The revolutionary social visualization website Many Eyes has just a released a new visualization technique, called Phrase Net [manyeyes.alphaworks.ibm.com]. Phrase Net is based on simple pattern matching of words or phrases (advanced users can use also "regular expressions") within text, to detect the relationships between concepts hidden in books, speeches, and other unstructured documents. For instance, one can search for the term "and", and the resulting graph will show the "linking phrases" of the form "word1 and word2″. Each of these phrases defines a connection between two words. The resulting graph shows clusters of related concepts and people.
The size of each word shows how many times it occurred in a linking phrase. The arrows show the order of the occurrences. Darker blue words occurred typically as the first word in the linking phrase, while light blue words occurred in the second position. Other connectors give different views. For instance, the word "at" provides a network of places in the book. One can also input custom phrase patterns or use an asterisk to stand for a word.