Research Assistant, Applied Logic Lab, Computer Science, Stony Brook University. (May 2008 - May 2012)
A. HearSay Non-Visual Web Browser.
Areas: Machine Learning, Information Retrieval, Web Mining and Web Accessibility.
Project Description: Visually disabled Individuals use screen readers to browse the Web. Sequential processing of screen readers (e.g. JAWS) makes Web browsing time-consuming, strenuous and causes significant information overload. HearSay is a multi-modal non-visual web browser, envisioned as the next-generation assistive technology that will alleviate many Web Accessibility problems. Two years in the making, Hearsay is now a well tested working system. On the algorithmic side it embodies several techniques including: content analysis that partitions web pages into meaningful sections for ease of navigation; context-directed browsing that exploits the content surrounding a link to find relevant information as users move from page to page; process models that help users to quickly get to content fragments needed for doing web transactions; detection and handling of changes in web pages that helps users stay focused; statistical models for associating labels with web elements even if they are missing, as in images without alternative text; personalized audio macros for doing repetitive tasks; statistical language detection; etc. Information Retrieval and Machine Learning principles underlying these techniques make them robust and scalable.
On the interface side, HearSay supports multiple output (audio, visual, Braille) and input (speech, keyboard, touch, phone keypad) modalities. HearSay can be used as a desktop application or can be used remotely via the plain phone. The browser also supports the IBM’s Social Accessibility network that brings together end-users and supporters who can collaboratively create and utilize the accessibility metadata.
1. Muhammad Asiful Islam, Faisal Ahmed, Yevgen Borodin, I.V. Ramakrishnan, "Thematic Organization Of Web Content for Distraction-Free Text-To-Speech Narration". In Proceedings of the 14th International ACM SIGACCESS Conference on Computers and Accessibility, (ASSETS'12).
2. Faisal Ahmed, Yevgen Borodin, Andrii Soviak, Muhammad Asiful Islam, I.V. Ramakrishnan and Terri Hedgpeth "Accessible Skimming: Faster Screen Reading of Web Pages". In Proceedings of the 25th ACM Symposium on User Interface Software and Technology, (UIST'12).
3. Muhammad Asiful Islam, Faisal Ahmed, Yevgen Borodin, I.V. Ramakrishnan, "Tightly Coupling Visual and Linguistic Features for Enriching Audio-Based Web Browsing Experience". In Proceedings of the 20th ACM Conference on Information and Knowledge Management (CIKM'11).
4. Muhammad Asiful Islam, Yevgen Borodin, I.V. Ramakrishnan, "Mixture Model based Label Association Techniques for Web Accessibility". In Proceedings of the 23rd ACM Symposium on User Interface Software and Technology, (UIST'10). [pdf]
5. Muhammad Asiful Islam, Faisal Ahmed, Yevgen Borodin, Jalal Mahmud, I.V. Ramakrishnan, "Improving Accessibility of Transaction-centric Web Objects". In Proceedings of the 10th SIAM International Conference on Data Mining, (SDM'10). [pdf]
6. Y. Borodin, F. Ahmed, M.A. Islam, S. Feng, Y. Puzis, V. Melnyk, G. Dausch, I.V. Ramakrishnan, "Hearsay: A New Generation Context-Driven Multi-Modal Assistive Web Browser". In Proceedings of the 19th International World Wide Web Conference (WWW'10). [pdf]
7. Faisal Ahmed, Muhammad Asiful Islam, Yevgen Borodin, I.V. Ramakrishnan, "Assistive Web Browsing with Touch Interfaces". In Proceedings of the 12th International ACM SIGACCESS Conference on Computers and Accessibility, (ASSETS'10). [pdf]
8. I.V. Ramakrishnan, Jalal Mahmud, Yevgen Borodin, Muhammad Asiful Islam, Faisal Ahmed, "Bridging the Web Accessibility Divide". In Proceedings of 4th Int'l Workshop on Automated Specification and Verification of Web Systems (WWV'08), Elsevier ENTCS, Volume 235, Pages 107-124, April 2009. [pdf]
B. Inference and Learning in Probabilistic Logic Programs with Continuous Random Variables.
Areas: Statistical Machine Learning, Logic Programming, Statistical Relational Learning.
