Browsing by Author "Silva, T"

Now showing 1 - 8 of 8

item: Conference-Full-text
Brain-computer interface for controlling cursor movements – a review
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2016-12) Hettiararchchi, EP; Silva, T; Fernando, KSD
Brain-Computer Interface (BCI) is a rapidly evolving technology that builds a direct channel between the human brain and the computer. It acquires brain signals, analyzes them and translates these analyzed brain signals into commands which can carry out a specific action in an external output device. Therefore BCI is a different approach which makes the communication between the external world and brain without using the normal output pathway which is composed of neurons and muscles. Electroencephalography (EEG) is the widely used brain signal due to its fine temporal resolution and low cost. The acquired EEG signals are pre-processed to extract the features before classifying them. Finally, the classified EEGs are converted to commands which can be used to carry out a specific action in the external world such as controlling the cursor movements. This paper reviews the literature related to BCI and presents how BCI has contributed to alleviate the challenges in cursor movements controlling.
item: Article-Full-text
Identifying Social Roles Using Heterogeneous Features in Online Social Networks
(Wiley-Blackwell Publishing Ltd, 2019) Liu, Y; Du, F; Sun, J; Silva, T; Jiang, Y; Zhu, T
Role analysis plays an important role when exploring social media and knowledge-sharing platforms for designing marking strategies. However, current methods in role analysis have overlooked content generated by users (e.g., posts) in social media and hence focus more on user behavior analysis. The user-generated content is very important for characterizing users. In this paper, we propose a novel method which integrates both user behavior and posted content by users to identify roles in online social networks. The proposed method models a role as a joint distribution of Gaussian distribution and multinomial distribution, which represent user behavioral feature and content feature respectively. The proposed method can be used to determine the number of roles concerned automatically. The experimental results show that the proposed method can be used to identify various roles more effectively and to get more insights on such characteristics.
item: Conference-Full-text
Inpra – an intelligent system for writing while doing research
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Karunananda, A; Silva, T; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
Research students have difficulty following at least some steps in research methodology. Nowadays, computing technology has made it easy for researchers to execute the steps with the support of software tools. Although many research students gradually manage to conduct the research, some find it challenging to write the thesis in time. This issue stems from separating the conduct of research from writing the thesis. This paper presents the extension to our Intelligent Personal Research Assistant, InPRA, to enable a researcher to write the thesis incrementally while doing the research. Here, we have introduced an intelligent writing template, which guides the writing process differently from the order of chapters of the thesis. The intelligent template is also integrated with tools such as Zotero, MS Word, and Grammarly software. The incremental writing process enables the generation of research proposals, progress reports, conference papers, theses, and journal papers. The writing extension to InPRA has been evaluated with research students at the postgraduate level.
item: Conference-Full-text
Ontology-based knowledge modelling for handling criminal law cases in Sri Lanka
(Faculty of Information Technology, University of Moratuwa., 2021-12) Amarasena, K; Weerasinghe, U; Weththasinghe, D; Silva, T; Ganegoda, GU; Mahadewa, KT
Law can be considered as one of the required fields, which controls the behaviour of people within the society to a greater extent. In most legal systems, constitutions, legislation, case laws etc., are documented in natural language. Lawyers, law students and other related parties who need to access these legal documents for a specific legal case encounter the difficulty of extracting only the relevant data from these documents. Systems to manage criminal cases and facilitate extracting crucial facts regarding past cases are rare in practice. On the other hand, some systems which focus on general document retrieval utilize traditional classification techniques to facilitate document retrieval. Such a system overlooked domain-specific attributes such as dynamic knowledge updates for knowledge retrieval. To address this deficiency, an ontology-based knowledge modelling system is proposed to automatically identify and extract the necessary information based on the legal cases provided.
