Search and You Will Find – Maybe
Today, one of our technical people said that search was a commodity. That is probably valid. But I am hoping that executive management now sees that search, even if it is a commodity, needs to find what you are looking for and only what you are looking for.
It seems so simple. Maybe that’s part of the problem. People expect search in the enterprise to work in the same way it does on the Internet and in social applications. I hate to burst their bubble, but it doesn’t.
Management of information is a shambles – really. Did you know that poor information quality costs an organization 30 to 35 percent of its operating revenues? Now that’s a hefty bill for not managing information. Take a moment and figure it out. Done yet? The only action organizations take is to switch search engines. It’s time to take a step back, and think how can you improve what you have, not keep starting again.
The problem, and I am sorry to keep harping on about this, goes back to the quality of your metadata. Did you know that if end users are presented with a list, 90 percent of the time they will select the first option, regardless of whether or not it is correct? It’s no wonder organizations that use manual tagging can’t find anything.
Create a taxonomy and clean up your metadata. Not possible it? It would be quite a bit of work to go through every piece of unstructured and semi-structured content and retag it, now wouldn’t it? Our technology uses compound term processing. Don’t want to hear a vendor pitch? It isn’t. It’s explaining what we do that addresses the metadata issue.
Traditional search assumes the end user knows what they are looking for, or must enter the right combination of words to get the right result. In the example below, a search engine relying on keywords identifies all documents that contain the words ‘triple,’ ‘heart,’ and ‘bypass,’ instead of documents that contain the concept of ‘triple heart bypass.’ Through metadata, compound term processing associates that ‘triple heart bypass’ represents a concept, and returns documents that have those words together.
Since the software understands the meaning of the phrase, it will retrieve any document that contains similar content, such as heart surgery, heart attack, open heart surgery, and coronary artery bypass. It will return content based on the probability that it contains the information you are seeking, but your initial search string will be at the beginning of the list. Pretty smart.
We are the only statistical vendor with technology that can do this. OK, that last part was a vendor pitch. How do you tag documents for retrieval?