Tagging content

January, 2001

Original question
A Society member is setting up "knowledge centers" -- collections of documents on a particular topic, such as customer information or financial information. Authors, content owners, and Web site publishers are being asked to "tag" their documents by selecting terms from a "pick list" before publishing them to one of the knowledge centers.

Why tag documents?
Tagging is used to make it easier to find and access documents. Tags can be assigned to descriptive records (e.g. card catalog cards in a library) that are linked to documents, or they can be assigned to the documents themselves. 

A simple form of tagging occurs automatically in desktop programs like Microsoft Word. To see an example, open a Word document and select File/Properties. You'll see that the title, author, and company fields are already filled in. You can also fill in values for other tags, such as subject, category, and keywords. To see how tags are used to help you find documents in Word, go to File/Open and select the Advanced option. At the bottom of the Advanced screen, you can select a property, enter a value, and ask Word to find all matching documents.

The index at the back of a book is another form of tagging, although in this case the goal is not to find one document among many but instead to find a word or idea within a single document.

Tags are also common in databases, such as the address book files in e-mail programs such as Microsoft Outlook. At minimum, an e-mail address book contains a person's name and e-mail address, but it can also includes other descriptors, such as job title or department. In some applications, a person's department or access code is used to determine whether someone can view a document.

Who does the tagging?
Tags can be assigned by document authors, information professionals, editorial assistants, or even by computer programs. Artificial intelligence programs are the fastest and easiest -- but not always the most accurate -- method. Information professionals can produce highly accurate and effective tags that take into account all the nuances of language and subject matter -- but there are a limited supply of people with these skills. Getting authors to assign tags to their own documents should be both relatively inexpensive and accurate -- but they often lack the time, skills, and perspective to do a good job.

What members say
Here are some comments from members on this issue.

Management consulting firm
"A small percentage of our authors tag their documents, mainly because we have asked "experts" to do it. We didn't think the "masses" would actually do it. In fact, the experts are stretched too thin and don't do it judiciously either.

Those that take the tagging role seriously do it well. Those who don't leave lots of tags blank.

Training and incentives won't work for us. It has to be driven down as a management objective with negative consequences if it's not done. One of our groups has taken that approach, with excellent results. We are trying to cajole other managers to do the same."

Pharmaceutical company
"Our experience is based on the Documentum software.  Those who write documents that are entered into the system will, at the same time, add meta tags. In most cases, the secretaries enter both the documents and meta tags. However, in many departments, secretaries have developed a list of relevant meta tags. The authoring scientist indicates on the list those tags/values which the secretary should enter for the current document.

Tags were developed in consultation with users. We modified Documentum's metadata structure, adding our own tags. We do not use 'keywords,' since that structure is too broad. We have more specific tagging related to our industry.

We kept tags to a minimum as much as possible. However, we continue to receive complaints that there are too many tags. When we ask what can be deleted, the answer is always 'none, we need them all.' We haven't yet done an analysis of searching by tags. But the search engine is slow, especially when multiple tags are entered, so most people don't enter more than one tag when searching. I'm not sure how much valid information we would get with an analysis.

All content owners must enter tags. We removed any tags that were "optional." The majority of people won't enter anything more than they need to, so any optional tags are not reliable. All tags are mandatory, and the document won't 'issue' in the system without the tag being completed. This does not ensure accuracy, but the idea is that as long as something has to be entered it's as easy to enter the correct data as it is to enter incorrect data.

Also, whenever possible, we use pick lists rather than free text. For example, we have a pick list for document type (study report, general communication, etc.), discipline (toxicology, clinical research, manufacturing, etc.). Other tags are free text (title, author). Dates are also inserted easily.

There was a great deal of training when introducing the system as well as weekly user meetings. We now have monthly user meetings. When changes are made to the system, additional meetings are held, and we send updated information. Users were constantly asked for feedback and suggestions. Documents must be issued to be used in our regulatory submissions, and since they can't be issued without tags, participation is good. In some cases, however, we had to go directly to management to have use of the system made a performance issue.

We also did a lot of individual training and hand-holding. However, once the first group of people was trained, they tended to train others and became the resident experts. There is still a lot of discussion (6 years later) as to whether or not documents should be issued centrally in order to maintain quality.

The number of tags is related directly to the document type. There is a base group of tags including title, author and date. Study reports, depending on discipline, have more tags than for example a memo. A clinical research study report will have more subject tags than a chemistry report. Again, these tags were designed by talking to customers, both those who input the documents and those who use/retrieve them."

Technology consulting firm
"Our experience around tagging of content is that the contributor or author of the content is the best person to rely on for tagging -- no problem here as long as there are clear tagging standards or criteria that everyone understands. These tagging criteria are part of an asset submission form that everyone has to fill out when they contribute an asset [document]. As long as the submission form is well done -- it took us several revisions before it became simple enough, yet comprehensive -- then training was not necessary. On the other hand, it's problematic to ask someone to tag assets [documents] which he/she has not created. Furthermore, we have not had good experience with tagging something after the fact -- i.e. go back into the knowledge base to tag assets that have already been created some time ago. We have no specific incentives around tagging, but provide management-level incentives around contribution of knowledge assets -- senior managers are measured (bonus depends on it) on the percentage of their staff who contribute assets."

Management consulting firm
"When our company first began an online repository of our codified client work it was all volunteer based. That is, a "case team historian" was responsible for summarizing the work done and entering it into the system. The historian had a menu of terms to choose from that were company-specific as to study type, industry, etc. They could check off as many or as few as they wanted to. There was one editor who looked at submissions for general form but was not responsible for verifying tags. 

We found after several years of this procedure that the tagging was not reliable. People tended to choose too many or too few terms and frequently when they could not find terms that fit what they wanted they became frustrated and this discouraged them to contributing. We now have a team of several Knowledge Brokers who work closely with case teams to make sure their summaries and other documentation are codified for our KM platform. The team still picks the tags but the Knowledge Broker acts as a quality control filter before the information is entered into the system. If necessary they call the submitter back and verify or clarify what tags are the best match. This increases the quality of what is in the system, helps people to retrieve materials and helps to encourage contributions through increased credibility. 

Participation is always difficult, since it seems to be directly related to how much time someone has before they move on to their next project. It's typically very short. There are a number of incentives in place including dollars, contributions to personnel reviews, etc."

Technology consulting firm
"This reminds me of what I've heard about Xerox's Knowledge Pump. We don't do this at our company, but it's an excellent idea. It would probably be difficult to implement, but the results should make the effort very valuable."

Created on February 1, 2001 l Updated on November 19, 2012