Capstone Final Exam Schedule

 

Final Examination Schedule

PLEASE JOIN US AS THE FOLLOWING CANDIDATES PRESENT THEIR CULMINATING WORK.

Winter 2016

 

Tuesday, March 1st

Salma Bashar
Chair: Dr. William Erdly
Candidate: Master of Science in Computer Science & Software Engineering
3:30 PM; DISC 464
Tribal Education Network (TEN):  UI Design and Development of a Semantic Web Based E-Learning System

 

The goal of this project is to develop an e-learning system called Tribal Education Network (TEN) that intends to make the learning process more effective for Native American students. TEN uses both semantic web technologies and traditional web technologies in its infrastructure. Semantic web is about building systems that can enable intelligent agents to perform searches or link data based on context and meaning. TEN uses ontology based description of content, context and structure to form links between learning materials from tribes to academic courses and learning objectives. In this project, the author presents a design and implementation of a Course Builder that uses semantic web technology to enrich cultural content from tribes with annotations and link these cultural content to academic courses.

Wednesday, March 2nd

Jay Vyas
Chair: Dr. William Erdly
Candidate: Master of Science in Computer Science & Software Engineering

1:30 PM; DISC 464
Tribal Education Network (TEN) - Delivering Multimedia Content and Video Feature Extraction

The Tribal Education Network (TEN) project was initiated with the goal to assist Native American students in learning STEM-related content by using examples and approaches of relevance to their specific culture. Learning may be enhanced by including examples, stories, and experiences from within the tribal community scenarios into generic college course content.  This allows for a more tailored education that is more interesting and relevant to the student learner. My specific focus is to add multi-media extraction to identify and include specific culturally-relevant video to students that might be linked to course activities. The main functionality is an extraction tool and environment that provides related snippets of video clips for semantic tagging – and future use by instructors and students. The method uses frame-to-frame comparison (Twin comparison approach) for pulling these clips and annotating (adding context/meaning to) the main video content by linking it to a Cultural Learning Object (CLO) thereby providing culturally-specific, experience based learning. This is finally presented to the student via HTML5 video as a means for online content delivery.

Thursday, March 3rd

Subha Vasudevan
Chair: Dr. Hazel Asuncion
Candidate: Master of Science in Computer Science & Software Engineering

1:30 PM; DISC 464
Optimizing Provenance Reconstruction using Genetic Algorithm

The world has been generating a lot of data from the day internet was invented. There is no way of determining whether a piece of information is genuine without looking at it's provenance. Provenance can be defined as the history of an artifact. It traces the origin and lineage of a piece of information. If the source is trustworthy, the artifact is genuine. But until a decade ago, the importance of provenance was underestimated and ignored. So, a lot of data on the internet are void of provenance. My research is aimed at reconstructing the provenance of such data. I have introduced a Topic Modeling - Genetic Algorithm technique to reconstruct lost provenance.

Tuesday, March 8th

Wenbo Guo
Chair: Dr. Hazel Asuncion
Candidate: Master of Science in Computer Science & Software Engineering

1:00 PM; DISC 464
3-dimensional Visualization Tool of FACTS Project

During a software development lifecycle, a large amount of data has been generated. Among them, changes on source code and various types of artifacts help developers and product managers identify the rationale behind these changes, and improve the quality of their products. Usually, source code changes and artifacts are viewed separately by developers, and there is a view neatness issue of visualize them for a big software project. Here I present a hybrid visualization approach that combines treemap-like view and 3-dimensional view, to visualize changes made over the whole software project and traceability links to the related artifacts.

Pragya Upreti
Chair: Dr. Hazel Asuncion
Candidate: Master of Science in Computer Science & Software Engineering

3:30 PM; DISC 464
Analyzing Changes and Finding Reasons for Refactoring PAtteRns Changes (AC-PAR)

Changes are integral part of software evolution and different techniques are used to analyze these changes.  Changes can be of many types. Refactoring changes, does not alter external behavior of code and improves the structure of software, often performed to improve understandability, maintainability, and quality of software project. These changes are present in the commit log as simple line changes therefore finding them among numerous other changes is daunting task to perform.

This project takes the refactoring changes and connects them to specific reason as described by developer when performing those changes (commit statements). These connections are important as they fill the gap between the refactoring changes committed and future refactoring changes.

Aim of this file is to connect the statements with changes file, which represent official changes in the source repository. These statement show number of refactoring changes performed on files and type of refactoring changes done in single reason statement. In addition, summarize refactoring changes, so that it is easy to understand changes without going through all the data.

