The Charades Syndrome. Remember the game Charades? (Movie, three words; first word, two syllables.) Watching people play it, particularly children, you can learn a great deal about how we communicate and about metadata, taxonomies and indexing. My first revelation was watching a nine-year old trying to communicate "Terminator." He formed his hand into a pretend pistol and mocked shooting at us. From his perspective this made perfect sense. Terminator made him think of shooting. A implied B. The problem for the rest of us was that hundreds of As implied the same B. He properly categorized (or applied metadata to) Terminator, but this was of little help as an information retrieval tool. It might have been useful had we wanted to run a query such as "list all films containing guns," but that wasn't the task. What's required is finding an A-implies-B where B is realtively unique. It requires understanding the information retrieval process and the uniqueness of the application. Observing lots of kids in our family, it appeared this was impossible for younger children. They just couldn't do anything except act out memories inspired by the movie rather than doing something that would cause us to think of the movie title.
Jump forward ten years when I received the draft of the Index for my first book, written by a professional Indexer. Oh m'God! What a disaster. Here are some examples:
Balance, vendor “sweet spot” and, 84
Just think about it. If you want to learn about vendors' sweet spots, are you going to look under Balance? Would a need for information about outsourcing cause you to start your search with Confidence? How about Unneeded to learn about services?
Confidence, in outsourcing, 7
Design, application and database, 193
Flexibility, vendor, 46
Dissatisfaction, customer 92
Unneeded services 96
So what's my point? It's that even professionals are often clueless when it comes to managing metadata for searching. While metadata can be a big help for topic-based searching (e.g., you already know the customer name, you just want to find the contract), it's almost useless for retrieving general knowledge. Most people tend to describe an item rather than place themselves in the mind of someone with a question for which the item may provide an answer. User-created metadata is notoriously useless for this reason among others such as inconsistency. Look at the metadata describing most documents and you see titles, subjects and descriptions. These are valuable for scanning by a search engine, but the full text is usually even better. Typical metadata only serves to help a user, upon identifying a document, determining whether the document will be of interest. But it's far less valuable for actually locating the document in the first place. [discuss]
Posted Tuesday, December 04, 2001 9:57:18 PM