make use of much 
representation for many natural language 

7
00:00:28,193 --> 00:00:32,752
processing problems. 
And we'll see that they use some rather 

8
00:00:32,752 --> 00:00:37,99
interesting and very powerful ideas about 
how to estimate the parameters of these 

9
00:00:37,99 --> 00:00:40,208
models. 
So we'll be describing the basic 

10
00:00:40,208 --> 00:00:43,589
framework of log-linear models and then 
we'll be looking at their applications in 

11
00:00:43,589 --> 00:00:47,214
a number of problems in natural language 
processing.