Graduate Student Research Talks -- Ottawa Room (2nd floor)

​Approximate Time	​Session Title	​Speaker
10:00	Real-time Emotion Recognition In this work, we have implemented a real-time emotion recognition system to generate emotion sensitive responses to utterances. The main focus of this implementation is to monitor the emotional state of the user.  The recognition system is developed using a text-based, audio-based and a combination of boths the models to identify the most effective approach.  As a proof of concept, the methodology is implemented on the Sentrybot robot which is an autonomous robot intended for deployment in the security domain.​	Deeksha Chandola York University ​
10:10	Detecting Abnormal Weather Patterns using the Local Outlier Factor algorithm Year by year, we can see some abnormal weather during all the year in Vancouver, BC.  For example, in summer it can be extremely hot or unexpectedly cold which can lead to natural disasters and economic issues in the province and country. The main purpose of this research project is to investigate the number and characteristics of the outliers that we can find in the datasets about the daily climate data in Vancouver, BC provided by the Canadian government for the period 1940-2021.  For the conduction of the research and finding outliers by the Local Outlier Factor algorithm and building the graphs with results, I was using the free open-source Data Mining Software Orange.  Orange is a comprehensive, component-based software suite for machine learning and data mining, developed at Bioinformatics Laboratory, Faculty of Computer and Information Science, University of Ljubljana, Slovenia, together with an open source community. As a result of the project, I show graphs that clearly show the climate change issue and the weather patterns in Vancouver, BC.	Elina  Kuznetsova NYIT Vancouver
10:20	Neural Inertial Localization We propose the inertial localization problem, the task of estimating the absolute location from a sequence of inertial sensor measurements. This is an exciting and unexplored area of indoor localization research, where we present a rich dataset with 53 hours of inertial sensor data and the associated ground truth locations. We developed a solution, dubbed neural inertial localization NILoc which 1) uses a neural inertial navigation technique to turn inertial sensor history into a sequence of velocity vectors; then 2) employs a transformer-based neural architecture to find the device location from the sequence of velocities. We only use an IMU sensor, which is energy efficient and privacy-preserving compared to WiFi, cameras, and other data sources. Our approach is significantly faster and achieves competitive results even compared with state-of-the-art methods that require a floor plan.	Sachini Herath Simon Fraser University​ ​
10:30	Causal Investigation of Auto-tuning Tools for Spark (Causal Tune) As data analysis is becoming an integral part of every industry, more and more developers are using platforms like Spark to run their applications faster. However, their goal of running their applications as fast as possible might not be achieved if they don't choose a good set of values for configuration parameters. Choosing the optimum values for the configuration parameters is a demanding task since we have a large search space for configurations to search. Moreover, the change in each configuration parameter's value might have a completely different effect on the execution time of different applications. Hence, the effect of each parameter on each application should be investigated. As a result, there is a growing need for auto-tuning tools. The problem of building an auto-tuning tool has two dimensions: 1) the number of experiments needed to find the optimal configuration values and 2) the generalizability of auto-tuning tools and being able to perform better on unseen conditions using previous data. In this work, we try to build an auto-tuning tool that considers runtime system performance metrics to improve the modeling of the application. Moreover, We focus on making our tool more generalizable by considering dynamics in the system that might affect application execution time by leveraging causal inference techniques.	Sadaf Sadeghian University of British Columbia   ​
10:40	Exploring Relevant Artifacts of Release Notes:  The Practitioner's Perspectives A software release note is one of the essential documents in the software development life cycle. The software release contains a set of information, e.g., bug fixes and security fixes. Release notes are used in different phases, e.g., requirement engineering, software testing and release management. Different types of practitioners (e.g., project managers and clients) get benefited from the release notes to understand the overview of the latest release. As a result, several studies have been done about release notes production and usage in practice. However, two significant problems (e.g., duplication and inconsistency in release notes contents) exist in producing well-written & well-structured release notes and organizing appropriate information regarding different targeted users’ needs. For that reason, practitioners face difficulties in writing and reading the release notes using existing tools. To mitigate these problems, we execute two different studies in our paper. First, we execute an exploratory study by analyzing 3,347 release notes of 21 GitHub repositories to understand the documented contents of the release notes. As a result, we find relevant key artifacts, e.g., issues (29%), pull-requests (32%), commits (19%), and common vulnerabilities and exposures (CVE) issues (6%) in the release note contents. Second, we conduct a survey study with 32 professionals to understand the key information that is included in release notes regarding users’ roles. For example, project managers are more interested in learning about new features than less critical bug fixes. Our study can guide future research directions to help practitioners produce the release notes with relevant content and improve the documentation quality.	Sristy Sumana Nath University of Saskatchewan