Sentiment Analysis Parser - IEEE and NASA Standards
Training models used
Categorical - IEEE Standards Abstracts Model: a model trained by extracting abstracts from IEEE standards and using the most common keyword in the respective standard as its category (14 categories).
Categorical - IEEE Standards Abstracts Model with merged categories: a model trained by extracting abstracts from IEEE standards and using the most common keyword in the respective standard as its category (12 categories).
Categorical - NASA Technical Standards Model: a model trained by taking 80% of the given 42 NASA technical standards (=> 33), extracting the Scope of each and using the division number in the respective standard as its category (9 categories).
Datasets evaluated
IEEE Standards Abstracts (new): the new (different) abstracts from IEEE standards.
NASA Technical Standards (20% of 42): the remaining 9 standards' scopes.
Running IEEE Models (#1 and #2) on 1,273 standards abstracts
Results: Categorical IEEE Standards Abstracts Model
Networks: 53
Software: 23
Devices: 39
Interface: 3
System: 835
Circuits: 45
Management: 17
Equipment: 81
Metadata: 3
Wireless: 12
Security: 10
Design: 117
Data: 15
Radio: 20
Results: Categorical IEEE Standards Abstracts Model trained on some categories merged
Wireless Networks: 95
Software: 13
Devices: 54
Interface: 1
System: 892
Circuits: 45
Management: 11
Equipment: 52
Security: 5
Design: 67
Data: 12
Radio: 26
IEEE Model: Accuracy
Training Set - iteration 1: 0.254
Training Set - iteration 100: 0.968
Test Set - iteration 1: 0.166
Test Set - iteration 100: 0.573
Running NASA Standards Model (#3) on 9 standards abstracts