ClassMate

1
ClassMate

• A System for Automated Event Extraction from
  Course Websites
• Ashutosh Kulkarni Harry Robertson

2
ClassMate Pipeline

• Web crawler

3
ClassMate Pipeline

• Web crawler
• NER to tag text
• ANNIE system
• Specialized tags

4
ClassMate Pipeline

• Web crawler
• NER to tag text
• ANNIE system
• Specialized tags
• Event extraction
• Centered on date tags
• Variable size window

5
ClassMate Pipeline

• Web crawler
• NER to tag text
• ANNIE system
• Specialized tags
• Event extraction
• Centered on date tags
• Variable size window

• Features
• Words, tags, etc.
• Key uni/bigrams
• Heuristics

6
ClassMate Pipeline

• Web crawler
• NER to tag text
• ANNIE system
• Specialized tags
• Event extraction
• Centered on date tags
• Variable size window

• Features
• Words, tags, etc.
• Key uni/bigrams
• Heuristics
• Classification
• Kmeans clustering
• Derandomized start

7
ClassMate Pipeline

• Web crawler
• NER to tag text
• ANNIE system
• Specialized tags
• Event extraction
• Centered on date tags
• Variable size window

• Features
• Words, tags, etc.
• Key uni/bigrams
• Heuristics
• Classification
• Kmeans clustering
• Derandomized start
• Category labelling

8
Results
Event Precision Recall F1 score
Office Hours 70 93.33 0.79
Exam dates 80 93.02 0.86
Due dates 88 92.63 0.90
Overall 79.33 92.99 0.85
9
Feature study
10
The Future of Time Management
11
The Future of Time Management