Reseach Project by Bale Chen from New York University Shanghai. If interested, please contact via email (bale[dot]chen[at]nyu[dot]edu).
The emoji2vec twitter sentiment analysis data is originally from the paper Sentiment of Emojis by Petra Kralj Novak, Jasmina Smailović, Borut Sluban, Igor Mozetič. Due to convenience, I access the data from another research project, emoji2vec, that uses their data. Their repository can be accessed here.
The SemEval-2015 data is the SemEval-2015 Task 11: Sentiment Analysis of Figurative Language in Twitter by Aniruddha Ghosh, Guofu Li, Tony Veale, Paolo Rosso, Ekaterina Shutova, John Barnden, and Antonio Reyes. Please refer to their original paper for more information. Note that many tweets have perished or been deleted. Thanks to the SemEval-15 organizers who created a Twitter bot and posted those tweets again to avoid perishing.
Note that this SemEval-2015 dataset is deprecated on Jan 29, 2023 and is no longer available in this repo. Please contact me if you have any question regarding this dataset.
As a form of dissemination of knowledge, the research work and findings are presented in from of an academic report. Please refer to the Report.pdf file to see the complete paper. If you want to understand the main idea of this project, I suggest to read it first!
In the scripts folder are the multiple scripts that I wrote for different functionalities. The train.py is essentially the main idea of the whole piple, though it's not the final version I ran to generate results. Yet, it's adequate for understanding the whole process. train_robust.py contains the code I ran for the final experienment on each model-method pair. train_meta_robust.py is a complementary script for train_robust.py that trains the models specifically with the meta-feature method (refer to Section 3.3.2 in the report). contraction_map.py and utils.py have some handy functions for processing data. Note that all scripts are executed on NYU HPC Greene Cluster, so personal computers might not be able afford such heavy computation.
In the visuals folder I provided some visualization of results and data.
- python 3.10.4
- emoji 1.7.0
- numpy 1.22.3
- pandas 1.4.2
- pytorch 1.12.0
- openpyxl 3.0.9
- regex 2022.3.15
- gensim 4.1.2
- nltk 3.7
- scikit-learn 1.1.1
- scipy 1.7.3
- tqdm 4.64.0
I would like to extend my warmest gratitude to my research supervisor and mentor Professor Mathieu Laurière. He provides me with scholarly and resourceful advice, guides me through this summer research. It is my great honor and pleasure to finish this study with him and receive his email greeting on my birthday. This work was also supported in part through the NYU IT High Performance Computing resources, services, and staff expertise. Besides, I genuinely appreciate NYU and NYUSH for offering me the DURF research opportunity. Thanks to all my friends and family who helped me throughout this summer. The research would not have been possible without any of you ❤️.