Enhanced Content-Based Similarity Detection for Book Recommendation

Recommendations is one of the main ways Kobo users discover content on the platform. By using purchase history, Kobo can suggest other books similar to a certain item. However, this does not provide meaningfulrecommendations in some cases, especially for bestsellers and fiction books. Currently, only for books that have no purchase history does Kobo supply recommendations based on text.

Development of an at-home diagnostic test for COVID-19 using a multi-layered tablet

The intern will be responsible for developing an at-home test within a tablet/pill. Tablets are a familiar house-hold product, that would be intuitive to use and handle. The tablet would automatically break apart the virus, and detect viral genetic material and result in a blue colour. To use the at-home test, the patient would place sample/swab in a reaction tube, place the tablet and add water. After 30 minutes, the patient can take a picture of the solution and upload the result to a database with a barcode identification unique to the patient.

Indoor Mapping based on recorded videos

Knowledge of indoor spatial information is vital to stores, warehouses, industries, and homes alike. It is used to optimize layouts to achieve easier navigation for humans, machines, and autonomous robots. Maps provide limited data about the specific placement of objects in the environment and inferring information about the physical space can be impossible.

Short Text Similarity Calculation and Related Question Recommendation in Customer Service Chatbots

We build up chatbots for commercial companies to serve their needs, such as customer services. Within the whole chatbot building platform, there is one core component which is the short text similarity calculation component. We would like to improve our calculation capability for matching similar questions, as well as recommend related questions for the customers while they are chatting with the customer service agents.

The design and implementation of a gamified eHealth movement and mindfulness solution for school-aged children

The online delivery of primary school curriculums may work well for subjects like math and science, but not so well for physical education. Now more than ever, it is crucial that we ensure that children continue to benefit from the countless positive mental and physical health outcomes associated with regular involvement in physical activity.

SOTI SNAP SDK For creating third-party widget

This project dedicates to leverage the use of SOTI snap app to a greater community by providing more flexibility for third-party users and reducing required prior programming knowledge. To do this, we will focus on building a Software Development Kit (SDK) so that third parties can create and publish new widgets allowing SNAP apps to be quickly created that support a wide range of business needs. Human-Computer Interaction and Software architecture principles will be employed to make the SDK more user-friendly.

SOTI SNAP Blockly – Building Apps Without Programming

SOTI SNAP is an application development platform that allows users to create applications with little or no programming knowledge. By utilizing a block-based approach, users can drag and drop blocks and pre-made widgets and connect together to create applications in minutes. Apps made using SOTI SNAP can run on both Android- and iOS-based devices. The aim of this project would be to improve upon the existing SOTI SNAP platform to make it easier for users to tinker with and learn.

Spoken Language Identification for Children

While taking foreign language tests, people may respond in languages other than the expected one. Typical scoring systems are trained only on the expected language, so unexpected language responses can have unusual results in speech recognition and scoring. Pearson would like to develop a more robust system for the automated speech recognition machine to know up front if the response contains non-target language content. Common language labels are English, Spanish, Chinese, Japanese, etc. Audio files are typically from 5 to 90 seconds long.

Learning to Rank through User Interest Mining: Towards Search Personalization

In online shopping, search results often have inherent ambiguity. Two customers using the same term as search query might have completely different expectations of the displayed results. For example, when the users type in the query “headphone”, some of them might expect over-ear headphone with passive noise isolation, while others might expect in-ear headphone with better portability. This project aims to extract users’ interests or preferences and understand what they want.