Code repo for paper titled 'Balancing between Holistic and Cumulative Sentiment Classification' published in Online Social Networks and Media - Journal
When using the code-base please use the following reference to cite our work:
DOI: 10.1016/j.osnem.2022.100199
1: The code-base is set to run without additional path settings, if it is downloaded and placed at the downloads folder
2: The data folder must contain the datasets in excel format. The columns must be arranged in the folowing format:
- 1st column : user's opinions,
- 3rd column : labels for three classes classification task i.e.:[0,1,2] (for binary simply set n_classes=2 at the 'configs.py' file,
- 5th column : labels for five classes or fine-grained classification i.e.:[0,1,2,3,4], if exist,
- 6th column : labels for six classes, if exist,
- for many classes text classification, simply place the labels at the corresponding column.
3: The "configs.py" file, includes the hyperparameters for training the model.
4: The embeddings folder must contain the crawl-300d-2M-subword.zip
5: The train_holc.py is the main file to load & train the model.
- tensorflow version: '1.12.0'
- numpy version: '1.16.1'
- sklearn version: '0.19.0'
- nltk version: '3.2.4'
The framework is open-sourced under the Apache 2.0 License base. The codebase of the framework is maintained by the authors for academic research and is therefore provided "as is".