1
00:00:00,343 --> 00:00:04,183
[MUSIC]

2
00:00:04,183 --> 00:00:07,522
And as part of using this data set,

3
00:00:07,522 --> 00:00:13,625
the first thing that we're gonna do,
just like in the module,

4
00:00:13,625 --> 00:00:18,013
we're going to build a word-count vector.

5
00:00:18,013 --> 00:00:25,513
Build the word count vector for
each review.

6
00:00:28,463 --> 00:00:33,357
Now, normally, you'd have to implement
this and explain something that goes

7
00:00:33,357 --> 00:00:38,410
review, separates words,
called tokenizing, build the count vector.

8
00:00:38,410 --> 00:00:43,230
But one of the nice things about
using the tools here in this

9
00:00:43,230 --> 00:00:48,190
course is that with just one command,
we can build that word count vector.

10
00:00:50,030 --> 00:00:53,650
So in products,
I'm going to add a new column

11
00:00:53,650 --> 00:00:58,910
called word_count,
which is gonna start my word count.

12
00:00:58,910 --> 00:01:02,782
And if you just call
graphlab.text_analytics,

13
00:01:02,782 --> 00:01:07,102
it's a text analytics toolbox for
a bunch of functions,

14
00:01:07,102 --> 00:01:10,080
there is one called count_words.

15
00:01:10,080 --> 00:01:13,182
Notice there's also one called
count_ngrams if you wanna use bi-grams,

16
00:01:13,182 --> 00:01:14,110
tri-grams and so on.

17
00:01:15,388 --> 00:01:20,530
And as input, I'm going to give
the same products as frame,

18
00:01:20,530 --> 00:01:25,070
but I'm going to ask it to count
the words in the review column.

19
00:01:26,600 --> 00:01:29,020
And so we're gonna execute that and
it's done.

20
00:01:29,020 --> 00:01:33,000
And so now, if we take another look

21
00:01:33,000 --> 00:01:37,610
at the products table, so
at the head of the table you'll

22
00:01:39,240 --> 00:01:43,910
see that now we have a fourth
column with the word_count.

23
00:01:43,910 --> 00:01:46,750
So we're gonna explore that
a little bit more soon.

24
00:01:46,750 --> 00:01:51,022
But you see, for this first review,
include the word and

25
00:01:51,022 --> 00:01:53,310
five times, stink once.

26
00:01:53,310 --> 00:01:58,470
Probably that's why it wasn't a good
products review, but there's others.

27
00:02:01,560 --> 00:02:05,429
[MUSIC]