Wednesday, March 9th

James Yoder
Chair: Dr. Brent Lagesse
Candidate: Master of Science in Computer Science & Software Engineering

11:00 AM; DISC 464
Anonymous Voting in Collaborative Document Editing

This paper describes an implementation to allow anonymous voting by contributors to resolve conflicting commits in the context of real time collaborative document editing. By using this approach to reach a consensus among users, it is possible to overcome issues in collaborative editing such as social barriers to contribution, bias in review, and fairness in editing drafts of documents. The implementation of the algorithm applies stand-alone customizations to popular tools such as Git and the Atom text editor.  Considerations are given to user level issues such as the technical knowledge required by users and the amount of effort required to participate in order to reach consensus.  Additional considerations cover the technical challenges and approaches taken to conduct anonymous elections while at the same time providing a real time environment for collaborative text editing.

Thursday, March 10th

Naveen Mohan
Chair: Dr. Min Chen
Candidate: Master of Science in Computer Science & Software Engineering

10:00 AM; DISC 464
Voice Biometric Authentication for PELDA

Platform for Endangered Language Documentation and Analysis (PELDA) project aims to create a cloud-based, centralized platform for endangered language research and preservation. My focus in this project is to develop a voice-based login module for PELDA by using a two-phase authentication technique.

In the first phase, a data mining process is applied to capture each speaker’s voiceprint. Specifically, during the signup process, each user is required to read a key-phrase, which is recorded by the system and processed by a noise filtering component to improve signal quality. Then, its critical voice information represented by Mel Frequency Cepstral Coefficients (MFCC) is extracted and passed to K-Nearest Neighbor classifier to build the speaker classification model. This model will be applied to validate the authentication of a user during login.

The second phase is developed to further improve the system security. It requires a user to verbally respond to a specific security questions, which will be recognized by a speech recognition component and validated against the correct responses established during user’s signup.

Shalini Ramachandra
Chair: Dr. Min Chen
Candidate: Master of Science in Computer Science & Software Engineering

1:30 PM; DISC 464
Automatic Laughter Detection in Meetings

In this digital age, video and audio formats are being increasingly consumed. The goal of this project is to implement a system to detect laughter events. The audio recordings were obtained from the corpus of video data recorded by Professor David Socha for his software engineering research. Audio features relevant to laughter detection are extracted and k-Nearest Neighbors (kNN) is used to produce a training model. To address the challenge associated with the skewed data distribution (i.e., the number of non-laughter events being significantly more than that of laughter events in the recordings), balancing techniques were employed during the training process to improve the classification accuracy. Filtering process was applied to preprocess the testing data, on which the trained model is then used to locate laughter segments.

We believe that this research work can lead to various benefits including (but not limited to): (a) Ability to automatically identify laughter in comedy genre in audios; (b) Identification of semantically meaningful occurrences such as topic changes or ideas; (c) Identification of the speakers via their laughter; and (d) Detection of stress level of meetings.

Friday, March 11th

Aparajita Sahay
Chair: Dr. Min Chen
Candidate: Master of Science in Computer Science & Software Engineering

11:00 AM; DISC 464
Windows Phone based Plant Recognition Application Using Leaf Analysis

For inquisitive gardeners who wish to gain knowledge about their garden, gathering information about unknown plants manually is time-consuming. Through my capstone project, I have built a windows phone based plant identiļ¬cation system by analyzing visual information of leaves. Given a single leaf specimen on a white background, the application classifies the unseen leaf against a set of known data set and returns top three results to the user.

Leafs Identification process consists of multiple stages. First, the query image is pre-processed to remove unwanted noise. The second stage includes identifying interesting points and computing SIFT descriptors. Finally weighted K-Nearest Neighbor classifier compares features of query image with training data set and returns top three plant species as the result.

Rashmi Sandeep
Chair: Dr. Min Chen
Candidate: Master of Science in Computer Science & Software Engineering

1:30 PM; DISC 464
Unsupervised Feature Extraction for Data Mining of Endangered Language Audio Data

The primary goal of my Capstone project is to design and implement deep learning framework to extract useful features from the audio dataset in unsupervised manner.  The features extracted are then verified using data mining algorithms for sound classification and retrieval. The results from this project would then be compared against those from existing feature extraction frameworks.  This project would provide a framework for PELDA(Platform for Endangered Language Documentation & Analysis) researchers and users to apply and evaluate applicability of deep learning on endangered language recordings.  The system evaluation is based on its ability to increase precision and recall compared to other methods already implemented.

