Sequence classification model for sentiment analysis of Russian conversational text
Project structure
notebooks
.ipynb
files
- csv_preprocessing.ipynb - pre processing initial
csv
files - eda.ipynb - exploratory data analysis
- training.ipynb - model training
- evaluation.ipynb - model evaluation
src
.py
files
training.py
containsTrainer
class to train BertForSequenceClassificationprediction.py
containsPredictor
class to predict sentiments from text data using pre trained modelmetrics.py
contains functions to computeprecision
,recall
, andf1
scoresevaluation.py
contains a function to evaluate pre trained model using test datainput_data_preprocessing.py
contains utility function to fix initialcsv
files
Metrics
Positive | Neutral | Negative | Average | |
---|---|---|---|---|
Precision | 0.744 | 0.717 | 0.799 | 0.753 |
Recall | 0.811 | 0.623 | 0.833 | 0.756 |
F1 | 0.776 | 0.667 | 0.816 | 0.753 |