Thursday, December 10, 2015

Implementing User-Generated Tagging with Digital Photographic Collections

            Given the nature of today’s information focused and technologically based society, it is expected that folksonomies would feature prominently in many aspects of the field of information science. Folksonomies have become popular in many familiar places on the Web, including, but not limited to, YouTube, Twitter, Instagram, and Pinterest. Social tagging is also beginning to be implemented on various library catalogs (Peterson, 2008).
            Folksonomies enable users to assign their own subject oriented tags for whatever video, post, photo, or other digital file, and then have the ability to retrieve information from searching these and others’ created tags. They allow for the potential of more access points in information retrieval and the fact that they are beginning to be used in library catalogs emphasizes their potential usefulness when thinking of other collections. However, there is a reason that folksonomies haven’t taken over in other areas; their very structure goes against the nature of traditional catalog indexing using controlled vocabularies (Menard & Smithglass, 2012; Peterson, 2008). There are many aspects of user-generated tagging that make it challenging to implement.
            The present challenge comes in obeying traditional cataloging rules, while also making index terms engaging to current users. Users today expect interaction on the web, because of the culture of social networking, and social tagging fulfills this need (Peterson, 2008). Controlled vocabularies remain the primary indexing technique because of the weaknesses associated with folksonomies, such as their lack of specificity and lack of control over the terms (Noruzi, 2006). However, with the right techniques and developments, folksonomies could be helpful with visual materials in institutional collections, as they could help reach a broader audience (Beaudoin, 2007).  
            There are various digital collections that have made some attempts at implementing social tagging, most notably, Flickr. The hierarchical tagging technique has been implemented quite successfully with Flickr, allowing relationships to be captured between the different levels of the subject terms (Plangprasopchock & Lerman, 2009). Flickr tags appear to utilize a wider range of index terms than might be suspected. One study, looking at the Library of Congress stream on Flickr, compared the tags to terms using the controlled vocabularies of the Thesaurus for Graphic Materials and Library of Congress Subject Headings. The comparison suggested that extending LCSH and TGM to incorporate user generated terms, such as those used by Flickr users, will make these vocabularies more accessible to a wider user base (Stivilia & Joergensen, 2010). Social tags on Flickr have shown success in some ways, but it remains to be seen if they really can, or should, replace traditional controlled vocabulary terms.
Using the two as complements to one other can help increase the information known and make the collections richer by gaining general knowledge about the collection from the user tags. However, although user tags may help to identify people and places, they don’t appear to improve retrieval of images (Menard & Smithglass, 2012). Yet, as the tags appear to increase public knowledge and interactions to improve tags, in this way they are improving the quality and quantity of the metadata surrounding the images.
            There are a number of online photographic collection projects that have attempted to implement the use of social tagging techniques, as Flickr has done. The CoDesPho Project (Collaboration for the Description of Photographs) was started in Belgium and centered on a collection of digitized photographs from 1890-1960. The premise was to allow users to enter comments about the images in order to increase the available metadata. Of course, these comments would need to be screened carefully by an expert reviewer in order to eliminate those that were not helpful. This project determined that there was a strong correlation between the comments and enriched metadata, suggesting that the user added comments did help fill an information need (Durieux, 2010). Unfortunately, this project was discontinued in 2007, for reasons that were not discussed, but it seems likely that the time commitment to screening the comments might be one reason.
            The National Archives of the Netherlands launched a similar project in 2004, using a digital collection of 500,000 images. Like the CoDesPho project, this collaboration also included a free text comment form that would be screened by expert reviewers and the user comments were evaluated based on their relevance towards user queries. The comments were found to add significantly meaningful, specific, non-abstract terms to the existing metadata for the images (Van Hool, 2006). Similarly, the Picture Australia project collaborated with universities, city libraries, and museums to use a technique that would combine traditional cataloging with social tagging. The images were described using traditional cataloging methods, but were also added to a Flickr stream, where general users were encouraged to add more images, as well as assign subject tags. This project successfully uses traditional methods to satisfy cataloging, but also recognizes the importance of user interaction to add to the digital collection (Peterson, 2008).
            Naturally, the integration of social tagging with traditional cataloging is not a seamless process; there are many challenges that still exist. Some professionals in the field of information and library science even maintain that it is not possible to incorporate the two because they are two distinct terms and serve two different purposes (Peterson, 2008). Others are less dismissive, but still skeptical, maintaining that there is still a fundamental difference between the two because of different underlying structures (Rorissa, 2010). Tags are assigned freely and without restrictions on quantity, while indexers are required to minimize the number assigned; tags are made within some sort of personal context, while indexers create subject terms after a thorough examination of context, leading to higher retrieval rates.
            The most challenging barrier to implementation of social tagging seems to be the lack of a controlled vocabulary and Noruzi (2006) lays out the four main problems behind this: polysemy, synonym, plurals, and depth/specificity. Polysemy (words with two or more meanings), synonymy, discrepancies using plurals, and the level of depth/specificity of the terms all lead to inconsistences with the socially tagged terms, all of which can lead to inaccurate retrievals or low retrievals rates (Menard & Smithglass, 2012; Noruzi, 2006). Another challenge to overcome has to do with the language that the indexers use compared to the language of the social tags made by the user. As alluded to previously, both of these groups describe image content in such different ways, as they have a different framework in mind in approaching the subject terms and/or tags (Stewart, 2013). However, it seems likely that these challenges can be overcome with greater understanding of user expectations and the language of the current user.
            The most likely solution seems to be double indexing, or combining the use of traditional cataloging with socially generated tags. This could mean a higher cost because of the time, energy, and resources needed to implement this, which ultimately is not helpful in increasing access and leads to many professionals being wary of its implementation (Menard & Smithglass, 2012). However, some solutions exist for decreasing the problems associated with social tagging, making it a more feasible solution. Flickr has adopted such developments; 1) an automatic ingest of information when images are uploaded, in order to reduce the likelihood of human error, and 2) using a schema to prompt the user to choose the “correct” tag from an expert created thesaurus of controlled terms, as a form of guided tagging (Beaudoin, 2007). Also, Noruzi (2006) has suggested coming up with a singular/plural tool that would seek to accommodate any form of the tag selected, decreasing the chance of terms being overlooked in retrieval because of the tag being only singular or only plural. Further, many studies have found the implementation of the collaboration of cataloging with social tagging to be worth the time and energy needed for its upkeep (Plangprasopchok & Lerman, 2009; Peterson, 2008; Rorissa, 2010; Stivilia & Joergensen, 2010; Van Hool, 2006).
            As social tagging is a relatively new development, especially in the world of digital photographic collections, it appears that there is still a lot of murkiness regarding the effectiveness of combining it with traditional cataloging. Some information professionals argue that it is cost-effective and worth the time and resources required, while others argue that it is not and that it is not feasible to combine the two because of their different forms and functions. This ongoing development also leaves a lot of questions unanswered: Where do we draw the line between user-generated metadata as added value, or as expressing a biased view that devalues the information professionals authority as expert? How will these “social tags” be successfully incorporated into existing metadata and do they have the potential to accurately retrieve results? It’s clear that more research about and more exposure to social tagging is needed to help bridge the gap between index terms, understanding users search language, and figuring out how best to implement tags.

