UCSC Digital Library Collection:

Short Essay Grading Using Automated Essay Scoring Techniques

Thu, 14 May 2015 00:00:00 GMT

Title: Short Essay Grading Using Automated Essay Scoring Techniques Authors: Perera, G.R.; Perera, D.N. Abstract: Assessment of the knowledge is considered as one of the most important aspect of the learning process. Conventional class room learning has been shifted and developed towards m-learning and e-learning concepts. Despite its development the underlying problem of ‘How to assess learning?’ still remains. Essay type questions are considered as the most appropriate question types compared to closed ended questions to evaluate the knowledge of the students. However the evaluation of those answers generally consumes huge time, effort and unavoidable human errors such as biasness, inconsistency, etc. Therefore development of an automated essay assessment is encouraged due to these reasons. The primary strength of automated scoring compared to human scoring lies in its efficiency, absolute consistency in applying the same evaluation criteria across essay submissions and over time and ability to provide instantaneous feedback. Computers are neither influenced by external factors (e.g., deadlines) nor emotionally attached to an essay. Computers are not biased by their stereotypes or preconceptions of a group of examinees. The main objective of this research is to present a novel approach towards developing an automated essay scoring (AES) system to evaluate short essays with the combination of natural language processing (NLP) techniques and Vector Space Models (VSM) to reduce time, effort and eliminate unavoidable human errors such as biasness and inconsistency in evaluation of students' short essay answers. This incorporates NLP techniques such as lemmatization, tokenization, handling of spelling mistakes, relation of objects, negation, sentence cases and short term resolution. Besides Vector Space Models are Latent Semantic Analysis (LSA), Latent Dirichlet Allocation (LDA), and Probabilistic Latent Semantic Analysis (PLSA). VSMs is used to provide a score for student answers by means of providing a score by computing similarities of the model answer across a collection of student answers in a vector space model and finally predict an accurate score by the help of the NLP techniques. Importance of this approach is that we do not utilize any domain specific corpus; thus, training the system for each prompt is not necessary. Instead a semantic space is built using students’ answers. A model answer is used to measure the coverage of the answer. Further we have discussed some important implications when implementing the system as well. We obtained a correlation of 0.813 and 0.773 for the two data sets (vision & hearing data sets) with average value of human raters’ score. Finally the results conclude that there’s a significant and strong relationship between average value of human raters’ score and the system score.

Enhancing the Skill Development of ICT/Computer Science Undergraduates in Large Practical Oriented Classrooms through an Empirical Pedagogical Model In Sri Lankan Context

