NATO Advanced Study Institute on Mining Massive Data Sets for Security

Filtering Multi-Lingual Terrorist Content with Graph-Theoretic Classifi-cation Tools

author: Mark Last, Ben-Gurion University of the Negev

Description

Since the web is increasingly used by terrorist organizations, the ability to automatically detect multi-lingual terrorist-related content is extremely important. In this talk, we present an efficient detection methodology based on the recently developed graph-based web document representation models. Evaluation is performed on corpora in English and Arabic languages.

You might be experiencing some problems with Your Video player.

*Slides*
0:00	Filtering Multi-Lingual Terrorist Content with Graph-Theoretic Classification Tools
0:34	Outline
1:32	Important Preliminaries (1)
3:03	Important Preliminaries (2)
3:36	Propaganda in Arabic Organization: Palestinian Islamic Jihad (1)
5:00	Propaganda in Arabic Organization: Palestinian Islamic Jihad (2)
5:52	Propaganda in Russian Organization: Hamas
6:31	Propaganda in English and Hebrew Organization: Hezbollah
8:14	Training Materials How to build a rocket engine
9:06	Training Materials (cont.) How to prepare a highly explosive acetone peroxide (the stuff is very serious)
10:39	Tactical Orders (?)
12:07	Challenges in Filtering Terrorist Content
15:29	Web Document Representation and Categorization
15:40	Text Categorization (TC) Basic Definition
16:15	Text Categorization (TC) Tasks
17:02	The Vector-Space Model (Salton et al., 1975)
18:03	The “Bag of Words” Approach A Practical Example (1)
18:21	The “Bag of Words” Approach A Practical Example (2)
18:53	The “Bag of Words” Approach A Practical Example (3)
21:58	Advantages of the Vector-Space Model (based on Joachims, 2002)
22:41	Limitations of the Vector-Space Model
24:05	DIVIDE ET IMPERA (“Divide and Rule”) The Word Separation in the Ancient Latin
24:48	Alternative Representation of Multilingual Web Documents:
25:13	Relevant Definitions (Based on Bunke and Kandel, 2000)
26:28	The Graph-Based Model of Web Documents
27:27	The Standard Representation
28:40	The Simple Representation
29:00	The n-distance Representation
29:27	The n-simple Representation
29:49	The Absolute Frequency Representation
30:02	The Relative Frequency Representation
30:10	Graph Based Document Representation – Detailed Example Source: www.cnn.com, May 24, 2005
30:38	Graph Based Document Representation - Parsing
30:58	Graph Based Document Representation - Preprocessing (1)
31:07	Graph Based Document Representation - Preprocessing (2)
31:13	Standard Graph Based Document Representation
32:00	Simple Graph Based Document Representation
33:29	“Lazy” Categorization with Graph-Based Models
35:05	Distance between two Graphs
35:36	Relevant Definitions (Based on Bunke and Kandel, PRL, 2000)
36:07	More Graph-Theoretic Definitions (1)
36:14	More Graph-Theoretic Definitions (2)
36:24	More Graph-Theoretic Definitions (3)
36:49	More Graph-Theoretic Definitions (cont.) (1)
36:59	More Graph-Theoretic Definitions (cont.) (2)
37:16	More Graph-Theoretic Definitions (cont.) (3)
37:23	MMCSN Distance Measure between two Graphs
38:38	k-Nearest Neighbors with Graphs
40:16	The Hybrid Approach to Document Categorization (Markov et al., 2006)
41:59	Predictive Model Induction with Hybrid Representation
42:44	Frequent Subgraph Extraction Example
44:21	Frequent Subgraph Extraction: Complexity
48:10	Case Study 1
48:16	Document Collection
49:00	Preprocessing of Documents in Arabic
50:10	Accuracy Results
51:27	Resulting Decision Tree
53:07	Does the word الصهيوني (“Zionist”) indicate a terrorist document?
54:17	Case Study 2 - Categorization of Terrorist Web Documents in English
54:24	Document Collection
54:53	Results for the Hybrid Smart Approach Maximum Graph Size: 100 Nodes
55:33	Resulting Decision Tree Subgraph Frequency Threshold: 0.55
56:59	Conclusions
57:43	Future Work
58:23	References (1)
58:53	References (2)
59:10	References (3)
59:30	Thank you!

Lecture rating

People found this lecture:
Worth seeing
because it is:
Valuable and informative
Well presented
Easily understandable
Acceptably recorded
You need to login to cast your vote.

Report a problem or upload files

If you have found a problem with this lecture or would like to send us extra material, articles, exercises, etc., please use our ticket system to describe your request and upload the data.
Enter your e-mail into the 'Cc' field, and we will keep you updated with your request's status.