Hindustani or Hindi vs. Urdu: A Computational Approach for the Exploration of Similarities Under Phonetic Aspects

Suffian, Muhammad; Ahmed, Tafseer; Yaseen, Muhammad

doi:10.14569/ijacsa.2020.0111191

Cited by 1 publication

(1 citation statement)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We maintain that the emojis are the concentric part of tweets (as we have seen their counts and distributions in tables 9, 10, and 9) therefore, we catered for their presence and absence in the experimental variation of the datasets. Details of these datasets are provided in table 12.…”

Section: ) Classifiersmentioning

confidence: 99%

Cognitive Relationship-Based Approach for Urdu Sarcasm and Sentiment Classification

Khan,

Ahmed,

Siddiqui

et al. 2023

IEEE Access

Self Cite

View full text Add to dashboard Cite

Humans have a natural tendency to express their emotions, but they are also skilled at using sarcasm to shape their feelings. In cognitive computing and natural language processing research, sentiment analysis and sarcasm detection are typically treated as separate tasks, with each text analyzed in isolation. However, this approach overlooks the connection between sentiment and sarcasm. We believe that sentiment and sarcasm are closely related and should be analyzed together to achieve a better understanding of context and natural language. In this paper, we propose a new framework that leverages the Cognitive Relationship (CR) between sarcasm and sentiment to improve the accuracy of classification. By taking into account the relationship between these two factors, we can achieve better results in sentiment analysis and sarcasm detection. We have also created a new and nearly balanced dataset for sentiment and sarcasm classification in standard Urdu that contains 7,000 tweets, which make up over 210K tokens. To gain a better understanding of the data, we conducted exploratory data analysis on words, hashtags, and emojis. The proposed methodology conducted a variety of classical machine learning classifiers and tested them with different variations of the dataset. After a thorough analysis of the results and errors, we found that the CR-based approach for sarcasm and sentiment classification performed better than the traditional stand-alone (SA) approach. Among the classifiers, Linear Regression and eXtreme Gradient Boosting proved to be the most effective. The sentiment classification based on CR has demonstrated a 9.3% enhancement compared to the stand-alone (SA) method while maintaining an overall improvement of approximately 22% compared to the baseline distribution. In the same way, the sarcasm classification based on CR has shown a 9.1% improvement over the SA approach and approximately 23.6% improvement over the baseline distribution.

show abstract

Section: ) Classifiersmentioning

confidence: 99%