Video Vortex: Jan Simons on the narrative of tagging
The presentation given by Jan Simons is called Weddings, Cities and Colors via tagging.
Simons has performed a short study into user generated content and user generated indexing. With a background in cinema and narrative, Simons is interested in tag activity and the problems of tagging in searching an ideal model of readers/speakers. Very often these models are defined on introspection, where it is assumed that minds of other people work just like your own. The internet and 2.0 provides us with the way users of the internet think and relate; it gives us traces of their thinking. What If tags that users attach could tell us something about how the ‘wisdom of the crowd’ actually works?
Simons presents a rapid-prototyped- kind of research, a proof of principle via Flickr. Flickr was chosen as object of study due to its tagging system which is a blind tagging system (so no recommendations. the tags used are really thought of by users themselves. Flickr also gives a tagclouds and provides clusters of tags. Lots of prelimenary research is already done in this sense).
The main question is how people categorize their thoughts and experiences. Within folksonomy he trouble is to know what users actually mean – do they have the same concept of the words and the content they put on flickr? Some problems with terms concerning tagclouds are mentioned (NYC NY and New York, for instance, leading to thesame content).
Within a tagcloud only nouns (no verbs) are used. The most popular tags in a graph creates a powerlaw graph. The most popular tags are japan, NY and wedding.
Tags are messy categories. Polysemy is a major problem (e.g. place-names) as well as synonemy (e.g. fall, autumn, city and urban) and homonomy. It gives an unrelated meaning (e.g. rock meaning stone, but also a music genre) The level of categorization is also a problem (from generic to specific). Another obvious problem is that of spelling. San is often used for instance, but what does it mean? San Francisco, or San Diego? San is split up, the same occurs with New York; New becomes a tag.
There are also users who abuse tags. One user for instance used all the popular tags from flickr and hooked it to his photo. Flickr says: use more than one tag to increase its searchability. He did, very literally. Other types of misuse are that of adding very strange tags in order to avoid censorship. This misuse pollutes the tag system. There is a difference in “in” and “about” England, for instance. Tags are unreliable as guide.
Simons continues by talking about the approach in looking at tags. Tags as labels for things, or names for objects or places. When you look at correspondence, tag and tagged – nouns and adjectives – tell something about the properties of the object. Tags are never used in isolation. Lots of users use Flickr as a backup system for their own pictures. (as a kind of life insurance). This private use is often used without tagging. This also pollutes the tag system.
Categories of tags:
1 geographical tags. names of countries (largest cat.) states cities,,. where california is the most popular.
basic level terms are most commonly used
2. by events. Querying via events (christmas, holiday) shows activity that is linked to that event.
3 by nature of the object.
These are the most used categories.
About tag semantics: most popular is temporal metadata that is copied with the picture (the camera holds this information).
This gives an argument structure (time, location, event).
A distinction exists between nuclear arguments (core) and satellite argument (more general). Via this information, one can see that they follow a very basic semantic structure.
In order to complete the categorisation, a distinction is made between states and events. Within events, there is also the event of making the picture (picture of an event or of a state. where to put this?). Now we can distinguish the manner of photography and the instrument. Another problem now is polysemy: does color say something about the picture or the objects in the picture? If you apply linguistic analysis to tags, we see that tags are highly structured and very consistent. They follow a pattern of natural language. This is very important in understanding and interpreting the content (in this case, pictures). This could tell us a lot about what users conceive of the world via their pictures/ videos.
Q&A: users give meaning to pictures, narratizing it. The idea is that inner speech evokes mini-narratives.
How can we use the semantic results from this research? Simons replies that Flickr was used as a research objects due to it being concised and accessible (more than YouTube). This research shows that users have a more sentence-like approach to tagging than just labeling. This is what it shows, possibly one could project this onto youTube, although youTube is a different medium.
Vocabulary of the amateur photographer? Isn’t that what this research shows? Simons replies: Yes and no, events do happen a lot around the photograph itself. Tagging is biassed by activity that users have in common on Flickr, which is in this case photography.
Picture by Anne Helmond