CURRENT PROJECTS

CURRENT PROJECTS

Spoken conversational research

This project will research a new interaction paradigm for search engines, where all input and output is mediated via speech. While such information systems have been important for the visually impaired for many years, a renewed focus on speech is emerging driven by the ever growing sales of internet enabled smart phones. The phones
allow internet access in new contexts that require both hands- and eyes-free interaction; one example being searching for information while driving. Also, smart phones are being accessed by a new and large population of users across the world many of whom struggle with literacy; again requiring access mediated by speech. Currently, search systems poorly serve such a mode. Recent research showed that one cannot just ‘bolt on’ speech recognisers and screen readers to an existing system: a fundamental change to the way search is conducted is required.Our Project Aim then is to research a new framework for effective information retrieval over a speech-only channel: Spoken Conversational Search (SCS), which provides a conversational approach to determining user information needs, presenting results, and enabling search reformulation.
The project is funded by the Australian Research Council Linkage projects scheme (LP130100563) and is a joint collaboration between RMIT University and RealThing Entertainment Pty Ltd.

Job and Talent Search

This project plans to develop a new evaluation framework to understand and characterise web users and their situation within complex, multi-faceted search tasks, exemplified through job-search. While evaluation of web search engine effectiveness is relatively well understood, measuring information retrieval performance in the context of complex tasks with heterogeneous users is a largely neglected problem. This project plans to mine user-specific characteristics and situations from complex profiles and interaction logs for online information services run by the industry partner, SEEK. The new techniques are intended to redefine understanding of task-oriented search, and have the potential to reinvent the user experience for complex search tasks. This project will transform how practical search systems are measured within complex task scenarios. This will result in substantial economic impact by enabling businesses providing task-based search services to provide more customized offerings. Within the target domain (job search), this greatly enhances a service highly relevant to Australia’s productivity.The project is funded by the Australian Research Council Linkage projects scheme (LP150100252) and is a joint collaboration between RMIT University, SEEK Ltd. and The University of Melbourne.

More info: http://www.rmit-ir.org/index.php/research-grants/job-and-talent-search/

Data Analytics for Intelligent Route Planning in Transportation

GPS enabled mobile devices continuously provide new opportunities to improve our daily lives. This project aims to study a new kind of query- Reverse k Nearest Neighbor
Search over Trajectories (RkNNT), which can be used as a core operator for route planning and capacity prediction.

Funding partner: CSIRO Data61

Postdoctoral Research Agreement - Infoxchange - Data Analytics

CMCRC will fund a full-time postdoctoral researcher at RMIT University in the area of Data Analytics, under the direct supervision of Dr Jeffery Chan, in the School of Science (Computer Science and Information Technology)
Funding partner: Capital Markets Cooperative Research Centre

Adaptive Talent Search

This project covers the application and the testing of new/and/or existing tools and techniques for feature engineering and for retrieval and ranking to improve the search experience of the SEEK Talent Search Product.

Funding partner: Seek Limited

Visual Analytics of Geo-related Multidimensional Urban Data

The project involves research into techniques to enable human visualisation of Geo related data in a way that makes it possible for analysis and interpretation by the end user.
This project is a single PhD project involving collaboration between RMIT School of Science and Data61 through their Engineering and User Experience Design and Analytics groups.

Funding partner:CSIRO Data61

Efficient Storage, Search and Visualization of Trajectory from Social Media

The project involves research in data analytics and decision systems. It aims to derive machine learning algorithms to enable use of social media data from subjects to provide suggested information tailored to their needs based on location and prior search and data history.

Funding partner:CSIRO Data61

Capital Markets Cooperative Research Centre - Umbrella Record

Other Participant Agreement for RMIT as a participant of the Capital Markets CRC.

Funding partner: Capital Markets Cooperative Research Centre

Efficient and effective multi-stage retrieval in rust

The goal of this project is to create the next generation open source search engine in Rust. Current legacy systems such as Lucene / Elastic Search do not currently focus on end-to-end multi-stage retrieval, which requires both fast candidate generation, and efficient learning-to-rank. RMIT has a long history of research leadership in the Information Retrieval community, and we would like to help seed the next generation search engine that properly integrates efficiency and security (Rust) with new machine learning models. I believe this project could be compelling testbed for Rust in HPC computing that other academics and industry partners can leverage for a wide variety of different search tasks. There are currently no native Rust-based implementations of fundamental LTR algorithms such as Gradient Boosted Regression Trees or LambdaMart that I am aware of, and few projects leveraging the language for efficient indexing and search. I believe Rust has the potential to become the next big systems programming language, and projects such as this one can help build that momentum.

Funding partner: Mozilla

Continuous and summarised search over evolving heterogeneous data

The project aims to advance a new class of search engine that monitors and retrieves over multiple social media and micro-blogging platforms. The engine presents search results in a personalised continuously updated summary. Drawing on a team of world class national and international researchers, the project will create a new inverted index structure that systematically captures and provides easy access of all monitored content including text, linked and relational data. The structure will enable a novel continuous search paradigm that keeps track of the updates on both the content and popularity of search results. Results are presented as a dynamic summary that are personalised to the user’s social network.

Funding partner: Australian Research Council

Personalized search over the streaming social media data

Social networks like Google+, Facebook and Twitter are hubs for users to acquire up-to-date information on the Internet [6]. As well as information generated within social networks, external information (from the Internet) of higher quality and selectivity is shared by users in social networks and can be accessed by the URLs in users¿ microblogs [4]. This project will explore how to provide an effective search over highly-updated yet short-texted social media data.

Funding partner: Google

Magnitude Estimation for the Evaluation of Search Systems

This project will explore the application of magnitude estimation for the improved evaluation of information retrieval (IR) systems.

Funding partner: Google

User-Adaptive Search and Evaluation for Complex Information-Seeking Tasks

While evaluation of web search engine effectiveness is relatively well-understood, measuring information retrievalperformance in the context of complex tasks with heterogeneous users is a largely neglected problem. Thisproject will develop a new evaluation framework to understand and characterize users and their situation withincomplex, multi-faceted search tasks, exemplified through job-search. User-specific characteristics and situationswill be mined from complex profiles and interaction logs for online information services run by the industry partner,SEEK. The new techniques will redefine understanding of task-oriented search, and has the potential to reinventthe user experience for such complex search tasks.

Funding partners:Seek Limited, RMIT University, Australian Research Council, University of Melbourne