A Sentiment-Analysis project that aims at determining the opinion of the user utterance if it is positive or negative. The project is written in Python language and requires the Natural Language ToolKit (NLTK) to be installed before usage.
Three main stages:
Normalization, stop-words removal and stemming.
The Term-Frequency-Inverse-Term-Frequency (IT-IDF) was used as a feature vector for the classifier
The Support Vector Machine (SVM) and Naive Bias classifiers were used in the experiment.
Six datasets were used for the experiment
| Data-set | Positive Samples | Negative Samples |
|---|---|---|
| Twitter tweets | (1000) 50% | (1000) 50% |
| Product Attraction reviews | (2073) 96% | (82) 4% |
| Hotel Reviews | (10775) 69% | (4798) 31% |
| Movies Reviews | (969) 63% | (556) 37% |
| Product Reviews | (3101) 72% | (1172) 28% |
| Restaurants Reviews | (8030) 73% | (2941) 27% |
| Unified data-set | (25948) 71% | (10549) 29% |
The following tables shows the results obtained by the experiment:
SVM classifier results:
| Data-set | Precision | Recall | F-measure | Accuracy |
|---|---|---|---|---|
| Twitter tweets | 88% | 77% | 81% | 82% |
| Product Attraction reviews | 99% | 100% | 98% | 96% |
| Hotel Reviews | 96% | 98% | 88% | 83% |
| Movies Reviews | 87% | 95% | 82% | 73% |
| Product Reviews | 90% | 98% | 86% | 78% |
| Restaurants Reviews | 94% | 99% | 85% | 75% |
| Unified data-set | 93% | 99% | 83% | 72% |
Naive Bias classifier results:
| Data-set | Precision | Recall | F-measure | Accuracy |
|---|---|---|---|---|
| Twitter tweets | 88% | 86% | 84% | 84% |
| Product Attraction reviews | 88% | 100% | 98% | 96% |
| Hotel Reviews | 88% | 99% | 83% | 72% |
| Movies Reviews | 88% | 100% | 77% | 63% |
| Product Reviews | 88% | 99% | 85% | 74% |
| Restaurants Reviews | 94% | 99% | 84% | 73% |
| Unified data-set | 92% | 99% | 83% | 72% |