Project Description: Statistical Relational Learning (SRL), an emerging area of Machine Learning, aims at modeling problems which exhibit complex relational structure as well as uncertainty. It uses a subset of first-order logic to represent relational properties, and graphical models to represent uncertainty. Probabilistic Logic Programming (PLP) is an interesting subfield of SRL. A key characteristic of PLP frameworks is that they are conservative extensions to non-probabilistic logic programs which have been widely used for knowledge representation. PLP frameworks extend traditional logic programming semantics to a distribution semantics, where the semantics of a probabilistic logic program is given in terms of a distribution over possible models of the program. However, the inference techniques used in these works rely on enumerating sets of explanations for a query answer. Consequently, these languages permit very limited use of random variables with continuous distributions. In this thesis, we extend PRISM, a well-known PLP language, with Gaussian random variables and linear equality constraints over reals. We provide a well-defined distribution semantics for the extended language. We present a symbolic inference and parameter-learning algorithms for the extended language that represents sets of explanations without enumeration. This permits us to reason over reason over complex probabilistic models such as Kalman filters and a large subclass of Hybrid Bayesian networks that were hitherto not possible in PLP frameworks. The inference algorithm can be extended to handle programs with Gamma-distributed random variables as well. An interesting aspect of our inference and learning algorithms is that they specialize to those of PRISM in the absence of continuous variables. By using PRISM as the basis, our inference and learning algorithms match the complexity of known specialized algorithms when applied to Hidden Markov Models and Kalman Filters.
1. Muhammad Asiful Islam, C.R. Ramakrishnan, I.V. Ramakrishnan, "Inference in Probabilistic Logic Programs with Continuous Random Variables". In Proceedings of the 28th International Conference on Logic Programming (ICLP'12).
Undergraduate Thesis, Department of CSE, Bangladesh University of Engineering and Technology, 2006.
Title: "A Hybrid Approach to Design Neural Network Ensembles." supervised by Dr. Md. Monirul Islam.
Abstract: This thesis presents two cooperative ensemble learning algorithms, NegBagg and NegBoost, for designing ANN ensembles. NegBagg and NegBoost train different individual ANNs in an ensemble incrementally by using the NC learning algorithm. Bagging and boosting algorithms are used in NegBagg and NegBoost, respectively, to create different training sets for different individual ANNs in an ensemble. The idea behind using NC learning in conjunction with bagging and boosting algorithms is to make interaction and cooperation among individual ANNs in an ensemble work better. Both NegBagg and NegBoost use a constructive approach to determine the ensemble architecture automatically. They have been tested on a number of benchmark problems in machine learning and ANNs, including Australian credit card assessment, breast cancer, diabetes, glass, heart disease, letter recognition, satellite, soybean and waveform. The experimental results show that NegBagg and NegBoost can produce ANN ensembles with good generalization ability by using a small number of training epochs.
1. Md. Monirul Islam, Xin Yao (Fellow, IEEE), S.M. Shahriar Nirjon, Muhammad Asiful Islam, Kazuyuki Murase, "Bagging and Boosting Negatively Correlated Neural Networks". IEEE Transactions on Systems, Man, and Cybernetics Part B: Cybernetics, vol. 38, NO. 3, June 2008. (http://www.ieeesmc.org/) [pdf]
2. S.M. Shahriar Nirjon, Muhammad Asiful Islam, Sarker Tanzir Ahmed. "Diverse Ensemble Creation by a Data Feeding Approach" In Proceedings of the Ninth International Conference on Computer and Information Technology (ICCIT), Bangladesh, 2006.