R Scripts for Error Analysis
A set of scripts that are used for analyzing output files of a machine learning system for NLP of review sites
What does it do?
It allows importing the output file in a given format for each of the test cases for analysis and visualization of the results:
- Percision and recall of each category/label
- Confusion Matrix between the categories/labels
- Diff analysis between different runs of the system
How to make it work?
Download R or RStudio from their open source sites
Use the sample files that have the following format
- V1 - Score - how certain is the classification (in the sample file we have only 1.0 and null)
- V2 - isMatch - is the classification correct (can be also calculated by comapring V3 and V4 below)
- V3 - Correct label
- V4 - Classified label as guessed by the system