Book tagging on LibraryThing: How, why, and what are in the tags?

Authors


Abstract

The purpose of this exploratory study is to understand book tagging by investigating LibraryThing (LT) members' purposes for tagging; exploring how tags are used; and comparing member assigned tags with terms in corresponding MARC records. The results show that LT members tag mostly for personal reasons, especially to manage their own collection, but they also tag to assist others to find a book. Once assigned, LT members use the tags to search and retrieve books, to gain information about books, and most importantly, to assist with personal collection management. Contrary to users of other popular tagging systems such as Flickr, LT members do not perceive social networking as an important factor when assigning and using the tags. The study also reveals that book tags by users vary greatly from MARC records. When tags overlapped with the terms used in MARC records, the majority of the matching occurred in the fields of 600s (Subject Access Fields) and 245 (Title Statement). The findings of the study contribute to the understanding of how books are represented in LT members' conceptual spaces, and have implications on how libraries could employ user created tags in OPACs.

Introduction

Tagging is the act of a person reflecting upon an object, then assigning free-form labels (tags) expressing ideas or notations the person deems relevant to the object. It is one of the features that are often associated with Web 2.0, promoting user contribution, and dynamic information creation and sharing. This social aspect of tagging emphasizes collaboration among users to create metadata for shared content, and is often referred to as “collaborative tagging,” “social classification,” or “social indexing” (Golder and Huberman, 2005; Voss, 2007). The vocabulary that results from collaborative tagging is referred to as folksonomy (Vander Wal, February 2, 2007).

To a certain degree, the process of user tagging is similar to traditional indexing practices; it depends upon the idea that information requires collocation by a person's decision about its content, potential usefulness, and relationships with other resources in order to be organized and findable (Grant, 2006). User tagging differs from traditional indexing practices by moving the power to make such decisions from library professionals into the hands of general Web users. Traditional subject cataloging is a top-down approach in which unbiased experts follow established rules to assign a limited number of relevant headings to an object (Peterson, April 2008). Completely cultivated from the collective intelligence of taggers, folksonomies are remarkably different from library classifications. Folksonomies are constructed from personal takes[?] on an object, reflecting user needs and how information is viewed, and are neither exclusive nor structured (Noruzi, 2007). In addition, tags are easy to enter and may constantly be added, keeping the tokens of folksonomies dynamic and up-to-date to embody the most current terminologies, expressions, and state of affairs. Consequently, folksonomies cover a wider breadth of topical “aboutness” than traditional cataloging (Golder and Huberman, 2006). Many researchers and practitioners have come to see tags as a great supplement to traditional indexes (Gordon-Murnane, 2006; Voss, 2007).

The popularity and use of tags have attracted those libraries that are moving towards a user-driven service environment. The Library 2.0 model emphasizes the importance in participatory services by enabling users to contribute to the information content of library collections, and thus providing personalized services and nourishing community engagement (Casey and Savanstinuk, 2006). Libraries have been taking the advantage of folksonomies to allow users to organize personal information spaces, provide tags to supplement existing controlled vocabulary and develop online communities of interest (Spietri, 2006). Many libraries have incorporated user tagging features into their Web systems. For example, the College of New Jersey uses del.icio.us tags as a part of traditional subject guides (http://www.tcnj.edu/∼library/moulaison/index.html). Ann Arbor District Library (http://www.aadl.org) and University of Pennsylvania library (http://tags.library.upenn.edu) implemented user tagging functions to their OPAC. Several libraries employ third-party services, such as LibraryThing (LT) for Libraries, to import user tags into their OPACs. As more and more libraries attempt to enhance their bibliographic records with user tags, it is essential to understand how user tags are generated and used, and to what extent the tags might enrich standard bibliographic records. As book tagging is a relatively young topic, there is limited empirical research in comparison with studies on photo tagging and social bookmarking (Angus, Thelwall, and Stuart, 2008; Golder and Humberman, 2006; Kipp, 2006, 2007; Lin, Beaudoin, Bui, and Desai, 2006). This study intends to fill such a knowledge gap by focusing on LT members' book tagging behaviors.

