Saturday, June 25, 2011

Implementing Location-Based Recommendation Services in Library Book Stacks

Jim Hahn

With the completion of a formative evaluation study of a collections-based mobile wayfinding application (Hahn & Morales, 2011) the need for recommendation services in libraries is uncovered as a desired and necessary component to library navigation services. To implement such a system, previous library and information science research on user context and system recommendations are reviewed for next directions for such services to mobile devices. Drawing heavily on already existing collocation attributes, this project will present a model for prototype implementation of location-based recommendation service that can offer greater access to print and electronic resources. Already existing iPhone Software Developer Kit (SDK) templates are leveraged for modeling data and interface prototypes. This paper reviews problems that will need to be solved before implementation of the production-level recommendation service and suggests possible implications which these location-based recommendation services may have on reference and instruction services.

1. Introduction

Printed material and digital content can exist side by side in the library. The virtual collocation of print and digital is possible and desirable to improve the experience of people browsing library book stacks. Collocating digital items alongside print can bring increased awareness of available information resources and promote a fuller understanding of what exists within the library collection.

A first-time library user trying to navigate a print collection will face an acute trepidation as described in (Mellon, 1986) as library anxiety. Recommendation services in library settings could help to alleviate the impairing aspects of library anxiety by delivering to the student suggested resources based on her physical location within the library. A mobile recommendation service lowers the barriers for receiving assistance at immediate point of need. Such a system is aware of the range of resources available to the student and suggests resources in a dynamic and responsive manner. A location-based recommendation service is feasible to implement in the near term, which is to say production systems could be online within a year or less.

Recommendation is a crucial and foundational component to information seeking. Librarians at the reference desk recommend resources to consult based on identified user needs. Further, the reference collection is also a part of the broader recommendation the library makes. The choice of the collection developer to include one item or type of item and not others is an act of recommending. Library services are systematically configured to recommend. One ubiquitous computing resource for which recommendation services have not yet been optimized is mobile computing technology.

The promise of such systems is realistic due to the advent of new mobile-networked technologies. With advances in wireless infrastructures that allow approximating the location of a mobile device with Wi-Fi triangulation methods as in the Skyhook service for wireless positioning (Skyhook, 2011) and software creation and customization with various device specific software developer kits comes increased opportunity to improve library services to mobile devices. The purpose of this paper is to inform general library practitioners of the opportunities and challenges that location-based mobile services present in the library setting. This paper will enumerate strategies for implementing a location-based recommendation service in library bookstacks. We focus on a case study that provides instructive lessons. The case study offers implications for reference and instruction service in contemporary academic librarianship.

Mobile technology product development is an area of ongoing innovation. Library patrons and library services can benefit from this field’s recent advances. The devices offer rapid, untethered connection to information resources. As associated hardware’s processing power increases there exists greater ability to manipulate and process data and documents from increasingly lightweight and powerful mobile computing tools.

Recently the two most significant entities in mobile software, makers of the iOS (Apple) and the Android (Google) systems have released new offerings for tablet computing. Emerging tablet computers offer users a greater reading experience than smaller mobile devices. The maturation of tablet devices is especially important in the library context, as it has been found that on small-format devices such as mobile phones, reading comprehension is about half that observed on desktop systems (Nielsen, 2011).

Librarians have a sense that today’s rapidly changing technological landscape should be reflected in the services they provide. But while enthusiasm and curiosity are in abundance in the library technical field, consensus on precisely where and how to merge library-specific expertise and emerging digital tools remains elusive. While locations outside of the library environment are a rich area for library service innovation, the scope of this project is limited to recommending resources to students in library book stacks. The print book stacks offer a fascinating research area for location-based services and mobile technologies, largely due to their intellectual organization by shelf classification and already existing collocation attributes (Svenonius 2000, p21-22); essentially, the experimental location-based recommendation service described in this paper is grounded in the advantages of collocation that support information discovery.

