I´ve worked as a media analyst categorizing press clippings. Except that it´s a bore, it´s actually also a truly interesting theoretical problem that arises when you do that. It´s very hard to categorize for instance a news story into the usual categories that are interesting to a marketer och PR consultant, namely topic and sentiment. The good things with beeing able to count something (quantitative research) is that it makes large amounts of data accessible for interpretation. The bad thing with it is that if there is no meaningful question that is answered OUTSIDE the original questions (such as what category a text belongs to) it is use-less and a waste of good peoples time.
Example: A brand wants to know if it´s positively mentioned among bloggers. You count how many that have mentioned them and then you categorize those mentions into positive or negative. The good people at the brand are looking at data that cannot be USED. Usefulness requires knowledge about a third factor (thanx Johan Ronnestam) that can be changed in order to achieve something. In this case create positive mentions among bloggers. For me the most interesting third factor that can ever be put into such research is pychographics or, in common lingo, descriptions of the psychological make-up of people and audiences. Or the noble art of trying to figure out what goes on inside our heads by watching what comes out of it. Words for instance.
Computers are the best analysts – people are the best synthesists!
Statistical analysis, such as what persons or brands are mentioned in the text and how often, is easily done. Especially with computers that don´t get tired in their eyes and start thinking about what to do on friday. Categorization on the other hand requires interpretation – and that is a complex task that computers are lousy at and humans are, at best, good at subjectively, but not objectively. The trouble is that two different persons often do two different interpretations of the same text.
pre-defined is productive – emergent is exciting!
On a deeper level the problem is about pre-defined categories versus emergent categories. A text might fall into several pre-defined categories. And while reading the text the person reading might come up with new fitting categories for it. Those two mechanism work simultainously and independent of each other. The result is, amongst others, that media monitoring companies that let people categorize the texts have a hard time keeping down both the time consumed on pondering about what categories it fits into and keep the analys from inventing new categories instead of reducing categories ath the expense of more complexity.
A computer in its right mind would be nice
So, the thing to do, has always seemed to be to let computers do the reading and interpretation of the texts. The digitalization of media has made that possible and more and more relevant as the news media distribute more and more of their original content in digital form. But, computers are fast, but still really stupid. They manage well what we humans can do with our left hemisphere; counting and sorting for example. But they are still almost use-less at what we do with tremendous ease and speed with our right hemisphere; find patterns of similarity and subtle harmonics such as over-all tonality of a text.
The signifier and the signified – hello postmodern schooling!
The problem can be divided into two separate stages of interpretation – the denotative level or the text itself and the connotative level or the meaning that a person interpret out of a text. Connotative analysis of a text is an abstract process and dennotative analysis is concrete. A person are able to do both really good; for instance see a drawing of a rose, state that on the dennotative level it signifies the flower named rose amd on the connotative level it symbols love. A computer is so-far un-imaginative and roams the realms of dennotative analysis where A = A and not the connotative level where A could be or is similar to B. There is a lot of reserach going on all over the world and I believe that we´re are not very far from having computers that actually are able to work like our whole brain and not just the left side of it.
Cool research I´d like to do with the Typealyzer
However, I believe that there might be interesting loop-holes that can be used in order to combine dennotative analysis and connotative analysis. It,however, requires some sort of a bridge between them that a computer can learn to follow. The idea is pretty simple and works especially well today when Internet has made it possible to determine connections between readers and content through linking and mentioning.
a) Analyze digital media content on a dennotative level (pre-defined categories or word frequency alone)
b) Analyze Myers-Briggs personality type of bloggers linking to that sample of media content
c) Translate Myers-Briggs type, via Jung to Archetypes which are strictly symbolic interpretations.
If there different Myers-Briggs links to different media content in a statistically significant way (chi2) it is possible to build categories based on word frequencies in the third category Archypes that might alllow a computer to analyze a never before seen text into it´s symbolic meaning (archetype). By just counting the most frequent words used in the 16 type-inking categories creates 16 new categories consisting of word frequency lists that can be used by computerized text classifyers. The next step is to assign those 16 new categories to relevant archetypes.
I believe it would work like a charm!
I´ve seen that there is indeed so that different Myers-Briggs types links to different media content whan analyzing swedish blogs and swedish online new media. And there is no shortage of data laying around on the Internet do multiple testing of different samples. The missing piece of the puzzle is to relate the Myers-Briggs personality types to relevant Archetypes. For a marketer it would be very usful to be able to categorize into the archetypes described by Margaret Mark and Carol S. Pearson in their eminent book and brand management The Hero and the Outlaw.
Is there someone who has done an Myers-Briggs and Marks-Pearson translation?
There are some work already done that might help. Here´s a post that links “community” archetypes to Myers-Briggs which gives an entrance into the interprative science of comparing archetype descriptions.
This is one type of reserach I think the Typealyzer can be productivly used to – performing brand audience research that might help brands understand themselves and their audiences better and, off course, find the right context to meet and interact with their most relevant audiencens.