LibraryThing is an online service that allows its members to catalog books and create their own collections. Expanding beyond personal cataloging, LT for Libraries integrates member generated tags and reading suggestions into a given library's bibliographic record. According to WikiThing (n.d.), 49 libraries from across the world are now using the application. In consideration of LT's growing popularity and broad user base, the present study choose this particular site to investigate book tagging behavior.

The following are the research questions for the study:

  • What motivates LT members to tag and how do they use existing LT tags?
  • Are tags originated from tagger's knowledge of the book, current tag clouds, or other sources?
  • Are there statistically significant differences between user tags and the terms in MARC records?

Literature Review

Tagging is not a brand new phenomenon. Consumer products that have tagging or annotation functions appeared as early as 1988 (Vander Wall, August 30 2007). But it was not until the launch of the social bookmarking website del.icio.us (www.delicio.us) in 2003 that tagging gained momentum. Since then, there has been much discussion concerning the validity of folksonomies, their advantages and disadvantages, the possible uses, and how systems can better facilitate tagging. The following sections provide an overview of literatures on tagging motivations, tag uses, and comparisons between tags and controlled vocabularies.

Motivations for Tagging

What motivates the general public to embrace tagging so quickly? According to Hammond, Hannay, Lund, and Scott (2005), tags collected from different systems showed that there are two dimensions involved in the tagging processes: who the tag is intended for, and who the content creators are. Hammond et al's theory does not fully capture the dynamics of tagging processes and tag usages, some of which is explored in Marlow, Naaman, Boyd, and Davis's (2006) content analysis of the photo sharing website Flickr (www.flickr.com) tags. Marlow et al. suggest that there are two kinds of tagging motivations: organizational and social. For organizational purposes, users tag to help with future retrieval of the material. With a social motivation, users tag to contribute and share their collections with the community, achieve self presentation, attract attention, interact with other users, and express opinion. Social motivations were identified as the primary tagging reason for Flickr users in Angus et al. (2008). The preceding studies drew their conclusions based on the researchers' subjective categorization of the collected tags, which may not comprehensively or accurately reflect users' intents.

To gain understanding directly from the users, Ames and Naaman (2007) performed in-depth, semi-structured interviews with 13 active Flickr users. A set of two dimensional photo-annotation motivations was developed as a result. The first dimension is “sociality” - who the tags are intended for. The second dimension is “function” - what the tags' intended uses are. There are two types of functions: organizational and communicational. For example, under a self-organization motivation, a tag would be assigned to help the tagger retrieve the item in the future. Similar to Angus et al., Ames and Naaman found social organization – assigning tags to help other system users locate the item – to be the most frequently mentioned motivation among Flickr users. While the study provides a unique, in-depth look into tagging motivations through interviews, it is limited by the sample size and targeted only to Flickr users. Marlow et al. (2006) points out that different architectural designs may support or retrain certain features on a system. It is therefore important to investigate a different kind of tagging on a different system for the purpose of broadening the current understanding of user tagging motivations and tag functions.

Functions of Tags

Tags may be used in many ways by the tagger and by other users. Golder and Huberman (2006) indicate that tags are used in many ways, including describing the tagged item and providing personal notes. However, the authors also point out that tag uses are not as definitive as they appear. Personal tags such as “to read”, for instance, could hold some unexpected values to others. Sen, Lam, Rashid, Cosley, Frankowski, Osterhouse, Harper, and Riedl (2006) propose the following categorization of tags according to their functions: (1) factual tags identify the facts about the item; (2) subjective tags express user opinions regarding the item; and (3) personal tags are used for task management or information organization. Additionally, the study recognized users using tags to complete five kinds of tasks: self-expression, organization, learning, finding, and decision support. Sen et al.'s tag categorization and task groupings are the basis for the questionnaire design and data coding of the study reported here.

