Thinking Out of the Meta-Box, Part 3

In the last two posts, I wrote about the disparity between user expectations and MPOW search, and about at least one approach to the issue (the Infomine metadata-scraper to add signficant words to item records). Today I’m going to lay the groundwork

Let’s start with the assumption that there is inherent value to a directory of very high quality Websites–a boutique portal on the Costco Web. MPOW does get plenty of use, and use has been consistently on the rise for the last decade, sometimes leaping 50% in usage from one quarter to the next. Our user surveys also indicate that we are valuable. Even if people didn’t vote with their feet, I would continue to believe in the potential of providing an expert-reviewed portal to the Web. (I also believe that users can provide their own expertise in MPOW through user rankings, but that is a feature for another discussion.)

I heartily believe in metadata. I do not believe in too many things. As an Episcopalian, I’m even skeptical about apostolic succession, though too laissez-faire to argue the point. But I do believe that attaching structured information to content is a good thing, and the more structure you provide content early on and the better you maintain it and add to it, the gooder it is. It’s like all the bottles in Alice in Wonderland, tagged with clear instructions, or Paddington Bear’s little label, “Please take care of this bear.” Metadata isn’t just a way to stuff content into directory databases; metadata fattens the content with guidance, context, and history.

Metadata is relevance insurance. In my talk to the mediarati this January at the Webcred conference, I made the point that for information, travel through time and space is a very lossy experience. Think about the shoebox of photographs most of us have somewhere in our houses, filled with quaint black-and-white pictures of babies none of us can recognize or old cars on mountains we cannot identify. When these photos were taken, they were rich with context and information, but when this information did not travel with these photos, they lost not only context but meaning. You write a speech, record a song, paint a picture; if you do not put a name on it, label it, scratch your initials in the lower right-hand corner, someone may someday treasure what you did but not with the same depth of meaning as if we knew where and when this object originated.

So MPOW has its values, not only for the selective quality of its information but also for the metadata we assign it. MPOW has a third value related to its browsing collection. Over thirteen years we have built a very high quality language for organizing our collection, which today’s hep meta-cats would call a folksonomy. Our strategery is simple but effective: we use public service librarians with a good feel for language to assign and maintain these labels. Recipes, not cookery; deer, not ungulates. (No language snobs need apply. If one is upset by the joy of common language, one must indulge one’s hoity-toitiness elsewhere.) I would like to tell you that search log analysis drives our selections, and it should, but we have never had good access to this data, so we go by faith and experience. I am pleased to report that early results from our usability tests conducted earlier this week–formal tests under the aegis of an expert, speaking of hoity-toity language–give an enthusiastic two thumbs up to the MPOW folksonomy; if the specialist had three thumbs, I think they’d all be hoisted, too, she was so full of praise.

But wait, there’s more! Our folksonomy both provides additional identification for the item–similar to the good intentions if poor execution of Library of Congress subject headings–but also functions as information about where the item stands within context of other items–its browsing, or shelf, location. (This is not really true with books, which lose the relationship between subject and call number as soon as more than one subject is assigned.) As is inherently true with Web directories, items in MPOW can be in multiple “shelf locations,” so shelf location is always faithful to intellectual assignment. (Is there a “mostness” we can assign to the location? Not now, but on the new site we will have to indicate the topic we used to publish an item. In other words, for publishing purposes, an item will appear under one topic when it debuts or is republished. I think the newsletter-assignment could eventually become a weighting element in the intellectual “mostness” of an item.)

So what can we do to leverage the power of our boutique content and structure–the metadata both within and without the item? How can we circumvent the mismatch between user search expectations and MPOW search functionality?

I have an idea–and that will be in the next post.

(N.b. Recently I was working with a programmer who was trying to present a database view of the items we had in each collection, and she kept saying “but it doesn’t ADD UP!” That’s right, if the goal is to have items match totals, it won’t add up. But the math works fine if you don’t think in terms of items but in terms of intellectual representation of topic areas. This reminded me of an incident several years ago, when I worked with a programmer who was convinced that somewhere we were holding out on the hidden location of the flat file that listed all the possible Library of Congress subject headings, despite our attempt to explain that subjects could be dynamically constructed and–at least by LC standards–were constantly evolving.)