Staff Papers

User collaboration in websites

Paper presented by Judith Pearce at Framing the Future, ARLIS/ANZ Conference, 21-23 September 2006.

Thanks to Michele Huston for lending me her Web 2.0 slides and to Fiona Hooton for letting me use material from the draft flickr Pilot Project Evaluation Report.

Introduction

Libraries serve as gateways to resources that meet the needs of their clientele. In recent years we have endeavoured to make our collections and services available online. Meantime the expectations of our users have been changed, first by Internet search engines and the lure of the simple search box, and now by what is being dubbed as Web 2.0, the next generation.

In this presentation I’m going to talk to you about what is meant when people talk of Web 2.0 and the Internet as platform. Then I will look at two services offered by the National Library of Australia – PictureAustralia and Australia Dancing – and some of the things we are doing through user collaboration to help them remain relevant in this rapidly changing world. Lastly I will look at the relationship between PictureAustralia, Australia Dancing and the proposed new People Australia service; and how this service will also need to embrace user collaboration.

Web 2.0: The Internet as platform

Web 2.0 is a term made popular by the publisher Tim O'Reilly [1]. He first used it at a conference brainstorming session in 2004 to distinguish companies that had survived the dot com crash from what had gone before. In simple terms it refers to websites that let people collaborate and share information in previously unavailable ways.

In August 2006 the Observer published a list of fifteen websites that changed the world, including sites such as eBay, Wikipedia, Blogger, Amazon and Google [2]. Sites such as these are seen as exemplifying the Web 2.0 concept.

All of these services share similar characteristics. Firstly, they are based on massive data repositories. Some of this data is being continuously created or uploaded by users. Some of it existed before but was not accessible or useful in its raw form. New licensing models are enabling this data to be replicated and used, or combined with other data, for new and unexpected purposes. The geospatial data used to create Google Maps is one noteworthy example. Google licences this data from its original creators, using profits from advertising to make it freely available.

Web 2.0 services also encourage user participation, either by contributing data or through tagging, comments, ratings and reviews. The user has a symbiotic relationship with the community formed around the application. This is based on trust. Often no more is required than to register with a valid email address but more complex trust models are also supported. The user’s interaction with the service may be enhanced through the use of technologies like Ajax that make web pages behave more like a desktop application by exchanging small amounts of data with the server behind the scenes [3].  This makes the services seem more interactive and easy to use.

Web 2.0 services also allow relationships to be built between services using programmable interfaces. Google Maps, for instance, offers an Application Program Interface (API) that lets you embed Google Maps in your own web pages [4]. This trend is enabled by cheap storage and computing power. Start-up costs are within reach of small players, even those without their own servers. Amazon has just released a new service that sells cheap web access to computing facilities on a per-hour basis [5].

flickr as example

Marcus Angermeier's tag cloud [6] gives a really good overview of the kinds of ideas encapsulated by Web 2.0. A cloud tag is a way of visually depicting the terms used to tag content on a website such as a blog or photo sharing service. The first widely known tag cloud appeared on flickr, a web-based photo management and sharing service now owned by Yahoo! that has been operating since 2004. What differentiates flickr from other photo management services is its focus on enabling new ways of organising photos through the assignment of descriptive metadata and tags. Features such as the most interesting photos for a given tag use an unpublished algorithm based on things like the number of people who have tagged, commented on or viewed a photo or marked it as a personal favourite.

The service continues to evolve in myriad ways, with latest developments including clustering based on some form of semantic analysis and geocoding for discovery via Google maps. Through services such as Groups, flickr also supports the development of online communities with common areas of interest.

flickr can be seen to have all of the characteristics mentioned before: a massive repository, user participation, a simple trust model based on community guidelines, programmable interfaces that enable flickr images to be re-used on other websites and cheap computer power and storage. Bandwidth is the limiting factor for the free service rather than the amount of data uploaded.

Wikipedia as example

Wikipedia is another often-quoted example of a Web 2.0 service. Wikipedia is a web-based encyclopedia that invites users to freely edit its content. It began in 2001 and now has over 1.3 million entries in the English language version. It is a fascinating social experiment, often criticised, as it says itself, for “its susceptibility to vandalism, uneven quality and inconsistency, systemic bias, and preference for consensus or popularity over credentials [7].” However, it is ubiquitous, with entries often turning up high in search engine result sets, and also an extremely useful means of getting a quick overview on topics for a paper like this.