Tag Sources

The growth of a folksonomy occurs in a non-linear manner. Many researchers observed that it follows a power distribution with a long tail: the most frequently used tags dominate the total number of tag uses, and the less popular tags remain seldom used. (Lin et al., 2006; Kipp and Campbell, 2006; Munk and Mork, 2007). The stabilized use of popular tags is hypothesized as being caused by reasons such as: users associating an item in similar ways, shared knowledge within the community, and the act of imitation (Godler and Huberman, 2005). Munk and Mork claim that imitation is the main reason for the power distribution, and they further suggest that it may be driven by the theory of least effort. Investigating whether taggers are affected by what they see, Sen et al.'s (2006) survey showed that users are indirectly but conclusively impacted on by community generated tags, such as tag clouds.

Folksonomies versus Controlled Vocabularies

Formed freely by users and not regulated through any standards, the quality of has been questioned by a number of researchers. Spietri (2007) examined tags based on the National Information Standards Organization guidelines for the construction of controlled vocabularies. Spiteri found that while ambiguity, polysemy, synonym and basic level variations are prevalent problems, tags “conform closely to the guidelines for the choice and form of controlled vocabularies” (p.23). Spiteri recommended that with the help of written suggestions on the selection and formation procedure, tags could be incorporated into public library catalogs and increase the catalogs' user-friendliness and interactivity.

Several researchers compared tags to a set of controlled vocabulary. An example is Lin et al. (2006), who matched tags in Connotea (http://www.connotea.org) with Medical Subject Headings (MeSH) and title based automated indexing terms. The comparison is based on strict, binary matching; any variation in spelling or form is not tolerated, and would result in non-matching terms. The result showed that there is less than 20% overlap among tags and professionally assigned or automated index terms. Kipp (2006) chose a seven scale thesaural relations system to compare the tags with content creator assigned keywords, and professionally assigned index terms. While there are some thesaural relations among the different types of metadata, few exact matches among tags and system provided terms were found. Furthermore, the various parties each created an assortment of tags. Keywords and index terms include geographic descriptors and specific details that are not seen in tags. In contrast, tags include time management terms and generalities that are not in keywords and index terms. Kipp's discoveries suggest that there are noticeable differences between the users', authors', and professional indexers' views of the concept spaces in which an article resides, and that tags can provide additional access points to system provided ones.

Similar studies were conducted on book tags as well. Smith (2007) collected the most popular tags assigned to five particular books on LT, and analyzed their content in detail against the Library of Congress Subject Headings (LCSH). The results suggested that tags are better at showing latent subjects and are more linguistically forgiving, therefore, tags used in conjunction with LCSH may increase recall for natural language retrieval. Weber (2006) also used LT as the basis for comparison, but in a distinct manner. Instead of comparing tags and professionally assigned subject headings by books, he compared each of the 75 most popular tags and subject headings on the system. The results again illustrated the differences between user's perception of concept space and what is represented in traditional metadata. The results showed that while tags and subject headings overlap to some degree, tags include a wider array of bibliographic data while subject headings make finer distinctions among topics.

The aforementioned studies used a variety of comparison methods to examine the differences of tags and various professionally assigned metadata. This study looks beyond subject headings and compares user assigned tags with the terms contained in a book's MARC record to see if the differences exist there as well. The comparison is based on a fuzzy matching approach that allows for contextual similarities.

Methodology

LibraryThing requires only a user name and password to sign up for a basic account, and an additional fee for lifetime membership; it does not track the accurate number of individual members or their demographic profile. In this study, 1000 LT members who provided email address in their membership profile were randomly invited to participate in a Web-based survey. To encourage enrollment, members who signed up for and completed the survey were given a chance to win one of ten $30 Amazon.com gift certificates. By September 8, 2008, 104 LT members had signed up and 98 surveys were completed.

The Web-based survey consists of three sections: background information, reasons for tagging and tag usages, and tagging practices. The complete survey is available at: http://web.simmons.edu/∼tsai/BookTaggingSurvey/.

