1
00:00:00,280 --> 00:00:02,540
Another machine learning service on AWS

2
00:00:02,540 --> 00:00:03,920
is called Amazon Kendra.

3
00:00:03,920 --> 00:00:06,720
So this one is a fully-managed document search service

4
00:00:06,720 --> 00:00:08,300
that is powered by machine learning

5
00:00:08,300 --> 00:00:11,940
and it allows you to extract answers from within a document.

6
00:00:11,940 --> 00:00:14,210
That document could be text, PDF, HTML

7
00:00:14,210 --> 00:00:17,480
PowerPoints, Microsoft Word, FAQs, et cetera, et cetera.

8
00:00:17,480 --> 00:00:18,790
So you have a lot of data sources

9
00:00:18,790 --> 00:00:20,330
where these documents may be

10
00:00:20,330 --> 00:00:23,280
and you see some of them right now in the screen

11
00:00:23,280 --> 00:00:26,570
and they're going to be indexed by Amazon Kendra

12
00:00:26,570 --> 00:00:28,500
which is going to build internally

13
00:00:28,500 --> 00:00:31,270
a knowledge index powered by machine learning.

14
00:00:31,270 --> 00:00:33,930
And how does it help from an end-user perspective?

15
00:00:33,930 --> 00:00:36,540
Well, we get natural language search capabilities

16
00:00:36,540 --> 00:00:37,790
just like you go on Google.

17
00:00:37,790 --> 00:00:39,450
So for example, if a user says,

18
00:00:39,450 --> 00:00:43,330
Hey, where is the IT support desk into Amazon Kendra?

19
00:00:43,330 --> 00:00:45,470
Kendra can reply, 1st floor.

20
00:00:45,470 --> 00:00:48,220
And this could be due to the fact that Kendra knows

21
00:00:48,220 --> 00:00:50,230
from all the resources that it took

22
00:00:50,230 --> 00:00:53,290
that the IT support desk was on the 1st floor,

23
00:00:53,290 --> 00:00:54,820
which is quite awesome.

24
00:00:54,820 --> 00:00:57,360
And also, you can just do a normal search

25
00:00:57,360 --> 00:01:00,260
and it will learn from the user interaction and feedback

26
00:01:00,260 --> 00:01:02,230
to promote preferred search results

27
00:01:02,230 --> 00:01:04,440
which is called incremental learning.

28
00:01:04,440 --> 00:01:07,610
Finally, you can fine tune the search results, for example,

29
00:01:07,610 --> 00:01:10,230
based on the important data, importance of data,

30
00:01:10,230 --> 00:01:13,700
the freshness, or whatever custom filters you have, okay?

31
00:01:13,700 --> 00:01:15,630
So from an exam perspective,

32
00:01:15,630 --> 00:01:17,990
whenever you see a document search service,

33
00:01:17,990 --> 00:01:19,430
think Amazon Kendra.

34
00:01:19,430 --> 00:01:21,743
That's it, I will see you in the next lecture.