A novel framework to improve the accuracy of detecting sentiments from textual data

This is a Centre for Decision Research (CDR) Seminar taking place at Leeds University Business School on Wednesday 4 October 2017

Dr Usman Qamar, Director of the Knowledge and Data Science Research Centre (KDRC) at the National University of Sciences and Technology (NUST), Pakistan, will be delivering the second presentation in the Centre for Decision Research's 2017/18 seminar series. 

Sentiment Analysis is increasingly becoming popular in the field of big data analysis where the aim is to analyse sentiments, opinions, or attitudes of people towards different elements such as individuals, topics or services. This can be achieved by either using machine learning or using a lexical-based approach. Recent research suggests that domain-specific lexicons perform better as compared to the domain-independent lexicons.

In this seminar, Dr Usman Qamar presents research proposing the use of machine learning with a lexical-based approach in an attempt to improve the performance of domain-independent lexicons. Their research introduces a novel framework to determine the features based on SentiWordNet – a well-known general-purpose sentiment lexicon. The features are selected based on their subjectivity and their part of speech information. Seven benchmark datasets have been used to assess the proposed framework using ten-fold cross validation. When compared in terms of accuracy and F-measure, the proposed framework outperformed other techniques for sentiment analysis. For example, the accuracy of proposed framework on large movies database was 86.44% as compared to the previously best reported accuracy of 80.4%.

For further information, please contact the Research Office at research.lubs@leeds.ac.uk

About the speaker

Dr Usman Qamar

Dr Usman Qamar has over 15 years of experience in data engineering and decision sciences both in academia and industry having spent nearly 10 years in the UK.

He has a Masters in Computer Systems Design from University of Manchester Institute of Science and Technology (UMIST), UK. His MPhil in Computer Systems was a joint degree between UMIST and University of Manchester which focused on feature selection in big data. In 2008/09 he was awarded PhD from University of Manchester, UK. His PhD specialisation is in Data Engineering, Knowledge Discovery and Decision Science. His Post PhD work at University of Manchester, involved various research projects including hybrid mechanisms for statistical disclosure for Office of National Statistics (ONS), London, UK, churn prediction for Vodafone UK and customer profile analysis for shopping with the University of Ghent, Belgium.

Dr Qamar also has the honour of being the finalist of the British Council’s International Professional Achievement Award as well as the recipient of the prestigious Charles Wallace Fellowship.