The survey used Ames and Naaman's (2007) sociality-function diagram as the basis for questions on tagging motivations, but added emphasis on communicational and organizational purposes for tagging. Sen et al.'s (2006) proposed five tasks for tag use were also used in questioning participants' habits of tag use. The tagging practice section asked users to tag three pre-selected books. The books selected for the tagging practice were from different genres, publication years, subjects, and perceived popularity. Participants were asked to follow LT's tagging conventions during the practice: tags can contain multiple words or symbols, but each must be under 30 characters long, and separated by a comma. For data analysis, the tags were processed and categorized by uniqueness and appearance in the content of the LC MARC record. All survey results were exported to Excel spreadsheets and SPSS for Descriptive and inferential statistical analysis.

Results

Description of the Participant Backgrounds

There were 50 male participants and 48 females, and 74.5% of the participants were under 45 years old. Over 84% received secondary and post secondary education. A small group of the participants (28%) have professional training in cataloging or indexing, about half of these participants (46%) also work in the library and information profession.

Most of the participants have a sizable book collection in LT. Approximately 75% of the participants have over 200 books in their LT accounts, and 33% of the participants have collections over 1000 books. Book tagging is a common practice among participants, 96% of the participants tag their books, and 43% of them tagged all of their collections. Aside from LibraryThing, 70% of the participants also use at least one other tagging website such as Flickr (36%) and del.icio.us (14%). The number of books tagged and the popular use of other tagging systems indicate participants' overall acceptance and comfort level with online tagging services and tagging.

Reasons for Tagging

The participants were asked to select or supply reasons that explain why they tag, and rank the top three most applicable ones. The data shows that 91% of the participants tag with specific goals in mind; for example, they tag to express an opinion or to organize a collection. On the other hand, a few participants (9%) tag because the system prompted them to do so. Figures 1 and 2 show the frequency distribution and ranking of tagging reasons as selected by participants as well as the “other” reasons specified. For a complete list of “other” reasons mentioned by the respondents, see Appendix A.

Figure 1.

Reasons for tagging.

Figure 2.

Other reasons as specified by Participants.

While multiple studies (such as Angus et al., 2008 and Ames and Naaman, 2007) have shown that social factors are the main tagging incentives some systems, LT members pointed out a different primary motivation for their tagging. Most of the participants (74%) indicated “collection management” - such as using tags to keep track of one's reading status or ownership - as a top reason for tagging. In addition, among the “other” reasons listed by participants, such as “to categorize the book by more detailed subjects” or “to record storage information”, most are also related to “collection management.” Together with “collection management”, “recording factual information” (62%) and “helping others find the book” (44%) are the top three cited tagging reasons. “Social networking” is identified by merely 18% of the participant as a reason for tagging. It is important to point out that tagging to assist in other people's book retrieval is different from “social organization” (Ames and Namman, 2007). Social organization occurs when tags are assigned to increase the findability of the tagger's own collection. When tags are added to help with other member's book search, the purpose is to serve as additional access points or to assist the recommendation algorithm link books. The tags are not assigned to promote a tagger's library or to garner attention. This is an altruistic reason for tagging that was not examined in prior research. This particular aspect demonstrates that LT members are conscious of the value of the tags in facilitating book discovery for the entire community.

Tag Uses

Participants were asked to explain how they make use of existing tags assigned either by themselves or others. Figure 3 displays the frequency of mentioning of different uses. For a complete listing of other reasons provided by the participants, see Appendix B.

Figure 3.

Reasons for perusing existing tags.

When specifying motivations for tagging, 77% of the participants assigned tags for “collection management”; similarly, 87% of the participants indicated that managing and keeping track of collections is a key use of existing tags. The 10% discrepancy between the number of participants who tag to organize their books, and the number of participants who use tags for that same purpose suggests that participants may come to appreciate the benefits of tags in collection management regardless of what the tags were originally assigned for.

