Can software automatically recognise celebrities? Hewlett-Packard (motto: Let’s do Amazing) believes so, and has recently published a paper on the feasibility of ‘Wikipedia-based Online Celebrity Recognition’.
Researchers Demiao Lin, Jianming Jin, and Yuhong Xiong, from HP Labs China have been investigating so-called Smart-Browsing on the www, and point out that : “Obviously, recognizing celebrity names is important for smart browsing.”
Clearly though, in order to implement Smart Browsing for Celebrities, a celebrity database is required. Fortunately, Wikipedia hosts many thousands of articles regarding celebrities. In fact, the team explain : “Wikipedia is a ready-made source for building celebrity base.” But the data needs to be somehow extracted from millions of other non-celeb Wikipedia items. For this purpose the team created special software to auto-query Wikipedia and perform -

• Personal Metadata Extraction
• Name Extraction
• Birthday Extraction and
• Occupation Extraction

The software then carries out a Celebrity Score Calculation – based around the following formula :

And finally, the identified celebrities are highlighted (in green) as the user browses the www. Thus “Applications are enhanced by the celebrity recognition service.”

The full research paper can be read here :