The National Library of Australia and Web 2.0

The pace of change seen in services such as flickr and Wikipedia has significant implications for the National Library of Australia. People of all ages are now coming to expect a degree of inclusiveness, interactivity and access from websites that they use on a daily basis that is not yet reflected in our services.

For this reason, when the National Library reviewed its Directions earlier this year, as well as including strategies for collecting, safeguarding and ensuring rapid and easy access to the Australian documentary record, we created a new set of strategies for ensuring our relevance in a rapidly changing world [8]. We saw that Web 2.0 services might provide models for enhancing our own services and sharing our data with other services, as well as opportunities to acquire data from new sources, participate in new online communities and enhance our visibility.

In summary our goals in this area are to explore new models for collecting material; to provide services that support the creation and dissemination of knowledge by users; and to enhance opportunities for users to share ideas by interacting with us and among themselves.

Today I am going to talk specifically about two projects we have undertaken to achieve these goals – a project to enable people to contribute images to Picture Australia through the flickr photo sharing service and a project to add information, Wikipedia-style, to Australia Dancing.

The flickr Pilot Project

PictureAustralia is a federated service that lets users search across a number of online pictorial collections at the same time. It began in 1998 as a pilot project called ImageSearch and now provides access to over one million pictures provided by 44 participants. PictureAustralia provides access primarily to historical material. While this is, in fact what differentiates it as a data repository from Google Image and flickr, a survey in 2004 found that 43% of users were looking for contemporary images [9].

Aim

The aim of the flickr Pilot Project was to increase the number of contemporary images in PictureAustralia and to engage with new audiences. flickr was chosen as a suitable partner because it already had a large following of Australians, because it encouraged the use of metadata to aid discovery and because it had a programming interface that could easily be mapped to Dublin Core, the metadata schema used to contribute records to PictureAustralia. The launch of the pilot was timed to coincide with Australia Day (26 January) 2006.

How it works

How it works is quite simple. Users upload their images to flickr and add them to one of the PictureAustralia groups. Once a week the Library harvests the metadata and thumbnails and adds them to PictureAustralia.

Two groups were set up in the first instance, one to collect images relating to Australia Day, the other a more general group based on people, places and events [10]. Since the pilot was launched over 9,000 images have been contributed to PictureAustralia.  In PictureAustralia the images show up in result sets side by side with images from other collections [11].

Outcomes

There is no doubt that this pilot project has increased PictureAustralia’s coverage of topical issues. It has done this in a very Web 2.0 fashion by harnessing the collective efforts of people already capturing events as they happen for their own purposes. Some of the contributors are now also actively taking pictures with contribution to PictureAustralia in mind. This has led to the introduction of many more full-colour images and also to some interesting juxtapositions of old and new pictures with the same subject. Studying the old and the new seems to have a universal fascination. Groups have already formed on flickr independently of this project, to collect Australian images around particular themes and it has been possible to invite photographers contributing to these groups to contribute their pictures to Picture Australia as well. There has been little need to moderate images, although some have been removed as out of scope, for example, pictures of pets.

The flickr images go into a kind of half-way house where they remain accessible through PictureAustralia for the life of the image itself on flickr, without any specific agency needing to take them into its collection. This is enabled by the PictureAustralia architecture, where only the metadata and thumbnails are harvested. Users link to flickr to see the full image. Significant images are, however, being selected for the National Library of Australia’s collection. The creators of these images are contacted to obtain a high resolution version of the image, which is then taken into the Library’s digital archive, to safeguard for future access. A high resolution version is needed because of the bandwidth limitations of uploading images to flickr.

There has been an unexpected increase in heritage images and this is a rich area for further mining. The thing that most distinguishes PictureAustralia from Google Image and flickr – is its historical material. Strict collecting policies have been applied until now to family history materials because of the costs of bringing it into the collection. The flickr halfway house approach will enable the Library to invite the uploading of such works to flickr and their contribution to PictureAustralia. This benefits family historians as well, by introducing them to flickr as a photo hosting service and to advice from the National Library on how to ensure the preservation of their personal collections, even if their images are not taken into the Library’s collection.