Tagging Practice

The three books provided for the tagging practice were known to some participants but not all. The first title, The World is Flat, was heard of by a significant number of the participants. The second title, The Da Vinci Code, was the best known among all participants and read by 57% of them. The third book, Team of Rivals, was not known by most. The variations in familiarities provide insights into whether familiarity is a variable that alters one's tagging behavior. Figure 4 illustrates the participants' knowledge of the books.

Figure 4.

Participant familiarity with the books preselected for the tagging practice.

A total of 1,119 tags were provided during the tagging practices: 342 tags (30%) were assigned to The World is Flat, 398 tags (36%) were assigned to The Da Vinci Code, and 379 tags (34%) were assigned to Team of Rivals. Each tag that is different from the other tags applied to the same book is viewed as a unique tag. Spelling variations of a term, including singular or plural forms and typos, are not viewed as being different. Unique tags make up for 31%, 37%, and 19% of the tags assigned to the three books respectively. However, most tags (71%) are repeated appearances of the unique tags.

The participants were asked to categorize their tags into the three types (per Sen et al.'s (2006) classification): factual information tags, such as “cultural criticism,” “suspense,” and “Presidents”; opinion tags (referred to as subjective tags by Sen et al.), such as “horrible” and “not recommended”; and personal note tags, such as “unowned,” “to read,” and “have.” If a tag does not fit into the pre-defined categories, the participants were asked to give it their own category. All “own categories” showed to be comparable to the three pre-defined categories. For example, “examples of where the book might be” is a personal note, and “descriptive term about the author” is factual information. Participant supplied categories are therefore coded into the three given categories during data analysis. The most commonly used tags are factual information tags (75%), followed by personal note tags (14.8%) and opinion tags (7.7%). The t-test shows that factual information tags are more likely to be repeated than personal note tags and opinion tags combined (df = 323, t = 2.6, p<0.05). The different likelihoods of repeated use could be attributed to the tags' characteristics. Factual information tags are more objective in nature and are anchored to the book's content, physical traits or its author; personal note tags and opinion tags originate from objective personal experiences. Consequently, taggers may share more similar views on what consists of factual information than on the experiences and thoughts they associate with a book.

There are also significant differences in the numbers of the types of tags assigned to each book (one-way ANOVA, df = 4, N=325, X2=23.698, p<0.001). There are substantially more personal tags and opinion tags assigned to The Da Vinci Code than to the other books. In fact, 92% of the opinion tags from the survey are for The Da Vinci Code. As participants have a higher degree of familiarity to The Da Vinci Code, a one-way ANOVA test was performed to examine the impact of familiarity in tagging. The result suggests that familiarity has a significant impact on the number of opinion tags assigned (df = 2, F = 4.278, p<0.05), but not on the number of factual tags or personal tags. These opinion tags, such as “terrible,” “avoid,” and “not that interested,” are often used to express strong feelings towards a book, implying that they are more often used when a tagger has a particular view of a book.

Tag Sources

Participants were asked to identify where the tag for a book comes from: whether the tag expresses something they already know (from memory), or was inspired by other resources (such as LT provided information). All participants tagged books relying first on their memories. Memory served as a main reference source even for participants who were not familiar with a title. For example, five of the participants who had not heard of The Team of Rivals prior to the survey tagged based on their knowledge of what can be discerned from the title or author alone. Tagging suggestions are not provided on LT, and tag clouds are not readily present at the initial tagging interface, but participants actively seek out information when they tag. When faced with an unfamiliar book, participants are likely to look for other resources to supplement their lack of knowledge of the book. A one-way ANOVA test shows statistically significant differences in the use of memory, LT provided descriptions, LT provided book information, and LT tag cloud as reference sources by participants with different familiarities with the books (F-values are 276.33, 21.60, 11.74, and 9.18 respectively; the p-value are all less than 0.05). Other sources of information, such as Amazon.com and Worldcat.org, were also relied on for tagging.

