Sarcasm Sniffer is situated at the intersection of natural language processing and sentiment analysis to address challenge of sarcasm detection within textual data, accurately identifying sarcasm presents a nuanced task that conventional sentiment analysis tools often struggle to accomplish.
Twitter Sarcasm Detection Using Machine Learning & Deep Learning Models This repository is focusing on the application and evaluation of machine learning and deep learning models for sarcasm detection within textual data enriched with emojis and hashtags.
- Build a good dataset which has sarcastic documents
- Annotate the data as 0/1 (sarcastic/not sarcastic)
- Identify different patterns in the text that reveal sarcasm
- Build a model that classifies a new unseen text or tweet as Sarcastic or not sarcastic
- Evaluate the model built using f1 score
- Machine Learning - Logistic Regression
- Deep Nueral Network - Long Short-Term Memory (LSTM)
-
Labelled Dataset The labelled dataset employed in this project was acquired from the GitHub repository crafted by Muhammad Adyl, accessible at Sarcasm Detection Dataset. Train.csv & Test.csv
-
Realtime Dataset I employed Scraper API as a robust tool for extracting a pertinent dataset of tweets. The utilization of Scraper API facilitated responsible and respectful data gathering, ensuring a seamless and uninterrupted process. Sarcastic_tweets.csv & Non_sarcastic_tweets.csv
-
Involves leveraging the 'nltk' library and the SentiWordNet lexical resource in Python. Assigns sentiment scores to words for a nuanced understanding of sentiment. - SentiWordNet_3.0.0_20130122.txt
-
Pre-trained Word Embeddings - glove.twitter.27B.100d.txt https://nlp.stanford.edu/projects/glove/
- Logistic Regression - 0.893
- LSTM - 0.754