The visibility of PictureAustralia has certainly been increased by the pilot, as the new contributors search for their own images and become fascinated with the juxtapositions that they find. In a survey recently conducted of participants in the project, 39% had not heard of PictureAustralia before, 64% heard about the project through flickr and other flickr members and 69% went to Picture Australia to see their images [12]. There has been a significant sustained increase in use of PictureAustralia as a result.

A survey has yet to be conducted of PictureAustralia users not involved in the flickr project to see if they are finding more contemporary images. Metadata quality is a significant issue, not so much because contributors are not using a standard thesaurus (PictureAustralia members already use a mixture of thesauri) as because they are not giving images a sufficiently descriptive title or assigning enough tags. Picture Australia administrators monitor the images and email contributors with advice on tagging. This has improved the quality of the metadata. As a sense of community develops there is also likely to be a level of self-education and management through the discussion forums associated with the PictureAustralia groups.

Future directions

To sustain the interest shown by Australians to participate in the pilot project, the Library will need to focus resources on development of these communities so that they can be self managing and assist in achievement of the Library’s objectives. A need for new groups has been identified that focus on targeted areas and for the Library to continue to provide guidance on the content that is being sought and how it should be tagged.

It is still the case, however, that images are often not easy to find in PictureAustralia, whether or not they have been contributed through flickr or another agency. The repository is already very large and will increase significantly in size as the flickr experiment proceeds. PictureAustralia itself may need to look at Web 2.0 service models in order to simplify access to the collection. What it has going for it is its ‘killer data’, a term used by Simon Greener in a recent article to describe what Google has been doing with Google Maps by exploiting hitherto inaccessible or usable reservoirs of geospatial data [13]. Picture Australia’s killer data is its heritage images, the juxtapositions these enable with contemporary material, the persistence of images selected by participants for preservation and the degree of ‘tagging’ and ‘geocoding’ already done by cataloguers for the permanent parts of the collection.

The Library is already looking at ways of enhancing access to this content through  ranking and clustering of result sets and the geocoding of images.  In an image-based service, getting this right depends on the quality and consistency of the metadata.  The Library has started to think about the benefits of thesaurus mapping to improve recall and to support the clustering of result sets. However, discovery still fails when an image is not appropriately or sufficiently described; and it is here that users are well-positioned to add tags and comments that might reveal previously unknown information about the image such as the location or name of a subject.  

There is a perceived risk that the authority of a service like PictureAustralia could be diminished by allowing users to add tags and comments. Existing services that don’t do this are now starting to look inert. While some user tags and comments might not have much to contribute to the discovery or understanding of a picture, they still provide metrics for determining its ‘interestingness’.

PictureAustralia, is an aggregation of data from many participants. Most of the institutions contributing data to PictureAustralia will also have websites providing access to their own images. The Powerhouse Museum has recently started experimenting with tagging [14] and the National Library has set up a working group to look at annotation requirements for its own collections. There is an emerging requirement for annotations of all kinds – tags, comments, notes, ratings and reviews -  to be shared to support features such as ranking and clustering or the browsingof images deemed most interesting or relevant, both at the local and aggregated level.  

Australia Dancing Takes Part

I would now like to turn to Australia Dancing and to work the Library has just begun to enable people interested in Australian dance to contribute to the Australia Dancing website. Australia Dancing  is a federated directory of resources held in Australian collections that are related to dance in Australia. Unlike Picture Australia, it takes a contextual approach to resource discovery. The primary objects for discovery are people, companies and performances. This approach allows it to operate as a dictionary of biography for users whose needs are met by the biographical entry, while also enabling users to navigate to linked resources from a number of Australian collections.

Aim

Almost as soon as Australia Dancing was released users started to write in asking why they were not in the directory.  The aim of the Take Part Project was to make Australia Dancing more inclusive by supporting the creation of entries by users. This could have been done by opening up the update interface to authorised users. However, the Library was curious to try and see if a simpler, less structured approach might help to create a more flexible online space for people interested in dance in Australia centred around the Australia Dancing service.