Sen et al. (2007) and Lin et al. (2006) surmised that tag assignments are influenced by system suggested terms or tag clouds, because taggers would imitate what they see. Analysis of the degree of matching among the content of the consulted resource and assigned tags showed that this may not be the case for LT members (df = 292, t = −0.624, p>0.05). The t-test result suggests that while members peruse a variety of sources for tag ideas, they often come up with their own tags based on how they perceive the information. Matches among tags and MARC records or tag clouds are most likely coincidental.

Comparison with MARC Record

The tags collected from the survey's tagging practices were compared with terms used in Library of Congress (LC) MARC records and LC call numbers (as dissected using Library of Congress Classification Outline (Library of Congress, n.d.)). The MARC records and LC call numbers for each book were obtained from LC OPAC.

As mentioned in the methodology section, tags and LC MARC records are compared in a fuzzy matching approach. A match occurs when the tag and a MARC record term are semantically similar. For example, “Political leadership” and the tag “leadership” are considered a match. In total, 50 factual information tags are matched to MARC record terms, accounting for 14.2% of the total number of unique tags and 19.8% of the total number of unique factual information tags. It seems inevitable that matching only occurred with factual information tags, since MARC record fields are also about subjective descriptions of a book's contents and physical attributions. The MARC fields with matched terms are listed in Table 1. The topical subject entry field (650) is the field in which term matching most frequently happened. Other fields where matches recur frequently are the title statement field (245) and personal name field (600).

Table 1. MARC fields that were matched by tags.
original image

Presumably, participants who have experience in indexing and cataloging should have a higher awareness of how professional bibliographic records are constructed, and might produce tags closely corresponding with contents of MARC records. However, participants' background in traditional indexing and cataloging practices did not induce significant differences in and the degree of matching between their tags and catalog terms.

Discussion

Without the precise demographic profile of LibraryThing population, the survey used a random sampling in hopes that this study would include participants that best represent the general LT members. It is possible that members who have listed their email on their profile and responded to the survey are also more avid Web users and more active in LibraryThing. These characteristics may limit the representativeness of the sample to LT member community. Nonetheless, the survey produced some interesting findings.

With regard to the first research question, the survey finds collection management to be the top tagging motivation indicated by LT members. This confirms both Golder and Humberman's (2006) speculation that users assign tags mostly for personal uses, and Kipp's (2006) conclusion that management is an important reason for applying tags. A considerable number of participants in this study also designated “help others find the book” as a major tagging motivation, showing evidence that LT members recognizes the use and value of tags in browsing and information retrieval. On the other hand, the “social” purpose of tags that have a significant role in Ames and Naaman's (2007) annotation motivation taxonomy and in Marlow et al.'s (2006) discussion does not resonate as strongly with LT participants, even though LibraryThing identifies itself as a service that “connects you to people who read what you do” (LibraryThing, n.d.a). This suggests that users of different tagging systems use and apply tags differently. The tagging behaviors are based not only on the taggers relationships with the items (Golder and Huberman, 2006), but also on the system's main function and how it was designed (Sen et al., 2006).

As for tag uses, LT members use tags in ways that are similar to Sen et al.'s (2006) observation. Collection management was identified as the predominate way of using tags. The second most popular tag use is to gain insight on and find a book, demonstrating that LT members have noticed the significance of tags in book searching and discovery, and that folksonomies play an integral part in LT members' information searching process. This particular finding adds substance to Golder and Huberman's (2006) theory that although tags are often biased by personal views and contain personal notes, other members find unexpected values in them as search tools and information sources. Conversely, social features are found to be unimportant to the participants as before. The indifference could be attributed once more to the specific system architecture and members' perspective of LT's main function. LT is first and foremost “an online service to help people catalog their books easily” (LibraryThing, n.d.a). LT's ability to build a large user base that generated over 51 million tags so far (LibraryThing, n.d.b[a?]) could be a result. Another appeal could be the fact that LT members are able to add a book that is cataloged in Amazon or in any of the 690 libraries LibraryThing connected to. Library OPACs, such as that of Ann Arbor District Library's (AADL), might not be able to amass a comparable number of tags because the patrons are limited to tagging books within the library collection, and because it is not aimed to serve as a personal collection building tool. Libraries would need to provide patrons with more incentives and motivations to congregate and begin the folksonomy creation process.