Thu, 14 May 2015 00:00:00 GMT

Title: Enhancing the Skill Development of ICT/Computer Science Undergraduates in Large Practical Oriented Classrooms through an Empirical Pedagogical Model In Sri Lankan Context Authors: Janaa, V.; Asanthi, H.H.A.T.; Yatigala, M.S. Abstract: Proper comparison between the skill development of Information and Communication Technology/ Computer Science (ICT/CS) university students of Sri Lanka and the industry requirements is essential in order to achieve an acceptable level of global competency. To leverage the skill development according to the global standards, the quality education which has five dimensions should be attained by the university system. The quality of the Teaching and Learning (TL) process is one dimension of the quality education. The quality development process consists of different sub processes such as curriculum design, pedagogical design, implementation quality, outcomes assessment and resource provision. Quality pedagogical design is one of the most significant components of the quality education. So it is chosen as a way to enhance the skill development of ICT/CS students in this study. In order to enhance the skill development of the students the following main objectives were used: (i) identifying the quality status of the pedagogical design of ICT/CS university education in Sri Lanka against the international standards (ii) specifying the quality gap if there is any, against the international standard, (iii) taking actions to minimize the quality gap and finally develop a model embedded with pedagogical techniques utilizing existing human and physical resources to enhance the quality of ICT/CS Education through an empirical approach to achieve an acceptable level of global competency. This type of empirical exploration is timely applicable to achieve the global competency in ICT/CS education, since Sri Lanka is experiencing a rapid development in the Asian region. To achieve the objectives of this study, it was designed in two phases. In the phase I, international standards of the pedagogical design were compared in order to identify whether there is a deficiency in the quality status of the pedagogical design of the ICT/CS education in the university system of Sri Lanka. In order to identify the current quality status of the local context, a preliminary survey was carried out with 297 student sample. In addition a selected sample of university lecturers were interviewed and class room observation was carried out by the research group. The existing quality gap was driven using gap analysis. In phase II, best possible pedagogical techniques were organized to formulate a model to enhance the skill development of the selected sample of ICT/CS university students, in order to minimize the specified gap in the local context in a sustainable way. Though this study mainly targeted for developing the skills of students, knowledge and attitude development of students also were achieved up to a certain level. Thereby this study involves Knowledge, Skills and v | P a g e Attitudes (KSA) development of ICT/CS students. To drive and test the model large group practical oriented classroom was chosen as the main target since most of the ICT/CS undergraduate courses are practical oriented and consists of large groups. Experimental approach was chosen as the approach to reach the target of the phase II. By incorporating the following pedagogical techniques; critical thinking exercises, activity based learning, collaborative learning, problem-based learning, peer-led team learning, journal writing, research oriented learning techniques and reflective learning, this experiment steered a series of continuous practical based assignments other than the lectures which was conducted using some selected pedagogies within the limited time. Moreover the Bloom‘s taxonomy was used to achieve the leaning outcomes. In the face-to-face lectures, a certain level of learning objectives was achieved due to the time and environmental constraints. The remaining learning outcomes were achieved through the practical sessions. Further, with valid testing, this study uses reflective practice theories, feedback of students and peer lecturers in a cyclic way to fine-tune the proposed model after the each experiment. Not only the quantitative but also the qualitative measures were taken in order to derive correct and accurate results from the evaluation. Ability to test students‘ own knowledge individually, improving the presentation skills, allowing them to learn their own are positive feedback received from the students and the peer lecturers. Results of the experiment I showed that expected skill development can‘t be obtained in the theoretical class and therefore applying the initial pedagogical model in the theoretical class is relatively low. According to the descriptive statistics of results, in the experiment II the average of the results has been increased to 73% from 51.20% and the standard variation has been reduced to 13.11% from 19.18%. In the experiment III the average has been increased to 68.5% from 58.36% and the standard variation has been reduced to 17.60% from 19.76%. According to the inferential statistics of the results, the Wilcoxon signed rank test showed that, the proposed pedagogical model elicited a statistically significant change (p< 0.05), which is positive, in students‘ results of assignments which were conducted in experiment II and experiment III. Furthermore the findings of the study showed that carrying out the practical sessions as assignments is a motivation factor to the students. At the end, this study revealed that the skill development of each individual can be improved through this pedagogical model which saturated with active learning techniques.

Recognizing the level of alcohol intoxication in Sri Lankan males through changes in voice suprasegmentals and reaction time

Wed, 13 May 2015 00:00:00 GMT

Title: Recognizing the level of alcohol intoxication in Sri Lankan males through changes in voice suprasegmentals and reaction time Authors: Abhayarathne, S.J.; Wakista, G.W.; Mendis, D.G.T. Abstract: Alcohol affects driving by disrupting the communication of the central nervous system and there by inhibiting driver’s physical and cognitive capabilities. The current methods used to identify drunk drivers in Sri Lanka are plagued with deficiencies. Blood alcohol testing is not routinely available due to its cost and the breathalyzer balloon does not have any reproducibility or reviewability of test samples. The focus of this research is to minimize the inadequacies present in the current process and create a scientifically valid, quantifiable, objective and economical alcohol detection framework which can assist the Sri Lankan Police, Judicial Medical Officers, courts and general public. Since motor co-ordination is impaired under intoxication, it affects an individual’s speech production and reaction time. One of the common questions which arise is; whether is it possible to detect whether a person is intoxicated by observing their speech patterns and reaction time if so can it be used to determine the degree of intoxication? To test this hypothesis, healthy male native Sinhala speakers were carefully sorted out. Speech recordings, reaction time (based on a mobile game) and associated Breath Alcohol Concentration levels were taken under sober and intoxicated condition. The suprasegmental features of the audio recordings were analyzed using Praat. Several statistically significant changes were found for increasing intoxication; primary results included increased Speech Duration, Pitch, Intensity, Shimmer and Degree of Voice Breaks, while Fundamental Frequency and Harmonics to Noise Ratio were decreased. Formant, Voiced to unvoiced ratio and Jitter appeared to fluctuate towards both end under various levels of alcohol intoxication. Reaction time generally increased under alcohol intoxication but in some instances it decreased. Then after, extracted voice features were processed and imported into the proposed training model. In the evaluation phase, 66.7% of accuracy level was obtained through the trained model. A Regression Analysis was conducted to predict BrAC value using Reaction Time. However, due to the fact that Reaction Time data deviates from normal distribution, we were unable to build a statistically valid mathematical model to predict the alcohol intoxication levels. Additionally, through our research, we found that by using a mobile game based approach, an additional variable ‘practice’ gets included and in return it affects the reaction time. There has only been handful of scientific research conducted on the behavior of suprasegmental features and reaction time under intoxication. Hence, our research findings will be a valuable contribution to the scientific community. Our project is the first scientific research that is conducted on Sinhala language to identify the behavior of suprasegmental features under intoxication. Additionally, by using the data collected through research, we have built an Alcohol Language Corpus for Sinhala Language so that future researchers can further research on this area.