The goal of this paper is not to present original research. This paper synthesizes a design framework for library staff that plan to provide access to information by way of mobile platforms. The report that follows can act as a blueprint in making design considerations for location-based library service. What follows is a lessons learned report of data and interface modeling and a suggested architecture for bringing experimental location-based recommendation prototypes online.

This paper begins by reviewing previous work on contextual information access and library implementations for recommendation service to mobile devices. The next section models the development of a prototype for implementation. Existing iPhone Software Developer Kit (SDK) templates (Apple, Inc., 2011) are leveraged for modeling the data and interface prototype examples. This paper concludes with a treatment of problems that will need to be solved before location-based recommendation services see production-level functionality. Final sections of the paper detail the ways in which recommendation services may impact how librarians do reference and instruction and crucially, how students will learn in the library with this new service.

2. Literature Review

In laying out a research agenda to understand the uses and components of library collections in a networked era, it has been argued that a collection is essentially an information context (Lee, 2000, p1112). The issues of tangibility associated with collections are of primary concern to my work here. Lee comments that collections are both tangible and intangible resources (p1108). The departure point for this project is to consider how those intangible parts of library collections might be more closely associated with the tangible; and to increase access to both.

Recommendation has varied and multiple research approaches. This project centers on collections-based recommendation; essentially treating the users location in the library book stacks as user query for relevance. Jonathan Furner (2002) reviews the development of recommendation services as evolving from work related to collaborative filtering. Furner posits a distinction between two broad types of recommendations which are either system derived recommendations or recommendations seeded by users of the system (p748). As an example of system derived recommendations consider those information systems which return results based on usefulness to the user – the online store at say Apple, Inc. which suggests peripherals for the computing resource the user intends to purchase. Examples of recommendations seeded by users of the system include Amazon-like environments where users are recommended items based on previous user purchases.

The recommendation problem is crucial for library settings. Examples of library recommendation systems from a library OPAC include the BibTip implementation developed in Karlsruhe University (Monnich & Spiering, 2008) this system is derived of three agents incorporating OPAC observation, an aggregation agent, and the recommendation agent (Monnich & Spiering, 2008). The architecture for the recommender system at Karlshruke is reported in Geyer-Schulz, Neumann, and Thede (2008).

Solomon, in Discovering Information in Context (2002), remarks that those structures of classification and categorization that are designed to aid in information discovery often fail in discovery (p240). His analysis of the structural ecosystem of the information context is particularly relevant to the location-based recommendation problem here, since a number of structures such as physical environment, staff situation, and user situation will affect contextual discovery (p241). The structural critique by Solomon suggests that structuring will promote certain views while obscuring others (p232). This structural analysis is an important consideration in the design of libraries since the structures in place in these settings may not be given analysis and prioritization for re-engineering for recommendation.

Goker et al. are concerned with “Context and Information Retrieval,”(2009). They argue that the concept of user context is not only crucial for information retrieval (a well studied position), but especially so in the context of mobile information management. Their treatment is instructive for librarians in that the authors present us (the community of librarians) with a model which we may base designs for contextual information access: “Context-aware information delivery methods use information about the user’s current situation as a means to deliver relevant content to that situation” (p133). The notion of context is a multidisciplinary area of study (Goker, et al., 2009, p132); “it has a range of meanings in information and computer science where it is also used to describe aspects of human-computer interaction and elements in natural language processing” (Goker, et al., 2009, p133).

A selection of research articles on location-aware and contextual recommendation services (Adomavicius, G., et al. 2005; Henricksen, et al., 2006) is indicative of the many and varied epistemological approaches to creating mobile information services. These approaches marshal a variety of evidence, theories and technologies. For example, some systems examine search logs for evaluating contextual attributes of search (Zhang & Kamps, 2010). Others rely on location sharing in social networks, studying the interconnections of physical location and user profiles (Cranshaw, et al., 2010).