Tags are most commonly generated from the participants' understanding and perception of a book. However, even with read books, LT members would seek information from other resources, sometimes taking the effort to search online in order to acquire tagging ideas. The findings suggest that users want to be informed during the tagging process, and that tagging systems can facilitate it by providing easily accessible information. Nonetheless, sources consulted have little impact on what the final tag choices are.

Kipp (2006) finds that time management (similar to personal note tags) and generalities (akin to factual information tags) are two types of tags usually unrelated to professional descriptors or author keywords. This study [which study, your or Kipp's?] discovers that personal note tags and a surplus of opinion tags are indeed unrelated to MARC records, but factual information tags are more likely to match the contents of MARC records. Nevertheless, the number of tags that match up with terms in MARC records remains low, and does not improve as a tagger's experience with cataloging or indexing grows. Tags of participants who use LT to manage books in professional settings or are skilled in cataloging and indexing do not have a higher match rate than tags of other participants. The results confirm Kipp's suggestion that tagging occurs at a more personal level, and embodies concept spaces that are different from what are represented in professional bibliographic records. User tags (such as “slavery” and “current affairs”) convey more subtopics, and reflect the cultural atmosphere (such as the tags “made into movie” and “overhyped”). Tags add to the richness of bibliographic records by enhancing professional records with user reflections and insights that are up to date and speak to current culture. The degree of differences between tags and MARC records also implies tags may resonate more strongly with members' takes on a book. Information professionals could learn through tags about how the general public associates a book, and could ascertain what important factors complement traditional book records.

Conclusion

This study finds that different tagging motivations may exist for different tagging systems for a number of reasons, and one would have to examine tagging systems of a particular type in order to construct a taxonomy that accounts for all tagging motivations. Although book taggers view personal use, such as management and organization, as the major reason for tagging, they also tag to assist others and help the system make correct connections among books. This tagging motivation may have transpired from participants' own use of tags as learning and finding tools. The finding attests to tags' potential usefulness in library catalogs and in the information retrieval process. On the other hand, social networking does not appear to be an important factor to LT members. The key tagging motivators and tag uses hint to aspects that users consider as most constructive in a system, and could help information professionals determine effective ways to promote tag creation. For LT members, collection management and the pooling of resources, such as LC and Amazon.com records, are important factors that contribute to the large collection of tags in LibraryThing. If a library is to accumulate a sizable folksonomy and make tagging a regular practice, it should start by making tagging beneficial to taggers in a personal way.

During the tagging process, most of the tags were derived from the tagger's existing knowledge and memory, but it is also common for participants to refer to other resources to develop accurately labeled, informative tags. This suggests that system designers ought to make the tagging process easier and help users to generate high quality tags by incorporating a variety of accessible reference sources in their systems.

The fact that user created book tags are significantly different from professional records is not surprising. The comparisons of tags and MARC record contents show that, regardless of background and experience, taggers make different choices about how to describe a book in simple terms. Tags add to bibliographic records by bringing in factual information (such as author background information) that is relevant but not included in the traditional records, by inserting personal touches through opinions or notes, and by adequately reflecting contemporary culture. The inclusion of tags in OPACs will not only increase access points to books, but also serve as effective reading suggestions from fellow users. Through tags, librarians would provide a new way to connect with their patrons and better comprehend how users think and act when searching for a book. It is apparent that there are several ways tags can be employed by libraries, and libraries will benefit significantly by integrating folksonomies into OPACs.

Although the chosen method of tag collection allows the researcher to observe the reasoning and judgments made behind each tag assignment, it is limited in that respondents were asked to tag books that they did not chose for themselves. A further study may include participants who have already provided tags to a book. While it might be difficult for LT members to recall the tagging process after the fact, the tags would accurately reflect those provided in a non-research setting. Another way to expand the current study is to adopt an extensive thesaural matching scheme, such as one used by Kipp (2006), to closely assess the relationships between professional index terms and tags.