item: Conference-Full-text
Personalized mood-based song recommendation system using a hybrid approach
(IEEE, 2023-12-09) Ranasingha, SS; Silva, T; Abeysooriya, R; Adikariwattage, V; Hemachandra, K
Music recommendation systems are becoming a crucial concern for the music industry because of the rise of digitization and the subsequent increase in music consumption. Music applications continuously strive to enhance their recommendation systems to ensure that users have an exceptional listening experience and remain loyal to their platform. In the early days, the recommendation system used collaborative filtering and content-based approaches to achieve this goal, but these approaches have an issue with a cold start, and context awareness of these approaches is less. Researchers identified in the context of the personalization of songs, Emotion, and mood can play a huge role. Research has shown that a user's current emotional state significantly influences their musical preferences in the short term. Therefore, the recommendation system moves toward mood-based recommendation approaches. The vast variety and context-dependent character of the data that must be considered present the main difficulty for moodbased recommendation systems. This information can vary greatly and is depending on several variables, including the user's environment and personal circumstances. Hybrid approaches have shown very good results in this domain. Therefore, in this paper, we are proposing a hybrid approach for a mood-based personalized song recommendation system. This approach combines content-based and context-based approaches together. The proposed solution produces the output as a personalized song recommendation for the music listener. This output is determined by several parameters including user mood, the profile of the user, and history of previously listened to songs. This solution impacts all the stakeholders. it improves the quality of service of music streaming platforms and improves the user experience.
item: Conference-Abstract
Semantic information retrieval based on topic modeling and community interests mining
Rajapaksha, M; Silva, T
Search engines or localized software systems developed for information searching, play an important role in knowledge discovery. Proliferation of data in the web and social media has posed significant challenges in finding relevant information efficiently even using those search engines or other software systems. Moreover, those systems or engines tend to collect large number of data, which could be useful for end users in various ways but have overlooked the meaning of the search phrases, hence generate irrelevant search results. A unit level searching i.e. searching information within a website or page is also not effective as they follow exact keyword matching techniques and ignore the semantic level matching of search phrases. In order to address those deficiencies, this research proposes a hybrid approach which use the semantics of data, community preferences as well as collaborative filtering techniques for semantic information retrieval. More specifically, Topic modeling based on Latent Dirichlet Allocation together with topic-driven based community detection methods are applied for identifying personalized search results generation and hence improve the relatedness of the research results. Based on the proposed hybrid approach a framework for semantic search that can easily be integrated to a software application has been implemented. The evaluation results confirm the effectiveness of search results which outperform benchmark approaches that follow traditional keyword search algorithms.
item: Conference-Full-text
Semantic learning for question and answering systems
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa, Sri Lanka, 2016-12) Dilmi, VKD; Silva, T; Fernando, KSD
Semantic similarity methods play a significant role in different areas including community mining, document clustering, automatic metadata extraction, information retrieval, document clustering, synonym extraction. In the recent past semantic similarity has been approved as a feasible and scalable alternative to grasp natural language. This review paper presents the existing techniques in semantic similarity and how these techniques are applied in question and answering systems. Furthermore, this illustrates the drawbacks of current techniques and recommendations will be presented to improve semantic learning for question and answering systems.
item: Article-Full-text
Topic-based hierarchical Bayesian linear regression models for niche items recommendation
(SAGE Publications Inc, 2019) Liu, Y; Xiong, Q; Sun, J; Jiang, Y; Silva, T; Ling, H
A vital research concern for a personalised recommender system is to target items in the long tail. Studies have shown that sales of the e-commerce platform possess a long-tail character, and niche items in the long tail are challenging to be involved in the recommendation list. Since niche items are defined by the niche market, which is a small market segment, traditional recommendation algorithms focused more on popular items promotion and they do not apply to the niche market. In this article, we aim to find the best users for each niche item and proposed a topic-based hierarchical Bayesian linear regression model for niche item recommendation. We first identify niche items and build niche item subgroups based on descriptive information of items. Moreover, we learn a hierarchical Bayesian linear regression model for each niche item subgroup. Finally, we predict the relevance between users and niche items to provide recommendations. We perform a series of validation experiments on Yahoo Movies dataset and compare the performance of our approach with a set of representative baseline recommender algorithms. The result demonstrates the superior performance of our recommendation approach for niche items.