Modelling and Forecasting Tourism Demand for Sri Lanka

Wed, 13 May 2015 00:00:00 GMT

Title: Modelling and Forecasting Tourism Demand for Sri Lanka Authors: Kodituwakku, W.H.; Wijesundara, W.M.D.M.; Hettiarachchi, C. Abstract: Tourism makes a substantial contribution to Sri Lanka‟s economy through generation of employment opportunities and foreign exchange earnings. During the past thirty years up to 2009 Sri Lanka‟s tourism has had many obstacles mainly due to the unstable security situation that prevailed. Additionally, the Tsunami disaster in 2004 and the world economic crisis started in 2008 also had adverse effects on the industry. At present, these holdups have either been resolved or absent. The improved political and security situation coupled with indefatigable efforts for post-war economic and infrastructure development has increased the attraction of international tourists towards Sri Lanka. Currently the industry has re-entered to a growth track and is poised to reach its full potential as a safe tourism destination. Accurate forecasts of tourist arrivals are of critical importance to the tourism industry. It would help to ensure the availability of required infrastructure and services when demand materialises. Therefore demand forecasting has become a very interesting topic in tourism research. This study aims to forecast tourist arrivals to Sri Lanka by using quantitative methods. It qualifies past information about a phenomenon by applying mathematical rules which take advantage of the underlying patterns and relationships in the data. To achieve this, monthly tourist arrival data from January, 2010 to October, 2014 are used to build models and evaluate the forecasting performance. The reason for selection and confinement to this duration is unique growth trend observed in tourist arrivals after conclusion of civil war in May 2009, and the absence of factors which contributed to the downward trend before 2009. From the analysis it appears that forecasts from Holt Winters Multiplicative Seasonal Model always outperforms those from Holt-Winters Linear Exponential Smoothing Model, Multiple Regression Model, Vector Auto Regression Model, SARIMA Model and Neural Network Models (i.e. Elman, Feedforward Backpropagation and NARX) in terms of out-of-sample performance used (for Total tourist arrivals, India, United Kingdom and Maldives) and most of the accuracy measures. But for France, SARIMA model generated the most accurate forecasts and for China and Germany, Holt Winters Additive Seasonal Method showed more accurate forecasts. Forecasting accuracy of the neural networks was relatively low due to the limitations of the available data set. But when such limitations are removed the use of neural network approach for forecasting may yield better results because neural networks are capable of representing knowledge based on massive parallel processing and pattern recognition based on past experience and are expected to be superior to statistical methods in forecasting. When the availability of data increases, forecasting accuracy will be higher. Since the ultimate goal of this study is to support information needs of stakeholders in Sri Lankan tourism industry by providing information about future flow of the tourist arrivals, a prototype web-based system was implemented. It provides forecasts for total tourist arrivals and six country-specific arrivals. The system also includes experts‟ opinions (as articles) about those forecasts in order to make such forecasts more reliable via judgmental adjustments.