Mahsa Mohaghegh

Cross-layer optimisation for quality of service support in wireless sensor networks

Wireless sensor networks need to deliver real-time services such as video, audio and traditional data services therefore providing efficient quality of services (QoS) support is essential. In this paper we aim to address the time-delay parameter of QoS this is implemented using a new cross-layer framework design. The concept of cross-layer design is based on architecture where different layers can exchange information in order to improve the overall network performance. Promising results achieved by cross-layer optimization initiated significant research activity in this area. We present results from simulations of the new cross layer design and traditional OSI model using the OMNET++ software. We show that the cross layer design provides a feasible and flexible approach to solving the conflict between different layers in a standard OSI model. We demonstrate that the cross layer optimization is a promising solution and that enhances the quality of service in wireless sensor network applications.

An analysis of the effect of training data variation in English-Persian Statistical Machine Translation

Globalization has made machine translation an attractive area of research and development. As technology opens up e-commerce opportunities, companies must overcome language barriers to reach new potential customers and partners. Web2.0 with tools like Google Translate makes the web more accessible. Statistical Machine Translation has been used for translation between many language pairs contributing to its popularity in recent years. It has however not been used for the English/Persian pair. This paper presents the first such attempt and describes the problems faced in creating a corpus and building a base line system. Our experience with the construction of a parallel corpus during this study and the problems encountered especially with the process of alignment are discussed. The prototype constructed and its evaluation is described and results analyzed. In the final part of the paper, conclusions are drawn and work planned for the future is discussed.

A hierarchical phrase-based model for English-Persian statistical machine translation

In this paper we show that a hierarchical phrase-based translation system will outperform a classical (non-hierarchical) phrase-based system in the English-to-Persian translation direction, yet for the Persian-to-English direction, the classical phrase-based system is preferable. We seek to explain why this is so, and detail a series of translation experiments with our SMT system using various bilingual corpora each with both toolkits Moses (non-hierarchical) and Joshua (hierarchical).

A Framework for Evaluating Anti Spammer Systems for Twitter

Despite several benefits to modern communities and businesses, Twitter has attracted many spammers that have overwhelmed legitimate users with unwanted and disruptive advertising and fake information. Detecting spammers is always challenging because of the huge volume of data that needs to be analyzed while spammers continue to learn and adapt to avoid being detected by anti-spammer systems. Several spam classification systems are proposed that use various features extracted from the content and user’s information from their Tweets. Nevertheless, no comprehensive study has been done to compare and evaluate the effectiveness and efficiency of these systems. It is not known what the best anti-spammer system is and why. This paper proposes an evaluation framework that allows researchers, developers, and practitioners to access existing user-based and content-based features, implement their own features, and evaluate the performance of their systems against other systems. Our framework helps identify the most effective and efficient spammer detection features, evaluate the impact of using different numbers of recent tweets, and therefore obtaining a faster and more accurate classifier model.

Parallel Text Identification Using Lexical and Corpus Features for the English-Maori Language Pair

Comparable corpora contain significant quantities of useful data for Natural Language Processing tasks, especially in the area of Machine Translation. They are mainly the source of parallel text fragments. This paper investigates how to effectively extract bilingual texts from comparable corpora relying on a small-size parallel training corpus. We propose a new technique to filter non parallel articles in Wikipedia based on Zipfian frequency distribution. We also use the SVM approach to find parallel chunks of text in a candidate comparable document. In our approach we use a parallel corpus to generate the required features for the training step. The evaluations of generated bilingual texts are promising.

Ensemble Statistical and Heuristic Models for Unsupervised Word Alignment

Statistical word alignment models need large amounts of training data while they are weak in small-sized corpora. This paper proposes a new approach of an unsupervised hybrid word alignment technique using an ensemble learning method. This algorithm uses three base alignment models in several rounds to generate alignments. The ensemble algorithm uses a weighed scheme for resampling training data and a voting score to consider aggregated alignments. The underlying alignment algorithms used in this study include IBM Model 1, 2 and a heuristic method based on Dice measurement. Our experimental results show that by this approach, the alignment error rate could be improved by at least 15% for the base alignment models.

An overview of the challenges and progress in PeEn-SMT: First large scale Persian-English SMT system

This paper documents recent work carried out for PeEn-SMT, our Statistical Machine Translation system for translation between the English-Persian language pair. We give details of our previous SMT system, and present our current development of significantly larger corpora. We explain how recent tests using much larger corpora helped to evaluate problems in parallel corpus alignment, corpus content, and how matching the domains of PeEn-SMTs components affect translation outcome. We then focus on combining corpora and approaches to improve test data, showing details of experimental setup, together with a number of experiment results and comparisons between them. We show how one combination of corpora gave us a metric score outperforming Google Translate for the English-to-Persian translation. Finally, we outline areas of our intended future work, and how we plan to improve the performance of our system to achieve higher metric scores, and ultimately to provide accurate, reliable language translation.