Semantic Browsing

An Algorithm to Surf the Meaning

Francesco Lentini

The Eloisa Project

Abstract

Every document and intelligent message contains a “secret map”. An equally intelligent algorithm can be used to decode this map and use it to tag, mark up, summarize, and generate hypertext links. This is called reading and interpreting text with a click, but I call it semantic browsing. The remarkable fact is that a machine does not possess a knowledge base: it contains neither vocabulary nor syntax rules. Despite this, It works on any kind of text, written in any language. Let us suppose that it is necessary to analyze a binary code signal originating from a remote region of cosmic space. All that we can see is a string of bits, or maybe a series of symbols (such as the Sumerian cuneiform script). To be certain that this is not merely a random series of symbols we require a meaning detector, such as a data-mining algorithm. Unfortunately, data-mining algorithms operate by applying the syntax rules of known languages. However, at the message-internal level, we do not even know whether any such rules exist, therefore we need a new kind of algorithm. My Semantic Browsing (SB) algorithm attempts to overcome the semantic barrier at the message-internal level, without applying syntactic or statistic rules. One way to understand this is to examine several messages where the meaning is already known. In the example, SB examines the famous "I have a dream" speech by Martin Luther King and then answers a little question: What is the most important information contained in this document? Answer: FREEDOM. SB then examines the Unabomber Manifesto and gives an answer to the same question: SYSTEM. SB next examines an Italian translation of the Iliad: ACHILLE. SB then examines this document itself: ALGORITHM. Finally, SB examines a document created by the computer and consisting entirely of random phrases. And the answer is: NONE. Today the big idea lies in allowing big servers to create and manage the hypertext links (semantic web). My little big idea involves assigning this role to a new kind of browser, powered by the SB algorithm and running on personal computers.