Linda Schamber, in a chapter from the Annual Review of Information Science and Technology on Relevance and Information Behavior, remarks that the models and contexts of relevance are “… entirely dependent on the context in which a judgment is made, the possible approaches to models and theories involving relevance are extraordinarily diverse,” (1994, p28). We must attend to the situation of the user in the library book stacks, or in the research trajectory sketched out by Schamber, the project at hand is specifically concerned with user criteria for relevance with an eye toward those “dynamic interactions,” of users, since as she indicates, “…relevance relates to people and their constantly changing perceptions,” (p35).

A definition of user needs within the bibliographic sphere is found in the Functional Requirements for Bibliographic Records (FRBR). FRBR is a bibliographic ontology of the universe of bibliographic access; it is a powerful conceptual model for the LIS community (IFLA, 1998). The FRBR specify four user tasks for bibliographic records: find, identify, select, and obtain (IFLA, 1998, p79). This chain of activities defines what it is users would do with bibliographic data in library settings. This project is centrally concerned with user tasks of navigate and obtain.

The obtain task suffers from inadequate design support as evidenced by many users who know a book is in the library but lack sufficient bibliographic skills or information to adequately retrieve the desired item. Recommendation services in particular also may help support the proposed navigation task, advocated by Svenonius (2000, p20) as being necessary yet missing from the four original tasks, where the navigation objective is needed “… to find works related to a given work by generalization, association, and aggregation; to find attributes related by equivalence, association, and hierarchy.” The associations that can be modeled into a recommendation application will be explored further in the project logistics portion of this paper.

The online public access catalog is a subset of the bibliographic universe. This subset can be extended with the recommendation of resources that are not traditionally associated with the user search experience. Traditionally, the public access catalog is a closed environment. The catalog intentionally omits data that lies out of the collection at hand; this precludes inclusion of external data such as geo-referenced information. While the catalog could make reference to a conceptual location such as the third floor of a given building – these locations are not typically associated with latitude, longitude, and elevation coordinates. A key argument in this paper is that physical space in a library is meaningful. This meaning is underscored by the arrangement of books on shelves according to subject-driven classification schemes. Helping users navigate the library’s physical space will also help them navigate the collection’s intellectual space.

A recommendation system tied to specific subject areas and those coordinates of library categorization could also incorporate content from a range of sources across the web, such as Amazon recommendations or Netflix suggestions. These suggestions would be filtered based on location of the client device (the location of the user) in the library. This mobile computing approach brings about a deeper collocation of digital content with the physical library navigation experience. The recommendation service proposed here will provide increased access to both the print and digital resources a library offers (and form interconnections of the two) students could therefore be exposed to the breadth of library collections, digital and print based on immediate need and interest; indicated by student location in the book stacks.

3. Mobile services

While the application of mobile technologies to problems of information seeking and access in libraries is new, several projects of varying levels of maturity have already been launched. This section outlines several of the most visible and influential of these.

A recent approach to mobile computing-based book recommendation in a library setting was developed at the University of Oregon. The book recommender suggests books through a mobile interface. The Book Genie1 service is not device specific, so while it was developed for mobile access, it can also be accessed from a desktop browser.2 However, Oregon’s system focuses on award winning lists and other elements that may not relate to the users immediate information need while omitting consideration of geo-referenced location information. Their recommendations are not tied to geo-referenced data, and omit design considerations for immediate user preferences.

The Orange County Library System Shake it3 app is a book recommender that allows users to search by filters just by shaking the phone. By shaking the device a user is given search results of recommendations based on specific genre, format or audience type. As with the Oregon State Book Genie, there are no geo-referenced data associated with the recommendation. In other words, the app is not designed to suggest material based on location. The designers view this service as another way of exploring content in the catalog based on user defined filters and allowing the option to browse titles from the application interface or choosing to view the catalog entry from a mobile safari browser.

