With the increasing trend of providing related information at one platform, by extracting it from different structured, semi-structured or even unstructured data sources, information extraction still have many challenging issues to deal with. This paper addresses the issue of university profiling, which means automatically building a semantic (based on lexical analysis) university profile, by identifying and annotating or labeling information from the web. Most of the profiling annotation tasks (e.g. job profile and resume extraction)were often undertaken in almost ad-hoc manner or a few efforts have been made by using sequential labeling approaches for information extraction. In this paper, we first give the formalization of said issue, and then a unified approach based on CRF (Conditional Random Fields) is proposed to perform all tasks by incorporating data dependencies. Experimental results show that, the unified approach has significantly outperformed rule based and classification based approaches, which have learnt independent rules and models for each task and cannot incorporate data dependencies.