How it works

The Take Part service is a stand-alone website based on the Confluence enterprise Wiki software. Users link to Take Part from the Australia Dancing home page. Anyone can register and contribute data as long as they agree to the conditions. These are simple: stay on topic, use your real name, respect the copyright and intellectual property rights of others, provide your contributions under the Creative Commons Share and Share Alike license and contribute content that will not offend.

Registered users can add a new page using a WISIWIG editor with capabilities to upload and link to images and documents. They can also add comments to or edit any existing page and view a history of the edits made to a page.

Outcomes and future directions

This project has not been widely promoted yet and so I can’t say very much more about its outcomes. What has been demonstrated so far is that the Confluence software is suitable for a Wikipedia-style approach to the creation of biographical entries. Building the community is the next step.

There are still issues to be resolved in terms of integrating Take Part with the main Australia Dancing service. There is certainly a requirement to enable users to comment on existing entries, to find people, companies and performances and to receive alerts from both collections of information. One option would be to migrate the Australia Dancing service itself to the wiki environment once the community is fully established.

This raises the question of whether a service such as Australia Dancing could be built through a fully open model. In fact, why not just use the Wikipedia itself rather than continuing to maintain a stand-alone service?  The science journal Nature recently compared natural science sections of Wikipedia and the Encyclopedia Britannica and found that the two were close in terms of the accuracy of their articles [15]. This study is used in the Wikipedia to counteract the criticisms that I quoted earlier. However, the Wikipedia entry fails to cite the response from the Encyclopedia Britannica which found the results “fatally flawed” [16].  

Other studies are starting to suggest that the “neutral point of view” policy adopted by Wikipedia is often weighted by the sheer volume of comment at one end of the seesaw.  Biography is particularly subject to interpretation and dispute. Paul Turnbull has reported recently at a recent conference on Biography and Technology on comparisons he has made with biographies in the Wikipedia and the Oxford Dictionary of National Biography that demonstrate the extent to which error can creep into an entry through politicisation and mythologising [17].

I did some testing on my own for this paper and found some interesting differences between the Australia Dancing entry on Paul Mercurio and the one in the Wikipedia. The Wikipedia entry has more up-to-date information but gives prominence to information that is of immediate public interest such as the fact that the subject is one of four judges on the Australian version of Dancing with the Stars. The Australia Dancing entry gives a more balanced assessment of the subject’s life as a professional dancer but does not mention the Dancing with the Stars connection.

Replicating the Australia Dancing data in the Wikipedia would mean enabling anyone to change the existing content. Paul Turnbull and others at the Biography and Technology conference argue that biography is an art and that the integrity of a scholarly biographical entry needs to be preserved. This is probably the single most important reason for continuing to develop the service as a stand-alone website, where an appropriate collaborative model can be tested and implemented.

The collaborative model currently being implemented for the Dictionary  of Australian Artists Online is based on registered users and peer review. It allows for multiple entries from different authors with different points of view and the addition of comments.  It also supports the recording of descriptive metadata such as gender, field of activity, training and awards; and links to related resources. This is the model that I suspect Australia Dancing will need to embrace if it is to meet user needs and be interoperable with other services.  

Biographies – killer data

A number of services now exist or are under development that provide access to authoritative biographical information, some of which also link to related resources. There are also valuable sources of information about people and organisations hidden in name authority files, finding aids and resource descriptions. A need is starting to be articulated worldwide for a way of resolving public identities, both as a means of managing intellectual property rights and of interlinking between services that contain information about parties or about resources by and about them.  For some time the National Library have been toying with the idea of repurposing and extending the Australian Name Authority File for this purpose.

The Australian Name Authority File is a database of 120,000 personal names and 90,000 corporate names used to collocate resources under a single form of name in the Australian National Bibliographic Database (ANBD). It is collaboratively maintained by Australian libraries. 

We have now initiated a project to implement a new People Australia service based on the Australian Name Authority File. The service will act as a portal for the discovery of Australian people and organisations by clustering entries created by partner agencies, with links to related resources. It will be based on a database of people and organisations with records contributed from a variety of sources including the Australian Name Authority File.