Outside of the library, the Wolfwalk application from NCSU4 features “a location-aware interface,” which uses current location information on the mobile device to suggest digital information about the surrounding campus environment. The digital resources Wolfwalk recommends are archival images and historical information about the campus. This is an example of reusing library data to create a location-based service. However, the space outside of a library setting includes multiple layers of data points to which library resources may be relevant. Given such an open environment over which to recommend, this research strand represents an area of multiple unknowns; such as knowing the information needs of users in the campus environment, the individual users’ familiarity with campus history, other dynamic (changing) elements of context on the NCSU campus that will influence a users experience in pedestrian navigation of the campus environment.

My previous work on the experimental app Library Helper5 for the Undergraduate Library at the University of Illinois Urbana-Champaign studied how students would be guided to the location of books in the library based on device localization. Of the formative evaluation comments collected, students reported wanting to know other resources in the area rather than simply locating one book; for example, students suggested the need for a service which helps navigate to items in the collections which are current New York Times best sellers, or those books that related to the individual student’s major. Students also asked if the app could recommend books based on items charged to their library account (Hahn & Morales, 2011). With the completion of our formative evaluation study, we now require a synthesis project such as this to sketch out design considerations for the ongoing iteration of wayfinding applications. Previous work has led us to consider that pure wayfinding is not enough. Students’ previous experience in online commerce (Amazon, in particular) influences the expectation for recommendations from a library navigation application.

In Amazon-like environments consumers want to know values of popularity of resources; this includes knowing which books in the library have the highest levels of circulation for a given time interval. Filters could be built which show the most popular resources by month, year, and overall (since statistics began) and these filters should be overlaid for a specific topical area too: i.e. showing the student what the most popular areas are for Literature, History, Economics or for specific information types, such as media or books on CD. Another data element to include would be a data element of popularity rankings from Netflix. One area that is particularly popular for the Undergraduate Library is the media collection; various film awards should be associated and geo-referenced for student browsing.

A location-aware book recommendation service that most closely resembles the work here is the example from C. Chen, & Y. Yang (2010) where the problem based learning approach is modeled into the service architecture (p855). This architecture recommends resources based on a “Problem-based Learning Agent,” comprised of data drawn from a “user portfolio database, learning task database, and a login account database,” (p855). It should be noted that this model was implemented in a vocational high school library in Taipei, Taiwan. Such academic-history based suggestion service based on student learning portfolios is not possible on such a large scale since the population of new undergraduate students at a research university makes such individual learning history prohibitively intensive for the library to maintain. Any kind of recommendation service implemented for larger research libraries will have to leverage already existing student profile data (courses enrolled, major of study) and make use of available data points from the user’s library account if the student chose to opt-in to such a service.

4. Use narrative for mobile recommendation service

This section outlines hypothetical use cases to help readers understand the goals of the current project. We start with a hypothetical given: the phone’s positioning software application is able to guide a student to the location of a known item in the stacks. After this initial navigation, the student expects that a wayfinding system should guide her to other books in the library that are relevant to her immediate information need, a need which may have shifted since entering the stacks. Further, she may now be interested in those resources that are available digitally. But navigating the print stacks does not currently support location-based discovery of relevant digital resources such as e-books and the full-text of scholarly articles indexed in online databases. The recommendation service therefore should instantiate digital resources into the physical space which new students will overlook if they concentrate solely on a browsing based search of the library book stacks.

A student’s major could be incorporated as an element of context, since the students major will impact the individual’s frame they bring to browsing the stacks. Previous wayfinding studies have uncovered that students prefer to find books that directly relate to their chosen major and current course assignments (Hahn & Morales, 2011). The system architecture could support various filters; where one filter a student could overlay onto her search would be assignment specific navigation of the book stacks.

A personalized recommendation service may not adequately meet all use needs. Personalized recommendations may unintentionally omit items that lie outside of interest models. This factor will need to be studied closely in actual use cases to gauge the need to correct for or supplement personalization filters.

