cv

Senior Software Engineer

Discourse Analytics

Full stack, Devops, Data analytics

• Data

Industry Advisor

FAST NUCES

Advisor to the Industry Advisory Board

MS

Lahore university of management science, Pakistan

Software Development, Data mining and A.I, Recommendation system

• Adv Topics in Data mining

Merit Scholarship

FAST - NUCES

Tuition Scholarship

An Unsupervised Method for Discovering Lexical Variations in Roman Urdu Informal Text

ACL

We present an unsupervised method to find lexical variations in Roman Urdu informal text. Our method includes a phonetic algorithm UrduPhone, a feature-based similarity function, and a clustering algorithm Lex-C. UrduPhone encodes roman Urdu strings to their phonetic equivalent representations. This produces an initial grouping of different spelling variations of a word. The similarity function incorporates word features and their context. Lex-C is a variant of k-medoids clustering algorithm that group lexical variations. It incorporates a similarity threshold to balance the number of clusters and their maximum similarity. We test our system on two datasets of SMS and blogs and show an f-measure gain of up to 12% from baseline systems.

Understanding Types of Users on Twitter

Researchgate

People use microblogging platforms like Twitter to involve with other users for a wide range of interests and practices. Twitter profiles run by different types of users such as humans, bots, spammers, businesses and professionals. This research work identifies six broad classes of Twitter users, and employs a supervised machine learning approach which uses a comprehensive set of features to classify users into the identified classes. For this purpose, we exploit users' profile and tweeting behavior information. We evaluate our approach by performing 10-fold cross validation using manually annotated 716 different Twitter profiles. High classification accuracy (measured using AUC, and precision, recall) reveals the significance of the proposed approach.

Personalized versus non-personalized tag recommendation: A suitability study on three social networks

IEEE INMIC

Tag recommendation systems are either personalized or non-personalized. Personalized tag recommendation utilizes a user's tagging behavior from her tagging history for predictions. Whereas non-personalized recommendation systems recommend what is popular and relevant to the user. In this study, we have analyzed the role of personal tagging history in recommending tags. The experiments are done on three folksonomy datasets: Delicious, Flickr and Bibsonomy. Important results for three popular tag recommendation algorithms: PITF, FolkRank and Adapted PageRank are reported in terms of prediction quality. It is found that users' history usage preferences change across all data sets; hence overall prediction quality of personalized recommendation system may suffer. We discover a generic life cycle of folksonomy users on the basis of their history usage. We propose this life cycle can be used to improved.

Legat Assist: RAG based LLM

UK legal proceeding data used to built LLM

• Hullucination
• RAG