MSc Data Science graduate with research interests in Natural Language Processing (NLP), Explainable AI (XAI), temporal NLP, adaptive machine learning, and social-media text analysis.

I completed an MSc in Data Science at the University of South Wales, United Kingdom, with a background in Computer Science and Engineering.

My MSc dissertation investigated large-scale sentiment classification using approximately 1.6 million tweets from the Sentiment140 dataset. I developed a TF-IDF-based NLP pipeline and compared several traditional machine-learning classifiers, including Logistic Regression, Linear SVC, Bernoulli Naive Bayes, Multinomial Naive Bayes, and SGD.

This work developed my interest in how NLP models behave when language changes over time, how model predictions can be better understood, and how models can be adapted efficiently to evolving social-media data.

• Natural Language Processing (NLP)
• Sentiment Analysis and Text Classification
• Explainable Artificial Intelligence (XAI)
• Temporal NLP
• Adaptive Machine Learning
• Trustworthy and Robust NLP
• Social Media Analytics
• Semantic Representation and Model Explanations

MSc Data Science Dissertation Project

Repository: sentiment-analysis

Key highlights:

• Processed and analysed approximately 1.6 million tweets from the Sentiment140 dataset.
• Developed an end-to-end NLP pipeline for sentiment classification.
• Applied text preprocessing including lowercasing, URL and user-handle masking, emoticon mapping, character normalisation, and lemmatisation.
• Represented social-media text using TF-IDF unigram and bigram features, with a feature space of up to 500,000 terms.
• Implemented and compared Logistic Regression, Linear SVC, Bernoulli Naive Bayes, Multinomial Naive Bayes, and an SGD classifier.
• Evaluated held-out predictions using accuracy, precision, recall, F1-score, and confusion-matrix analysis.
• Logistic Regression achieved the strongest held-out test performance among the evaluated classifiers, with 82.6% accuracy.
• Saved trained classifiers and the fitted TF-IDF vectoriser using Python pickle and used Logistic Regression in a simple user-input sentiment prediction demo.

My current research interests extend this work from sentiment classification on fixed datasets towards NLP systems that can be analysed and adapted as language changes over time.

I am particularly interested in:

• Semantic explanation shift in NLP models
• Temporal language change
• Model explanations and feature attribution
• Label-efficient model adaptation
• Active learning
• Contextual language models
• Semantic representations
• Trustworthy and robust NLP

My proposed PhD direction investigates whether changes in the linguistic evidence used by sentiment classifiers can be compared at the semantic level, whether harmful changes in model reliance can be distinguished from normal language evolution, and whether the affected examples can be prioritised for efficient model updating.

MSc Data Science
University of South Wales, United Kingdom
Merit
Research Project: 77% (Distinction)

BSc Computer Science and Engineering
North South University, Bangladesh

• Python
• scikit-learn
• pandas
• NumPy
• NLTK
• TF-IDF
• Logistic Regression
• Linear SVC
• Naive Bayes
• SGD Classifier
• SQL
• R
• Jupyter Notebook

📧 Email: [email protected]

🔗 LinkedIn: linkedin.com/in/akiba-mehzaveen-mon

BSc Computer Science and Engineering

North South University, Bangladesh

📧 Email: [email protected]

🔗 LinkedIn: https://linkedin.com/in/akiba-mehzaveen-mon