Monday, March 14th

Usha Basannappa
Chair: Dr. William Erdly
Candidate: Master of Science in Computer Science & Software Engineering

4:00 PM; DISC 464
EYE Platform - A platform for screening and treating Binocular vision dysfunction

A binocular vision impairment is any visual condition where in binocular visual skills like tracking, fusion, stereopsis, convergence, visual Motor Integration are inadequately developed.Research estimates that 25% of all the school children in the United States are affected by Binocular Vision Impairment. Reading and learning becomes a challenge for those kids with vision problems since both eyes are not teaming to work together properly causing them to drop out of school. Such children are frequently placed in special education programs.Hence it is very important to have vision screening to identify the children at the risk of vision problems and provide treatments. 

EYE platform is a full fledged software platform designed to help detect and treat binocular vision issues in children. This platform is a suite of various applications. First, an API interface for any therapy game to interface with and use the central data store in cloud to save game scores. Second, API interface for any presentation layer to retrieve and present data points, Third, A data store hosted on cloud to store patient details, therapy details and also record the therapy results. This will enable data sharing among medical providers, scientists and parents. Lastly, A Web App consisting of parent portal to manage and track the treatment of vision issues for their children and a provider portal to manage patients, assign therapies and view therapy progress.

Tuesday, March 15th

Ashik Samad
Chair: Dr. Min Chen
Candidate: Master of Science in Computer Science & Software Engineering

11:00 AM; DISC 464
Web-services for Endangered Languages Analysis

We aim to design and develop web services on a cloud platform to process endangered language recordings and extract audio features to facilitate effective language documentation and analysis. To achieve this goal, in my project, we deployed Praat audio analysis components into a cloud-based web application. Praat is a linguistics tool that has been widely used by researchers to analyze sound samples. However, Praat is a standalone application, and does not expose any libraries or services to be used in third-party applications. We deployed Praat audio analysis components as web services in the Azure cloud platform that can be integrated into any application that supports HTTP communication. We have also designed a web UI for general users to access Praat component via web browsers.

Haihong Luo
Chair: Dr. Hazel Asuncion
Candidate: Master of Science in Computer Science & Software Engineering

1:00 PM; DISC 464
FACTS - Policy Based Artifact Changes Description between Bugs and Codes

During software development lifecycle, changes on source code are frequently happened. The reason why codes changed is most related with requirements or bugs. Traceability between codes and JIRA issues can help user to do analysis, e.g. help verifying that a JIRA task has been implemented or not, the reason why these java files changed.

Policy Based Artifact Changes Description Tools captures JIRA issues using web scraper, generate links with codes (git commit) by different policies. It gives traceability between JIRA issues, codes with changed files. The policies are pluggable and easy for user to create a new policy with few changes.

Wednesday, March 16th

Jon Brammer
Chair: Dr. William Erdly
Candidate: Master of Science in Computer Science & Software Engineering

3:00 PM; DISC 464
AudioScreen: Unsighted Game Mechanics on Mobile Devices

Historically, most video games lean primarily on visual communication, making them inaccessible to the blind. Non-sighted gaming is beginning to gain some momentum, particularly with the proliferation of mobile devices and their abundance of sensor inputs allowing for inventive types of input controls. The arrival of built-in 3D audio support on iOS also represents a significant opportunity to easily explore audio-only gaming on the mobile platform. This study tests two variations on a paradigm for aurally representing a 2-dimensional screen in a virtual 3D audio world, as well as the interaction effects with three input control schemes and the presence or absence of a game mechanic called “auto-lock.” Users take on the simple task of finding and selecting stationary targets on the screen. The primary hypothesis is one variation will allow for faster and more accurate targeting than the other, but interaction with the control schemes and “auto-lock” mechanic may cause particular combinations to be more intuitive and/or enjoyable. This study will inform the further development of blind-accessible games and will provide quantitative feedback on the accuracy and speed that can be achieved with different configurations of this paradigm.

Thursday, March 17th

Pooja Shetty
Chair: Mark Kochanski
Candidate: Master of Science in Computer Science & Software Engineering

10:00 AM; DISC 464
Goods Next Door

"Goods next door" is an android mobile application for giving, lending and sharing items in the neighborhood.It provides a simple user  interface for searching items. It has a list view where the users can view the search results and has a map view which makes it easy  to locate the desired item. It allows users to post items for lending or giving away.When the requester requests for an item, a notification is sent to the owner. The Owners can as well see all the requests for their posts in their home page and can choose the recipient from the list. In case of lending items the owner can  prioritize requests as who would be the next recipient. The requester can also track the status of his requested item.This app creates a private network for sharing and loaning items among neighbors and community.

Questions? Please email Megan Jewell, CSS Graduate Advisor at mjewell@uw.edu