WEBVTT

1
00:00:13.516 --> 00:00:14.956
My name's David Weller-Fahy.

2
00:00:14.986 --> 00:00:15.966
I'm associated

3
00:00:15.966 --> 00:00:18.016
with MIT Lincoln Laboratory

4
00:00:18.016 --> 00:00:19.816
as an associate staff member.

5
00:00:20.296 --> 00:00:21.536
CHARIOT is an enrichment

6
00:00:21.536 --> 00:00:22.926
and filtering technology

7
00:00:23.206 --> 00:00:25.796
that allows analysts to use it

8
00:00:25.796 --> 00:00:27.726
to select only those documents

9
00:00:27.726 --> 00:00:29.296
out of a huge stream of documents

10
00:00:29.346 --> 00:00:31.166
that are most relevant to them

11
00:00:31.506 --> 00:00:33.416
and then highlights those documents

12
00:00:33.466 --> 00:00:35.246
to minimize the amount of time they have

13
00:00:35.246 --> 00:00:36.286
to spend reading them.

14
00:00:36.726 --> 00:00:38.406
CHARIOT solves the problem

15
00:00:38.406 --> 00:00:40.426
of analysts not having enough time

16
00:00:40.426 --> 00:00:43.196
to review all of the documents and all

17
00:00:43.196 --> 00:00:44.766
of the content that are coming at them

18
00:00:44.766 --> 00:00:45.546
on the internet.

19
00:00:45.956 --> 00:00:47.706
CHARIOT works using a combination

20
00:00:47.706 --> 00:00:49.366
of standard machine learning techniques

21
00:00:49.746 --> 00:00:51.726
as well as two novel techniques,

22
00:00:51.726 --> 00:00:54.336
one in knowledge graph generation

23
00:00:54.486 --> 00:00:57.376
and one in active learning or adapting

24
00:00:57.376 --> 00:00:59.166
to new sources of information.

25
00:00:59.766 --> 00:01:02.156
CHARIOT's benefits are really twofold.

26
00:01:02.516 --> 00:01:04.986
On the one hand you have the benefit

27
00:01:04.986 --> 00:01:06.076
of being able to go

28
00:01:06.106 --> 00:01:07.396
through more documents

29
00:01:07.666 --> 00:01:09.036
than you would otherwise be able

30
00:01:09.156 --> 00:01:09.826
to review.

31
00:01:10.156 --> 00:01:11.516
On the other hand you have the benefit

32
00:01:11.516 --> 00:01:13.676
of being able to review documents faster

33
00:01:13.676 --> 00:01:15.156
than you would otherwise be able to.

34
00:01:15.816 --> 00:01:18.346
Our system automatically highlights the

35
00:01:18.346 --> 00:01:20.326
documents and certain terms

36
00:01:20.326 --> 00:01:22.796
that may be important to their analysis

37
00:01:23.176 --> 00:01:25.266
so that analysts can very quickly

38
00:01:25.266 --> 00:01:27.246
determine whether a document will be

39
00:01:27.246 --> 00:01:28.826
useful in the future or not.

40
00:01:29.226 --> 00:01:31.756
In our trials we've increased the

41
00:01:31.756 --> 00:01:33.706
relevant content in a data stream

42
00:01:33.886 --> 00:01:36.616
from one percent to nearly 80 percent.

43
00:01:37.366 --> 00:01:40.206
CHARIOT makes an analyst's job easier

44
00:01:40.736 --> 00:01:42.466
by enriching and filtering

45
00:01:42.556 --> 00:01:44.086
relevant content.