Stephen Akuma

Work place: Department of Mathematics and Computer Science, Benue State University, Makurdi, PMB102119, Nigeria



Research Interests: Data Science, Information Retrieval, Machine Learning, Human-Computer Interaction


Dr. Stephen Akuma is currently a Lecturer of Computer Science in the Department of Mathematics/Computer Science at Benue State University. He holds a PhD in Computing and a Master’s degree in Software Development (Distinction) from Coventry University, United Kingdom. Stephen also holds a Bachelor's degree in Computer Science (2.1) from Benue State University. He has a history of working in data science, human-computer interaction, machine learning and information retrieval, focusing on user searching and retrieval behaviour, classical implicit feedback indicators, eye gaze-enhanced interaction, user modelling and personalization. Stephen has been able to continuously take ideas from conception, development, to deployment in lab-based and large-scale environments. He has published in leading scientific proceedings and journals.

Author Articles
A New Query Expansion Approach for Improving Web Search Ranking

By Stephen Akuma Promise Anendah

DOI:, Pub. Date: 8 Feb. 2023

Information systems have come a long way in the 21st century, with search engines emerging as the most popular and well-known retrieval systems. Several techniques have been used by researchers to improve the retrieval of relevant results from search engines. One of the approaches employed for improving relevant feedback of a retrieval system is Query Expansion (QE). The challenge associated with this technique is how to select the most relevant terms for the expansion. In this research work, we propose a query expansion technique based on Azak & Deepak's WWQE model. Our extended WWQE technique adopts Candidate Expansion Terms selection with the use of in-links and out-links. The top two relevant Wikipedia articles from the user's initial search were found using a custom search engine over Wikipedia. Following that, we ranked further Wikipedia articles that are semantically connected to the top two Wikipedia articles based on cosine similarity using TF-IDF Vectorizer. The expansion terms were then taken from the top 5 document titles. The results of the evaluation of our methodology utilizing TREC query topics (126-175) revealed that the system with extended features gave ranked results that were 11% better than those from the system with unexpanded queries.

[...] Read more.
Eye Gaze Relevance Feedback Indicators for Information Retrieval

By Stephen Akuma

DOI:, Pub. Date: 8 Feb. 2022

There is a growing interest in the research on interactive information retrieval, particularly in the study of eye gaze-enhanced interaction. Feedback generated from user gaze features is important for developing an interactive information retrieval system. Generating these gaze features have become less difficult with the advancement of the eye tracker system over the years. In this work, eye movement as a source of relevant feedback was examined. A controlled user experiment was carried out and a set of documents were given to users to read before an eye tracker and rate the documents according to how relevant they are to a given task. Gaze features such as fixation duration, fixation count and heat maps were captured. The result showed a medium linear relationship between fixation count and user explicit ratings. Further analysis was carried out and three classifiers were compared in terms of predicting document relevance based on gaze features. It was found that the J48 decision tree classifier produced the highest accuracy.

[...] Read more.
Factors Affecting Users‟ Measure of Interest: A Study of the Effect of Task, Document Difficulty and Document Familiarity

By Stephen Akuma Chrisina Jayne

DOI:, Pub. Date: 8 May 2019

Data on the web is constantly growing which may affect users’ ability to find relevant information within a reasonable time limit. Some of the factors previously studied that affect users searching behaviour are task difficulty and topic familiarity. In this paper, we consider a set of implicit feedback parameters to investigate how document difficulty and document familiarity affects users searching behaviour in a task-specific context. An experiment was conducted and data was collected from 77 undergraduate students of Computer science. Users’ implicit features and explicit ratings of document difficulty and familiarity were captured and logged through a plugin in Firefox browser. Implicit feedback parameters were correlated with user ratings for document difficulty and familiarity. The result showed no correlation between implicit feedback parameters and the rating for document familiarity. There was, however, a negative correlation between user mouse activities and document difficulty ratings. 

Also, the dataset of all the participants in the experiment was grouped according to task type and analysed. The result showed that their behaviour varies according to task type. Our findings provide more insight into studying the moderating factors that affect user searching behaviour.

[...] Read more.
Development of Relevance Feedback System using Regression Predictive Model and TF-IDF Algorithm

By Stephen Akuma Rahat Iqbal

DOI:, Pub. Date: 8 Jul. 2018

Domain-specific retrieval systems developed for a homogenous group of users can potentially optimise the recommendation of relevant web documents in minimal time as compared to generic systems built for a heterogeneous group of users. Domain-specific retrieval systems are normally developed by learning from users’ past interactions, as a group or individual, with an information system. This paper focuses on the recommendation of relevant web documents to a cohort of users based on their search behaviour. Simulated task situations were used to group users of the same domain. The motivation behind this work is to help a cohort of users find relevant documents that will satisfy their information needs effectively. An aggregated implicit predictive model derived from correlating implicit and explicit feedback parameters was integrated with the traditional term frequency/inverse document frequency (tf-idf) algorithm to improve the relevancy of retrieval results. The aggregated model system was evaluated in terms of recall and precision (Mean Average Precision) by comparing it with self-designed retrieval system and a generic system. The performance of the three systems was measured based on the relevant documents returned. The results showed that the aggregated domain-specific system performed better in returning relevant documents as compared to the other two systems.

[...] Read more.
Other Articles