minhpqn / sentiment_analysis_vlsp_2016

My submission system for sentiment analysis task at VLSP 2016

VLSP-2016 evaluation campaign: Sentiment Analysis

Reproduce the results of the technical report

Run the script SA_submission_method.py

python SA_submission_method.py [output]

[output] is the name of the output file. It is optional.

Evaluate the result

Use the script evaluate.py

python evaluate.py minhpqn_ftri_SA_submission.txt data/Evaluation/SA2016-TestData-Ans/test_tokenized_ANS.txt

Sample some errors made by my system on the test set

Use the script sample_erros.py

python evaluate.py minhpqn_ftri_SA_submission.txt data/Evaluation/SA2016-TestData-Ans/test_tokenized_ANS.txt N

In which N is the number of samples for each class (POS, NEG, NEU)

Official Evaluation Results

Our system rank third (system sa3) in VLSP Evaluation Campaign. The following table shows precision, recall, F1 score for positive, negative categories and the average F1 of the two categories.

The following tables shows the results for three categories positive, negative, and neutral.