Privacy and security
Mobile computing and pervasive computing
Machine learning intelligent programming system
This research consists of four distinct modules that are independentlystructured building blocks of the system. These four distinct modules are equally critical to the final prototype system. Depending on applicant’s strength of research and development, each applicant can select up to two modules to work on. These modules are described below:
Module 1: User Study
User study is a dominant method in the domain of HumanComputer Interaction to evaluate system performance on realworld users. In this study, before development of the actual system, we perform user study using the online crowd to validate user needs of personal assistant for instant messaging. The research involves brainstorming user scenarios, generating storyboards, launching online study, and statistical analysis. Basic experience in web development and graphic design is needed. It would be a plus if
Module 2: Mobile System
As a core component of this research, the mobile system module consists of two parts: 1) collecting personal data, and 2) design for personal assistant. In the first part, we will develop the virtual sensors to various types of personal data (e.g., event, physical activity, documents, emails, and location, etc. ). In the second part, we will design the software architecture as well as the user interface of the personal assistant so that it can deliver relevant answers to users in an efficient and effective way.
Module 3: Building Knowledge Graph
Knowledge graph is a powerful tool to surface complex relationships between personal data points. In this research, we leverage graph database to link rich types of personal data entities on smartphones. The applicant will need to transcribe the crowdsourced user scenarios into knowledge graph. This knowledge graph can be considered to cover most popular user scenarios that are related to information needs in instant messages. Our working personal assistant can therefore easily query for subgraphs of this knowledge graph in each scenario.
Module 4: Natural Language Processing
In order to be qualified to work on this module, the applicant should be familiar with common tools and algorithms of natural language processing. For each incoming message, we should be able to transcribe it into structural data and map it to a query on the constructed knowledge graph.