Works Cited
Beaudoin, J. (2007). Flickr image tagging: Patterns made visible. Bulletin of the American Society for Information Science and Technology, 34(1), 26-29.

Durieux, V. (2010). Collaborative classification of digital photographs: The CoDesPho project. Unpublished.

Menard, E. & Smithglass, M. (2012). Digital image description: A review of best practices in cultural institutions, Library Hi Tech, 30(2), 291-309.

Noruzi, A. (2006). Folksonomies: (Un)controlled vocabulary? Knowledge Organization, 33(4), 199-203.

Peterson, E. (2008). The coexistence of subject cataloging and folksonomy. Library Philosophy and Practice (e-journal), April.

Plangprasopchok, A. & Lerman, K. (2009). Constructing folksonomies from user-specified relations on Flickr. Proceedings of the 18th international conference on World wide web, 781-790.

Rorissa, A. (2010). A comparative study of Flickr tags and index terms in a general image collection. Journal of the American Society for Information Science and Technology, 61(11), 2230-2242.

Stewart, B. (2013). Pictures in words: Indexing, folksonomy, and representation of subject content in historic photographs. Retrieved from http://ro.ecu.edu/asu/theses/687.

Stivilia, B. & Joergensen, C. (2010). Member activities and quality of tags in a collection of historical photographs in Flickr. Journal of the American Society for Information Science and Technology, 61(12), 2477-2489.

Van Hool, S. (2006). From spectator to annotator: Possibilities offered by user-generated metadata for digital cultural heritage collections.


No comments:

Post a Comment