Constructing Personal Knowledge Base: Automatic Key-Phrase Extraction from Multiple-Domain Web Pages

In the paper, we proposed a general framework that could automatically extract key-phrases from a collection of web pages concerning a specific topic with the help of The Free Dictionary and then construct a personal knowledge base. Both the base and visual feature in a web page are used to calculate the weight of each candidate phrase. The system extracts top p% key-phrases for each web page based on these two features and then generates a term set using union operators. Next, the system builds the relationships between terms in the term set by referencing The Free Dictionary, and then generates a list of terms sorted by weights. With the top q terms specified by users, a semantic graph can be constructed to present the part of a personal knowledge base, which shows the relationships between terms from the same domain. Finally, the experimental results show that the key-phrases generated by the proposed extractor are with good quality and acceptable for humans.