Online recruitment services are among the most popular applications on the Internet. However, their usability is compromised by the information overload problem, as users must frequently search through hundreds or thousands of job advertisements for a given query. The CASPER project [1] seeks to enhance such services by providing users with more personalized information retrieval [3, 4] using JobFinder, an Irish recruitment site, as an application test bed.
The following scenario typifies the problems facing traditional information retrieval applications. A user specifies a query for a software development job paying a particular salary rate. However, the user is actually looking for a permanent C++ job with a salary rate exceeding the amount specified, located in Dublin, Ireland. Thus the query is incomplete, limiting the accuracy of the search results. Moreover, two users with the same query will receive the same results even though they have different preferences—a second user with the same incomplete query but looking for a LISP job in Cork, Ireland, will not want to see information on Dublin or C++ jobs.
CASPER’s solution is a two-stage search engine (see the accompanying figure) that selects job cases not just according to their similarity to the target query, but also according to their relevance to the specific user in question, based on that user’s interaction history [1]. During stage one, job cases are ranked according to their similarity to the query by using a standard similarity metric, which calculates a similarity score between query features and corresponding job case features. Each case is made up of a fixed set of features such as job type, location, and salary (see [1]). This server-side stage produces a ranked list of job cases according to their similarity to the target query (see [2]).
The second stage, a client-side process, emphasizes personalized information ordering. It reorders the results according to their relevance to the user by comparing each result to the user’s learned search profile. Each profile specifies the job cases the user has previously liked or disliked based on past feedback. Each stage-one result is associated with a relevance score by comparing it to the k most similar user profile cases. For instance, if the result is similar to many positive profile cases, it gets a high relevance score; whereas if it is similar to negative cases, it gets a low score.
Thus priority is given to jobs that are similar to the target query and relevant to the user. If a user has previously liked jobs in the Dublin area, the second retrieval stage will prioritize job cases from these locations in the future, even if the user does not specify a location constraint in a query.
Evaluation results for CASPER’s two-stage retrieval technique are positive when compared to a traditional single-stage retrieval technique.
Very briefly, evaluation results for CASPER’s two-stage retrieval technique are positive when compared to a traditional single-stage retrieval technique. In summary, CASPER is capable of generating results sets in which up to 70% of the cases are relevant to the user and query, compared to control results in which only 45% of the cases are relevant; see [1] for complete evaluation details.
In closing, it is worth noting that the benefits of the proposed approach described here are not just related to retrieval accuracy. Because user profiles are stored and manipulated on the client side rather than on the server side, an increased level of security is offered to end users due to the fact that their profiles are never exposed to the central server. Finally, the retrieval computation can be more efficiently distributed between server- and client-side processes.
Join the Discussion (0)
Become a Member or Sign In to Post a Comment