Librarians who work service desks observe students who frequently are in need of information resources that are immediately accessible. Students will prefer to access a digital copy of an article to the print copy when constrained for time. This places a constraint on search, such that the book must be nearby and the content of the articles must be full-text. The time context of the user should be inclusive of resources that are immediately available, where the architecture acts to filter out search results for books that would need to be requested from other campus locations and that are currently checked out. The proposed system takes account of the user’s temporal context to filter databases that do not provide full text access.

A data point to model in the database of suggested digital resources would be to classify those indexing and abstracting services for a key that indicated that these resources provide full text, immediate access. In terms of building hours, there should be recommendation services that are aware of multiple campus locations of titles, and should the user require immediate access to books, library collections which are not open or not quickly available should not be suggested as other available resources for students.

5. Project Development Logistics

Test site

The experimental setting for this project is the Undergraduate Library at the University of Illinois. This library primarily serves undergraduate students and features a print collection of about 130,000 monographs along with a popular open media collection of 19,674 DVDs and 12,142 VHS tapes. Items in the collection are a rich source of data for a location-based service. My argument is that the geography of the library is intellectually informative and can be a rich source of data for which to make recommendations; specifically, the individual’s position within the stacks is a query point. This query point relies on the relations with nearby classified resources as well as individual student interest. Motion of the user is important to underscore here since it is indicative of possible shifting user interests and needs within the library book stacks.

Developer environment

Software programming is not a widely held skill for librarians. For prototyping this recommendation service a web-based approach is advocated. This web-based approach would alleviate high-barriers of entry for general library practitioners. Librarians who have created web pages and who understand basic web design principles will find that Android or iPhone application development can be undertaken using web technologies, such as HTML, JavaScript and CSS. These platforms insulate the developer from writing code in compiled languages such as java or objective C.

For example, the Dashcode module of the iPhone Software Development Kit (SDK) uses web scripting (JavaScript) and web markup (XML or JSON) for creating applications. For most generalist librarians the use of web authoring tools is a core competency. Apple’s developer environment will provide dynamic JSON and XML data such that the web app would only require a web address (URI) for data. The proposed API architecture is modeled on the paper, “A Web Service Platform for Building Interoperable Augmented Reality Solutions,” (Belimpasakis, et al. 2010, p2) as shown in Figure 1.

Figure 1. The proposed architecture of the API for positioning and recommending resources in the library. The positioning service detects the client and delivers an XML string of detected Wi-Fi Access Point. The recommendation service matches this location with subject area of surrounding bookstacks.

The implementation uses RESTful (Representational State Transfer) web architecture; an architecture that relies on the architecture of the world wide web for implementation, and which further lowers barriers to entry for software development. Web service APIs make possible a recommendation system for any device that can communicate by way of standard web protocols (like HTTP) it allows the library to design for all devices rather than any specific software developer environment. If a mobile device can access web content, then it is able to access the recommendation and positioning web service. This would make the designed services accessible to the broadest range of diverse devices and mobile operating systems. Devices specific applications can still make use of HTTP requests by way of the device specific programming library for HTTP (Richardson & Ruby 2007, 24).

The broad outline for creating a web service with a RESTful architecture is as follows:

“Every web service request involves the same three steps:

1.Come up with the data that will go into the HTTP request…
2.Format the data as an HTTP request…
3.Parse the response data—the response code, any headers, and any entity-body—into the data structures the rest of your program needs.” (Richardson & Ruby 2007, p24-25)

This RESTful approach to building web services for mobile applications is detailed in a research article, “A web service platform for building interoperable augmented reality solutions,” (Belimpasakis, et al. 2010). Their paper explains an HTTP approach to building augmented reality systems, describing the architecture of the web service, which is developed using RESTful web services (p2).

Prototype implementation

The previous section features higher-level design ideas. This section will detail the step-by-step Dashcode workflow that I will use to implement subject-based recommendations. This workflow is informed by the research into best practices described above, and is intended to serve as a guide for other practitioners who hope to implement similar services.

