Category: Folksonomies

Continuing on the additions from yesterday.

Today we added the top three search terms for each object to the USER KEYWORD section of an object page. This section is where folksonomy tags can be added and deleted.

Here is a 1960s raincoat for example.

Why did we add this to the USER KEYWORDS section?

What we are finding through our ongoing analysis of folksonomy tagging behaviour is that users are generally adding synonyms. What folksonomies are doing in this instance is effectively ‘crowd-sourcing’ synonym generation. Now when a user searches for term and selects an object from a list of results they are making an association between the object and that term. In many instances these may be tentative associations and generally full of false drops, but when aggregated patterns begin to emerge (see Chan 2006 forthcoming). Effectively on our site we are noticing that the search terms are beginning to offer the same kind of synonym behaviour that tags do.

We are presenting them together as a way of making explicit the terms that are already associated with an object (automatically). This way we hope to improve the diversity of user tags (why tag with a word/term that already appears?). This is just a trial though and if we see little or no change then we may move the search terms off to a separate section.

I’d welcome any comments or thoughts on this as we’re experimenting here.

Take a look at the new ‘related searches’ feature on our collection search.

Now a search for ‘glass‘ will these ‘similar searches’ –

glassware vase bottles bowl bowls

This result will change over time. Hopefully we will implement a timescale simulator in our upcoming ‘experimental’ browsing section which will allow users to view the changing search language over time.

How does it work?

Because we keep a large store of relationships between search terms and clicked objects, we are able to reverse query terms such as ‘glass’ and see what terms have been used to find similar objects. We currently show only the top 5 terms – aggregated – which lessens the probability of ‘false drops’. False drops are most likely to occur for uncommon search terms although this will change over time, too.

Does this use folksonomy tags?

Yes. We allow user tags to be included in the search terms and from time to time certain objects will be most visited as a result of their user tag.

Today we finally ironed out one of the major problems with the search engine on our OPAC2.0/collection database.

There are still some tweaks to be done on the results, and the advanced search needs to implemented but the new search is much better than the original.

If you have some spare moments and feel like trying a few searches, please do so. If something odd happens then I’d welcome your thoughts in a comment on this post.

Next up for OPAC2.0 is the presentation of ‘other search terms similar to X’ and ‘others who searched for X looked at’ alongside search results. We have already implemented this on Design Hub and have the code and data ready to go.

Then it is on to adding a automatic spell checker for the folksonomy tags to reduce post-user tag editing.

danah boyd, Cameron Marlow, Marc Davis and Mor Naaman, all of Yahoo, explore social tagging in detail in their paper presented to ACM/Hypertext06 in Denmark.

Of particular note are the sections on system design and user incentives which cover the differing types of systems and methods behind different implementations of tagging. They also suggest that considerably more research is required into the different ‘lects’ used in tagging and the phenomenon of ‘vocabularly overlap’ between random users’ tags in a short Flickr case study.

Essential reading.

ABSTRACT

In recent years, tagging systems have become increasingly popular. These systems enable users to add keywords (i.e., “tags”) to Internet resources (e.g., web pages, images, videos) without relying on a controlled vocabulary. Tagging systems have the potential to improve search, spam detection, reputation systems, and personal organization while introducing new modalities of social communication and opportunities for data mining. This potential is largely due to the social structure that underlies many of the current systems.

Despite the rapid expansion of applications that support tagging of resources, tagging systems are still not well studied or understood. In this paper, we provide a short description of the academic related work to date. We offer a model of tagging systems, specifically in the context of web-based systems, to help us illustrate the possible benefits of these tools. Since many such systems already exist, we provide a taxonomy of tagging systems to help inform their analysis and design, and thus enable researchers to frame and compare evidence for the sustainability of such systems. We also provide a simple taxonomy of incentives and contribution models to inform potential evaluative frameworks. While this work does not present comprehensive empirical results, we present a preliminary study of the photo- sharing and tagging system Flickr to demonstrate our model and explore some of the issues in one sample system. This analysis helps us outline and motivate possible future directions of research in tagging systems.

A few weeks back I posted an initial chart showing distributions of object usage on our OPAC2.0.

Here’s a quick updated chart but done with logarhythmic scales on both axes.

MS Excel seems to only cope with 32,000 values on one axis so it cuts off artificially at 32,000 (out of 55,134 objects viewed of the total 61,780 currently available)

Some other useful data:
Total object views to date = 1,776,259
Max views for single object = 2104 (Delta Goodrem dress)
Average views per object = 28.751
Standard deviation = 40.819
Median views = 19

Popularity drops below 10 views at rank 39,313 (not shown on graph as a result of Excel limitations)

Already there is a clear line emerging which droops around the 20K rank point – which indicates that there is still some way to go with driving traffic down to the more obscure objects in the tail. The bump at the head is the result of objects that are receiving abnormally large amounts of traffic – the Delta Goodrem dress, the Nu-U bra (the one from 1957) – as a result of time-specific cultural factors.

Lynda Kelly at the Australian Museum has relayed some reporting on tagging from a recent Web Usability seminar. (Lynda is part of an ARC project we are collaborating on.)

Roger Hudson took us through a brief history of classification and taxonomy(Linnaeus I think, Dewey, etc etc), making mention of an interesting Indian historical figure who had introduced the idea of classifying by “facets”. This idea was not widely taken up but is now highly relevant to the ways that tags are used. He also presented some *very* preliminary research with punters about tags – what they were and how they were being and could be used. The messages for me from his talk were (with apologies in advance to Roger as I am just outlining my impressions which could be wrong!):

1. Little understanding of the concept of tagging

2. Little understanding of why some words were larger in a tag cloud that others

3. A wide variety in the ways that people could potentially tag something. For example a picture of a redback spider was tagged as a spider (obviously); redback (also obviously); however other tags were Slim Dusty and dunny (think about it…) which i thought were pretty cool

4. The potential that as tag clouds make the “popular” tags the biggest, there could be “expert” tags that are lost (as in the above example where only 3 or so people used the word “arachnology” as a tag which is something that other experts may seach on)

When it comes to collections we are noticing some different trends emerging – mainly because tags on our site are combined with controlled vocabularies and are thus enhanced in this way, the end result for users is better/broader.

The stats we are accumulating are now showing a clear preference for tag as entry point, but interestingly enough, NOT necessarily tagged content as end point. Thus a user might click on the big tag MODEL TRAIN but then not view an actual OBJECT tagged as model train, but one of the results from a free text search for the term.

(I’ll be presenting some statistical evidence on these trends in future presentations and perhpas in a future post)

Unlike a lot of other sites that use tags we are not JUST using tags as a folksonomic classification system, we are also using them as search entry points. The use of tags as search entry points means that we are increasing the likelihood of users widening rather than narrowing their search.

Lynda has posted links to two excellent introductory pieces on folksonomies as well.

http://www.rashmisinha.com/archives/05_09/tagging-cognitive.html

http://infotangle.blogsome.com/2005/12/07/the-hive-mind-folksonomies-and-user-based-tagging/