Purpose: This paper seeks to describe a personal recommendation service (PRS) involving an innovative hybrid recommendation method suitable for deployment in a large-scale multimedia user environment. Design/methodology/approach: The proposed hybrid method partitions content and user into segments and executes association rule mining, collaborative filtering, and contents popularity algorithms over various combinations of content partitions and user groups. The process results in recommended content for end-users based on the linear combination of candidate data sets. Findings: This study reveals that: the use of usage frequency is an effective way to analyse user's behaviour patterns and their selection of content; the partitioning of content and users into meaningful groups and the identification of optimal parameter values of constituent recommendation methods, yields successful results in the implementation; the hybrid method performs better than any constituent methods in most evaluation metrics. Practical implications: The PRS system serves as a useful reference for electronic libraries or information centres considering the development of personalised information services. Originality/value: The PRS system is designed and implemented to work efficiently in the large-scale multimedia user environment. It can also be applied to small and medium-scale environments or mobile platforms.