To summarize the overall design:

1.A primary index of Wi-Fi regions to broad subject collections is modeled with a JSON array
2.A second JSON array is created to model associations of the identified subject with relevant books and electronic content
3.The JSON array is placed into the Dashcode environment (Figure 5)
4.The JSON data is linked with the application interface
(illustrated in Figures 6 & 7)

We begin the prototype with information modeling; to model the recommendation system, the library bookstacks will be divided into regions based on Wi-Fi data as shown in Figure 2. Previous wayfinding work (Hahn & Morales, 2011) used this methodology of dividing the library into regions; we find it to be a proven practical starting point.

Figure 2. Wi-Fi access points are assigned regions. Regions are further associated with subject areas. Collections map not to scale. In a rapid prototyping study on wayfinding in libraries with mobile technology it was found that CAD maps viewed from an Android developer phone were not useful for students to navigate the Undergraduate Library; a map with distorted scale and collections was preferable and increased student wayfinding to known items (Hahn & Morales, 2011).

In order to create an index to match the identified access point with a subject area we will associate the Wi-Fi access point regions with subject areas using a JSON array such as the information show in Figure 3.

Figure 3. Note that regions can include more than one subject area and some subjects may be found across different regions, i.e. Biography exists in regions 9 and 4. The Wi-Fi Access Point API will detect the user as being in a specific region. The API will then specify which Access Points the user is “connected to,” with respect to physical proximity.

In order to show the functionality of suggesting books and databases we assume that the client mobile device is detected in region 6. Location and primary level of recommendation data are linked by subject area to recommended resources (print or electronic) as shown in the subject area JSON array, Figure 4.

Figure 4. Subject areas to recommended resources.

The modeled arrays can then be incorporated into the Dashcode developer environment as shown in Figure 5.

Figure 5. Note that the red arrows show the web component files that are generated by Dashcode. Since this is a prototype or model, we are working with static data elements. In a production service this data would be generated dynamically based on approximate user location.

Figure 6. A graphical representation (based on our model above) of the subject area JSON data structure. Based on approximate location of the mobile device, the generated JSON can be linked to the subject area in the app interface using the dataSource tool within Dashcode.

Figure 7. A model for the association of JSON resource type (print or digital resource) data to the interface.

I make the assumption that an efficient mobile recommendation system will rely on the soundness of its underlying data. Designing the prototype system requires modeling associations for connecting digital and print recommendation. This primary information modeling happens before construction of the user interface; modeling associations for recommendation requires us to first annotate a geographic approximation of the library collection (Figure 2) and describe that annotation with JSON markup (Figures 3 & 4).

I then incorporate the JSON arrays into the Dashcode developer environment (Figure 5). Dashcode takes the textual representation of the associations and then represents the data model graphically. This re-presentation of the data is just a Dashcode specific operation, it does not change your data model; but rather causes the software to understand the relations you have modeled. Finally, I linked the modeled data to interface display (Figure 6) and interface performance (Figure 7).

As an important consideration in the efficiency of library systems, the modeled associations should be re-usable for other future library location-services. If JSON remains a web standard for the foreseeable future, the data models could be useful for other web based services. It is desirable that at some point in the future this work could also be implemented in the online catalog and online records could be enriched with geographic associations.

To summarize the overall operation of the recommendation app:

The Wi-Fi access point closest to the user is identified. Based on this, the location of the client device within the library is approximated.
Identified subject area is presented to student.
Student can then choose the type of information resource desired (print resources or digital).
Student is presented with suggested resources.
Student can continue to pan down for more suggestions as shown in Figure 9.

After modeling and connecting the associations to the interface, the generated web files are made accessible on a prototyping server: that can be viewed from a mobile Safari browser: When loading the HTML of the first level page, we see the following primary interface, Figure 8.

Figure 8. Appearance of the app on the student’s iPhone. This figure is based on a scenario where the student device is detected near the Science subject area and can now choose other resources based on his or her information need within that context.

Figure 9. Item suggestion interface. This model was created with the Dashcode RSS template and data from the University of Illinois Library new titles RSS feed.