A primary function of the People Australia service will be to identify and collocate information about the same party and to disambiguate parties sharing the same name. This has always been a challenge for national authority files. Web 2.0 services are now providing new models for harnessing the collective endeavour of online communities to achieve this purpose.  Cataloguers, reference librarians, curators, subject experts, the general user and parties themselves may all have knowledge that can contribute to the development of the service and be highly motivated to do so, given appropriate easy-to-use interfaces and trust models.    

Conclusion

Much work and thinking still needs to be done to ensure that the National Library of Australia’s services remain relevant in a rapidly changing world. The Flickr Pilot Project and Australia Takes Part show ways in which we might start going about it. The challenge will be to find ways of exploiting our own killer data through ongoing collaboration at both an institutional level and with end users.  One of our current priorities in this area will be to focus on a contextual approach to resource discovery through the People Australia service.

Endnotes

[1] Tim O’Reilly, “What Is Web 2.0: Design Patterns and Business Models for the Next Generation of Software”, O’Reilly, 30 September 2005.  Online. Available: http://www.oreillynet.com/pub/a/oreilly/tim/news/2005/09/30/what-is-web-20.html (viewed 21-Sep-2006).

[2] John Noughton, “Fifteen sites that changed the world”, Observer, Sunday August 13, 2006. Online. Available: http://observer.guardian.co.uk/review/story/0,,1843263,00.html (viewed 21-Sep-2006).

[3] “Ajax (Programming)”, Wikipedia: The Free Encyclopedia. Online. Available: http://en.wikipedia.org/wiki/Ajax_%28programming%29  (viewed 21-Sep-2006).

[4] “Put Google Maps on Your Own Web Site”, Google Maps Beta. Online. Available: http://www.google.com/apis/maps/ (viewed 21-Sep-2006).

[5]  “Amazon Elastic Compute Cloud (Amazon EC2) - Limited Beta”, Amazon Web Services. Online. Available:   http://www.amazon.com/gp/browse.html?node=201590011 (viewed 21-Sep-2006).

[6] “Web 2.0 Extended Mind Cloud Map”, flickr Gamma. Online. Available: http://www.flickr.com/photos/kosmar/62381076/ (viewed 21-Sep-2006).

[7] “Wikipedia”, Wikipedia: The Free Encyclopedia. Online. Available: http://en.wikipedia.org/wiki/Wikipedia (viewed 21-Sep-2006).

[8] National Library of Australia, Directions for 2006-2008. Online. Available: http://www.nla.gov.au/library/directions.html (viewed 21-Sep-2006).

[9] PictureAustralia flickr Pilot Project: Evaluation Report. Draft dated 18 September 2006.

[10][List of PictureAustralia flickr Groups], Flickr. Online. Available: http://www.flickr.com/search/groups/?q=pictureaustralia+library&m=names (viewed 21-Sep-2006).

[11][PictureAustralia Search on Opera House], Picture Australia. Online. Available:

http://www.pictureaustralia.org/apps/pictureaustralia?action=PASearch&mode=search
&term1=opera+house/
(viewed 21-Sep-2006).

[12] Draft evaluation report, op. cit.

[13] “Killer Data – LiDAR”, Directions Magazine, 12 October 2005. Online. Available: http://www.directionsmag.com/article.php?article_id=1986 (viewed 21-Sep-2006).

[14] Powerhouse Museum Collection 2.0 beta. Online. Available: http://www.powerhousemuseum.com/collection/database/ (viewed 21-Sep-2006).

[15] Jim Giles, “Special Report: Internet encyclopaedias go head to head”, Nature, 438, 900-901 (15 December 2005). Online. Available: http://www.nature.com/nature/journal/v438/n7070/full/438900a.html (viewed 21-Sep-2006).

[16] “Fatally Flawed: Refuting the recent study on encyclopedic accuracy by the journal Nature”, Encyclopædia Britannica, Inc. March 2006. Online. Available: http://corporate.britannica.com/britannica_nature_response.pdf (viewed 21-Sep-2006).

[17] “Biography and Wikipedia”, paper presented at Humanities Research Centre, Biography and Technology, Australian National University, 12-14 September 2006.

Back to the top