Acknowledgements

The author would like to thank Dr. Rong Tang for guidance throughout the research process, ranging from her advice on the research planning and design to her thorough edits on the revised paper. The author also wishes to thank the four anonymous reviewers whose detailed comments greatly improved the presentation of this paper. The study was funded by a research stipend from the Simmons College Graduate School of Information and Library Science.

Appendix

A. Participant listed reasons for tagging

Categorizing and IR

  • 1.I find it really helpful to be able to organize my books online the way I do in my home
  • 2.I tag to make identifying similar books easier to locate and group in various ways.
  • 3.I use tags to categorize the book.
  • 4.I use tags to divide my books into subject categories
  • 5.I use tags to help myself find the book
  • 6.I use tags to sort books by category so that I can more easily locate volumes in a specific field
  • 7.I would normally tag by subject.
  • 8.stay organized and find information by category later
  • 9.type of book, category
  • 10.I tag books by genre and period classifications.
  • 11.the format of the item (audiobook, hardback, etc…)
  • 12.I tag with more specific content for the type of book that is more detailed than the title or publisher's infomation
  • 13.i tend to tag by a combination of content and author to simplify search within a specific subject
  • 14.I use tags to enhance findability via categorization. This is similar to “organization” but a bit more robust.

Physical Location

  • 1.My books are not all in my apartment – I use tags to specify location of the book.
  • 2.to help with the physical organization of my items in my home (location codes)
  • 3.I use tags to help me with finding a specific book, which might be at home or at my office, and if at home, could be on any of three floors! I suppose this is a subset of “organize my collection.”
  • 4.track location of books (my apartment, my parents' basement, loaned to a friend, etc.
  • 5.Remember shelf location
  • 6.I tag to help me find the book in my home library's book organization/storage/shelving system
  • 7.so that I can instantly generate a list of all the books I want to read, to pick a new book to read when I finish one

Book Source

  • 1.where I got the book

Management

  • 1.I tag to keep track of subjects for purposes of creating syllabi in classes I might teach in the future.
  • 2.I tag to assign a reading level to the book that is more detailed than the publisher information (Lexile.com score for children's books);
  • 3.what the book was used for. Some of my books were purchased/read for specific reasons, such as a class or a project.
  • 4.I also use tags to indicate which LT records need to be edited in some way…i.e. add a cover picture, or add more tags, or fix the author's name, etc…

Track Trend

  • 1.I use tags to identify trends in my behavior, for example “I didn't realize I had so many books on X, perhaps I should pay more attention to that subject.”
  • 2.I tag to keep track of my interests; for instance, I hadn't realised I was that interested in economics until I found a little cluster of books with that tag.
  • 3.I tag to track trends in my collection via the tagcloud
  • 4.so I can understand my collection
  • 5.I use tags to get an overview of collection contents and material
  • 6.know how much poetry, fiction, non- fiction, etc

Help System

  • 1.I tag so the social software will organize them into alike groups
  • 2.to help with recommendations of similar books
  • 3.I tag so LibraryThing can make meaningful connections between books.

Others

  • 1.I very rarely use tags
  • 2.I use tags to get to know a book a bit before reading it or putting it on the shelf
  • 3.I like tagging

B. Participant listed ways of using existing tags

Current Tag Use

  • 1.I don't really use the tags as of yet. Some of these choices I may do in the future, however.
  • 2.I don't think I usually use tags (or if I do, I don't think of it as tagging on librarything. On flickr, I use it to search for photos of a particular subject matter.)

Reference Other's Tags

  • 1.I read my LT friends tags to see how they categorize/organize their books

Keep Track of Book's Physical Location

  • 1.I use them to remember where I go tthe book
  • 2.I've also used them to record where a particular digital item is stored, with the number of a dvd or cd.
  • 3.To show which library branch owns a copy if it isn't a book I currently own.

To Compile Reading List

  • 1.making a list of books I want to read

Ancillary