Duplicate detection for billing systems

Merging different sub-companies into TELUS caused some of customer records to be repeated through the merged data-set. Algorithms are needed to determine the duplicate records. Currently a deterministic algorithm is being used in TELUS. In this project, we will investigate if machine learning can help to detect duplicates. Solving this problem has several parts. We have to preprocess the data and select some features from the TELUS records that help us in our model. A probabilistic model should be selected, implemented and tuned.

Biomarkers of Spinal Cord Injury

In this project, we will establish biomarkers that objectively reflect the severity of injury, measure its progression, and predict neurologic outcome after acute spinal cord injury (SCI). This will be accomplished by comprehensively analyzing blood and spinal fluid samples from acute SCI patients. In addition, we will conduct a parallel experimental study in a large animal model of SCI with a similar analysis of blood and spinal fluid samples.

Improving the calibration of a multi-camera system for accurate tool tracking

This research aims at improving the accuracy of a 3D-vision tracking system. The physical set-up consists of a tool to be tracked, such as a drill, with one or more planar patterns attached to it and a set of cameras. This set consists of one to four camera clusters, where each cluster has one or more cameras. The current tracking system consists of several modules, including one for the calibration of the cameras (intrinsic and extrinsic), and another one for the calculation of the 3D coordinates of an unknown physical point, the tip of the tool.

Conversational Response Generation with Linguistic Context and Emotional Context

The objective of this project is to develop methodologies for automatically generating responses in a natural language to converse with humans. Responses directly generated from the question-answer database are inflexible and cannot meet users' needs. On one hand, the responses should take into account the previous utterances that can keep a conversation more active. On the other hand, the responses should be appropriate for the emotions conveyed in a conversation that can make a conversion more user-friendly.

Activity recognition using physical layer information from wireless communications infrastructure

Sensing technologies require the deployment and maintenance of complex and large infrastructures. This research proposal is focused on people’s activity recognition technologies though existing WiFi infrastructures. The information gathered by this technology can be applied to different industries like home automation, security, etc. In the future, this technology will powered applications in the home automation industry as the one described next. Mary comes home and leaves her cellphone on the couch. As the system recognizes her, no alarm is activated.

Operational analysis and optimization of the delivery of HIV treatment and care in Vancouver

The continuum of HIV care is highly complex. It includes prevention, testing, patient care, treatment, and support services. This project will help Providence Health Care utilise its limited resources to provide the best treatment and care for people living with HIV in Vancouver. Care for HIV patients includes antiretroviral therapy, treatment of co-morbidities, monitoring clinical markers of disease progression (CD4 count and viral load), and support services to ensure treatment adherence and retention in care.

Computational Approaches for Characterizing Non-canonical Tandem Mass Spectra

Extensive research has been conducted for the computational analysis of mass spectrometry based proteomics data, however most of the traditional computational approaches take the assumption that the acquired spectra are generated from the fragmentation of a single precursor and the peptide is simply a linear sequence of amino acid residues. This ubiquitous assumption is impeding the utility of those computational approaches, especially when handling those non-canonical tandem mass spectra.

Network Traffic Profiling for generating intrusion detection evaluation datasets

Intrusion detection has attracted the attention of many researchers in identifying the ever-increasing issue of intrusive activities. In particular, anomaly detection has been the main focus of many researchers due to its potential in detecting novel attacks. However, its adoption to real-world applications has been hampered due to system complexity as these systems require a substantial amount of testing, evaluation, and tuning prior to deployment.


Often, a single employment notice may receive hundreds of applications. Manual inspection of applications is extremely time-consuming, and may be approximated by a computer program. Such a program would automatically extract a number of features from each application. For example, relevant work experience, skills, and qualifications might represent appropriate features. After extracting these features, the system would be able to score and rank applications in an effort to reduce the number of applications that would then need to be reviewed.

Maritime Domain Awareness: A Service-oriented Analytic Framework

Maritime situation analysis is critical for dynamic decision-making in responding to real-world situations. Rapidly unfolding situations that pose an imminent danger or threat to critical infrastructure or public safety require interactive decision-making to enable a swift response. The main objective of this project is to design a robust methodical framework for the development of intelligent systems and services for real-time anomaly detection in marine traffic, applied to large volume maritime surveillance operations.