RMIT and University of Melbourne researchers have teamed up with SEEK in a three-year Australian Research Council Linkage project to apply their search engine and user analytics expertise to SEEK’s job-search system.
Now in its final phase, the project is showing significant results.
Project Lead, RMIT's Associate Professor Lawrence Cavedon, said the Seeksy query amendment system they had developed used text mining techniques, informed by analysis of user history and behaviours, to expand search terms in relevant ways for each query.
Seeksy implements the ability to cope with synonyms like “governess” and “nanny”, or to modify overly specific queries like “training train driver” into “train driver” in order to increase job matches.
Comparative A/B testing of the technique conducted by SEEK showed click-through rates increased four-fold, and actual job applications six-fold, for certain common queries that had proven hard to find relevant results for.
Cavedon said it was satisfying to know that the improvements had been based on evidence of actual user needs.
“Job search involves different behaviours and goals to standard web searching, so we started this project by exploring user behaviours on the site through log data analysis. This showed us where to focus on designing automated techniques that would really help people find what they were looking for more easily,” he said.
Meanwhile, researchers in the field of information retrieval are starting to take notice of the impressive results.
The team won best short paper at the latest International Conference on the Theory of Information Retrieval in China, and also presented their work in July at the top research conference in the field, the SIGIR Conference on Research and Development in Information Retrieval, held in the United States.
“We can proudly say we’re one of the strongest research groups on information retrieval systems in the world,” Cavedon said. “This is demonstrated not only by our performance at major conferences in the field, but also by the impact of collaborations with industry.”
Head of Product - Search Experience at SEEK, Dr Evelyn Balfe, said the partnership supported SEEK’s mission to constantly innovate and improve product offerings.
“The collaboration with RMIT and University of Melbourne helps us do that by giving us access to world class researchers as well as ensuring we’re kept abreast of the latest advancements in academic and industry research,” she said.
“It also gives us access to a great talent pool of PhDs and really smart early-career computer scientists and engineers."
The team includes leading search engine experts from RMIT’s new Centre for Information Discovery & Data Analytics, including Professor Mark Sanderson and Associate Professor Falk Scholer, along with junior postdoctoral staff and PhD students. Several PhD research graduates from the team have already gone on to work for SEEK.
The RMIT and University of Melbourne collaboration in search and information retrieval research spans more than three decades. It has generated hundreds of high-impact research papers, including seminal work in the area, and produced senior research and executive leaders at Microsoft, Google, eBay and IBM Research.
Story: Michael Quin