Graduate Student Talks - Ottawa Room (2nd Floor)
3:00 - 3:50 PM Saturday October 22nd
​

Approximate Time	Session Title	Speaker
3:00	Thoth: Experiment Reproducibility Using Provenance Data processing makes up a large portion of the time and effort that goes into computational experimentation. While there are tools available to maintain experiment code, much of the data preprocessing steps still go unrecorded. These steps are crucial to determining the final experimental results and without an accurate log of how the data was processed before the experiment, results can be difficult to explain and reproduce. Manually recording the data preprocessing requires significant researcher effort and is error prone. Thoth uses system-level provenance collection to automatically track the data processing workflow. Researchers will be able to use the Thoth to improve the explainability and reproducibility of their computational experiments.	Nichole Boufford University of British Columbia
3:10	Guilty Until Proven Innocent: Student Perceptions of e-Proctoring Software The use of e-proctoring software by universities across North America has been highly criticized in popular media due to their intrusive nature, and privacy-compromising features. Through qualitative interviews with 14 university students, we explored perceptions of e-proctoring software through the lens of their personal experiences and concerns. We share our analysis of student experiences using different types of e-proctoring techniques, from software-based monitoring  and tracking to remote video-based invigilation by human proctors.  Our preliminary findings showed that students were most concerned with misflagging for academic integrity issues or experiencing technical issues. Privacy was less of a concern, which may potentially be connected to their limited awareness of how online proctoring systems are collecting and storing their data rather than acceptance of the practices. Interestingly, students’ mental model relating to the software’s capabilities were largely based on assumptions. Overall, our results establish a link between students’ perceptions of e-proctoring, stress, technological concerns, and privacy.	Kazma Chaudhry Carleton University ​
3:20	Q-Learning Based Routing in Optical Networks The rapid increase in bandwidth demand has driven the development of flexible, efficient, and scalable optical networks. One of the technologies that allows for much more flexible resource utilization is Elastic Optical Network. However, there is a need to solve the Routing, Modulation and Spectrum Assignment (RMSA) problem. In my research, we use reinforcement learning to improve the efficiency of the routing algorithm. More specifically, we implement an off-policy Q-learning and compare it with the state-of-the-art algorithms. The results confirm that Q-learning is highly effective when optimal results need to be found in a large search space.	Sommer Harris Northeastern University ​
3:30	Exploration of glosses for gaze enhanced language learning applications Reading can be a daunting task for language learners, especially if the text involves a large amount of vocabulary that is beyond the reader’s scope of knowledge. An important aspect of language learning is reading texts written in their second language (L2). Moreover, the majority of available language learning tools target beginners, and few exist to support intermediate and advanced learners. A key challenge in creating tools to help intermediate and advanced learners is catering to individual needs, since people use varying sources to facilitate their learning, leading to disparate vocabularies and grammatical knowledge sets. In this project, we wish to explore ways an eye-gaze aware system can be used to create a personalized reading experience for learners. We are developing a gaze informed application that has the ability to augment articles with information about the text prior to reading, providing overlaid captions – or glosses – in-text, additional help in the sidebar, and a glossary at the end of the text. One of the goals of this project is to compare the performance of providing glosses and help to learners at different points during reading (i.e. before, during, after). A secondary goal is to test how gaze can be used to inform a system about what a user is interested in to personalize the system, and how gaze interactions can be used to control on screen content without the use of a mouse.​	Zixin Zhao Ontario Tech University ​
3:40	Improving The Usability of SWfMSs Using Group Discussion: A Case Study onGalaxy ​User forums have become a safe space to gather information, research and solve problems. Members share their constructive criticism as well as praise significant and noteworthy discoveries. Users post the problems that they face while using the system or performing their experiments on the system. Through the constructive analysis of those user posts, we can figure out the occurrence of any certain issue a system has, understand the difficulty level of using the system and also measure what varieties of problems users face regularly. When a system cannot fulfil the requirements of the users needs, it loses its users and gradually the system becomes a obsolete one. So monitoring the users’ needs, maintaining the system’s usability and fulfilling the users’ requirements are very crucial. User forums plays a remarkable role in this case as we can gather insights from the user posts. Thus, in our study, we have analyzed the user forum provided by Galaxy, which is one of the most popular scientific workflow management systems. Millions of users utilize this web-based computational platform to conduct their scientific analysis. The Galaxy users post their queries in the user forum on the domain of bioinformatics. However, those posts contain tags for categorization and we found that not all of the posts have tags. Moreover, if tags are found, they are not detailed and appropriate to the context. So primarily, in this study, we try to suggest tags based on the context by proposing a tag suggestion method and we perform a user study to evaluate our method. The findings from our user study show that our suggested tags are more relevant than the posted tags and also the suggested tags can properly represent the topic where users did not post any tags. We also found that SWfMSs are domain specific, so without prior knowledge, it is very difficult to search and find the appropriate workflow. That is why to help the users who are very new in their domains, we introduce a Natural Language Processing-based workflow recommendation system where anyone can write their queries using natural language and our system can recommend the users with the most relevant workflows in return. Our study findings can guide the Galaxy community to improve their features and extend the services according to the users' requirements. Moreover, the other SWfMSs can look into the findings to improve the usability of their services and take proper initiatives.	Shamse Tasnim Cynthia University of Saskatchewan