1
00:00:00,648 --> 00:00:03,804
[MUSIC]

2
00:00:03,804 --> 00:00:07,844
So let's start with describing this
task in a little bit more detail.

3
00:00:07,844 --> 00:00:11,861
Okay, so let's say we're sitting here,
reading an article.

4
00:00:15,019 --> 00:00:17,202
Here we are, reading our article.

5
00:00:17,202 --> 00:00:19,811
And actually it's Carlos
who's reading this article,

6
00:00:19,811 --> 00:00:21,720
because this is an article about soccer.

7
00:00:23,870 --> 00:00:28,670
Or as he would call it football if
he was wearing his Argentina jersey

8
00:00:29,670 --> 00:00:35,470
or footsie ball if he was
wearing his Brazil jersey and

9
00:00:35,470 --> 00:00:37,925
clearly, I'm not pronouncing
either of those words correctly.

10
00:00:37,925 --> 00:00:39,970
Carlos can correct me later.

11
00:00:39,970 --> 00:00:43,365
But the point is that he likes
this article and what we'd like

12
00:00:43,365 --> 00:00:47,505
to do is retrieve another article that
he might be interested in reading.

13
00:00:49,363 --> 00:00:50,746
But a question is how do we do this?

14
00:00:50,746 --> 00:00:54,123
There are lots and lots of articles out
there and I can't expect him to go and

15
00:00:54,123 --> 00:00:57,890
read each of them and say yes,
he's interested or no, he's not.

16
00:00:57,890 --> 00:01:01,913
So we like to think of a way to
automatically retrieve a document that

17
00:01:01,913 --> 00:01:03,619
might be of interest to him.

18
00:01:03,619 --> 00:01:07,612
By questions here are first, how to we
measure similarity between articles?

19
00:01:07,612 --> 00:01:11,851
We need to have that in order to say that
this article is similar to the one he's

20
00:01:11,851 --> 00:01:15,040
reading now and
might also be of interest to him.

21
00:01:15,040 --> 00:01:18,010
Or that, here's a large set of
articles that are very different and

22
00:01:18,010 --> 00:01:20,490
probably are not of interest to him.

23
00:01:20,490 --> 00:01:24,800
And then the second question is, how
are we gonna search over the articles that

24
00:01:24,800 --> 00:01:28,146
exist out there and
retrieve the next article to recommend.

25
00:01:28,146 --> 00:01:31,699
[MUSIC]