Saturday, October 29, 2011

Let's talk about semantic analyzers

I see a few of these in my human web search job. They are very good. Being my blog let me take a quick pass of what I hink is a standard semantic analyzer.

Its major function is to take plain text and generate dense ontology maps, using key words. What are key words? First, they are sequences of letters, without spaces, that can be looked up in a standard dictionary. Second they may be text phrases having spaces that have special connotation. Third, a key word may be a date, a location, a relative time. The third groups admits of Built in formats. Then there is a group of small grammatical connector words, usually dispensed with. And so on, various categories of phrase conversion may be the first phase of analysis.  The semantic machine reduces the text, partially, and generates the ontology of keywords, partially matching the order in which they appear.

In the Imagisoft system, I can imagine the system administrator having a slider knob that lets him set the density of output from the machine. Each piece of submitted corporate text will be inserted into the corporate ontology mesh, the raw material. How that corporate mesh evolves over time is set by the administrator with slider knobs. I would encourage employees o keep a bare bones blog in which to submit raw text.

No comments: