Islamabad, April 12, 2023 (PPI-OT): Sentiment analysis through artificial intelligence (AI) can be an effective way to analyse controversial topics. The utilization of efficient AI techniques and tools can also help plan policies as the observed trends can predict the direction a certain situation is headed to. This was highlighted by Sana Sohail, data scientist, who was presenting her MS research paper, titled ‘Investigation of Feminism Trends through Sentiment Analysis, Using Machine Learning and NLP Techniques’, during an in-house session held on April 6, 2023.
The scholar investigated the trends of feminism in Pakistan during the last five years using sentiment analysis as a combined AI-based technique of machine learning and natural language processing (NLP) and presented the insights as extracted by AI models. As per the data models, neutral sentiments on the issue of feminism were recorded to be the highest overall, while positive sentiments were the lowest in five years. The study was conducted by applying three AI models of sentiment analysis, namely VADER, pysentimiento and Roberta-large, on comprehensive Twitter dataset to see sentiments of people on positive, negative, and neutral scales, along with the visualization of words and emoji’s.
After data pre-processing and model implementation, the AI tools observed the trends across five years and extracted the findings as mentioned above. The study also compared insights of human data labelling and AI data labelling and concluded that out of three AI models, the findings of Roberta-large, with 130 GB vocab data, were closest to human findings showing similar figures as manual data labelling. This demonstrates that efficient AI tools with proper data handling can be a substitute for human intelligence, under specific conditions, it was concluded.
For more information, contact:
Outreach and Publications,
Institute of Policy Studies (IPS)
1, Street 8, F-6/3, Islamabad, Pakistan
Tel: +92-51-8438391-3 (Ext: 122)
Fax: +92-51-8438390
Website: www.ips.org.pk