Detecting Target of Sarcasm Using Ensemble Methods

Code related to the ALTA's 2019 Shared Task - "Detecting Target of Sarcasm using Ensemble Methods". Our work was presented in ALTA 2019 The 17th Annual Workshop of the Australasian Language Technology Association

The link to the paper can be found -> http://bit.do/pradalta

What's this Read Me about ?

Basically , it is to give you an overview on what is this project about. Keep in mind that I've built this very quick and dirty way. It is not in the best shape but I am trying to make sure that at least people can download it and make it easier for them to download and use it

Requirements

• python 3.7
• numpy
• sckikit-learn
• pandas
• ALTA 2019 Shared Data Challenge Dataset (Please kindly obtain for them)

Will update with pip requirements.txt so that you can have the exact one. I will make it easier for you all :). At the moment , I apologize - you just need to go through the files for the imports and install the pip To make your life a bit easier , the major imports or the stuff that I use can be found in the following files below :)

What are the files for

• DoItAll.py - It runs the rule-based system. Just load the CSV file with the title on the target. It will pick it up and get back to you
• LinearRegressionClassifer.py - The main linear regresson. You can take a look at the parameters that were used and also the way how we load the embeddings (.npy which was used from Google Sentence Encoder)
• lstm.py - This to generate the word embeddings - "Universal Sentence Encoder" . I should rename it , but this shows how you can quickly transform it
• rule_weighting.py - Our genetic algorithim in optimizing the weightages for it
• rules_implement.py - The rules from Joshi et al (2018) with our modifications. The original methods are still kept so that you may able to reproduce it

Problems Running ?

Raise an Issue or Reach out to me - I can help. Again it's my first time of publishing some public project. I try to follow best practices , so please pardon me if there are mistakes

TO-DO

• Improve the Read Me
• Refactor the Code - some of the file names do not even make sense or even the method. It can be optimized further
• Add PIP Requirements
• Publish Jupyter-Notebook (to help people step-by-step and explain it to them) QUICK and DIRTY INFO