ironySarcasmDetection

URL: https://github.com/teelinsan/ironySarcasmDetection

Description: This is the code’s repository of the paper “A Kernel-based Approach for Irony and Sarcasm Detection in Italian” presented at IronITA @ Evalita2018

Methods

This repository contains the implementation of the system described in the paper “A Kernel-based Approach for Irony and Sarcasm Detection in Italian”, which was presented at IronITA @ Evalita2018. The system uses a kernel-based approach for detecting irony and sarcasm in Italian text.

Key Steps:

1. Feature Modeling: The features for sarcasm and irony detection are modeled using a Jupyter notebook (GenerateKLPFile), as explained in the paper.
2. Preprocessing: Datasets must be preprocessed using a POS-tagger and lemmatizer to generate new dataset copies with preprocessed text.
3. KeLP: The system utilizes KeLP (Kernel-based Learning Packages) for applying the kernel machine approach to the task.

The approach achieved top ranks at Evalita2018, ranking first and second at the sarcasm detection task and sixth and seventh at the irony detection task.

Results

• Sarcasm Detection: Ranked first and second in the sarcasm detection task.
• Irony Detection: Ranked sixth and seventh in the irony detection task.
• The code implements the feature modeling and classification pipeline, using KeLP for kernel-based classification.

Dataset

• The dataset used for this task is from IronITA @ Evalita2018. The dataset can be downloaded here.
• The preprocessing requires converting the dataset into .klp format using a POS-tagger and lemmatizer, creating processed files like:
  • test_ironita2018_revnlt_processed.tsv
  • training_ironita2018_renlt_processed.tsv
• The system uses these preprocessed datasets for feature extraction and kernel machine modeling.