6. Future Work

This paper has detailed problems that will need to be solved for production-level implementation and practices based on prior projects that can guide as we work towards solutions. In this section I summarize additional challenges that practitioners (including myself) must negotiate in order to create production-level mobile recommendations in library settings.

Data feeds: online catalog and digital content

For a production level implementation the service still needs to be connected with OPAC data of subject areas, inclusive of both print and electronic sources. While this data does not yet exist in one index it will be preferable to pre-process all of the recommendation data into an index (the Recommendation Service entity modeled in Figure 1) that can be queried over subject associations modeled here; i.e. by subject area: print or digital resources.

Approximate location and user motion

The approximation problem includes solving issues such as granularity of knowing where a device is in the building. The prototype implementation described in the methods essentially uses broad approximations: the connected access point will only give approximate location of the device; further, we need to factor in the probability that the user of the client device will be in motion and the system will require low latency if precision of approximate location must be known on the order of seconds.

Administrators of library implementation will want to consider how abstract or precise their specific implementation of location should be. It may be up to the individual user to declare how precise they want the system to be able to identify their locations. The library can start with broad regions, and then move toward increasing levels of granularity, depending on needs and system objectives. Additionally, library administration will need to think about security: making sure the system never records logs of individuals, or that those with administrative rights to system information are never able to associate an individual with detected device.

Should there be a sharing mechanism for location, that option should include an opt-in sharing of location: if individuals wish to share previous query/location information that should be done as an optional feature. It should be noted that since this is an application that the user chooses to install on their mobile phone there is no building-wide tracking of patrons in the library, but rather a service the patron has chosen to access from their phone.

Mobile computing

There also exist technical problems inherent in smaller computing devices. Questions that will need to be examined include asking database providers if digital content is formatted for mobile display. Further investigation about clicked through resources and mobile display will be necessary. Depending on the underlying structure of the digital resource, XSLT transformations along with relevant style sheets can help to alleviate this display problem.

Implications for reference and instruction services

When location-based recommendation services become available for students, there will be implications for other traditional services of the library. Traditional reference service can be tailored to more specifically address the in-depth questions that students will have. If a library puts into place location-based recommendation service, librarians can expect to help users with queries that would perhaps be much more research intensive—users will come to the research desk with increasingly sophisticated queries. It may be the case that librarians will see a decrease (or perhaps more challenging types) of known title searches, or readers advisory questions may increase. Perhaps such a service would lead to librarianship that may be considered more fulfilling or at least more challenging – since directional questions based and shelf browsing questions (while important and fundamental) are rarely the most fulfilling aspects of public service work.

The reference desk expertise of locating relevant information based on identified sources can be incorporated into the system where librarians and teaching faculty can also recommend resources by course or by assignments. The modeled associations for relevant databases will need the relevant expertise of librarians who will want to be sure that the correct digital resources are associated with subject areas of the stacks. Within the realm of classification there may be implications for the descriptions of electronic books or other digital tools that may not be classified for the shelf, since they are not traditionally thought of as requiring shelf collocation. An app such as this does make libraries re-examine what it means to browse collections and the line between digital and print may become an antiquated concern if digital can be instantiated into the print browsing experience.

Library instruction services can be made more effective with location-based services. Those students learning styles that more closely align with hands-on and active simulations will benefit from tools that help them become aware of library collections in an interactive manner. With location-based service the user has an additional tool for beginning research; previous Information Literacy Research (Head & Eisenberg, 2009, p2) tells us that the single defining attribute of research in the digital age is that our students feel overwhelmed with knowing where to begin class research. A recommendation service such as this essentially creates various subject specific filters that may help to streamline the student search process while navigating the print collection. A further way to study such a service would be to observe and interview small groups of students working on a research paper with the aid of a recommendation application. Gathering this initial qualitative data would help us further understand the implications for learning in the stacks with mobile recommender tools.

7. Conclusion

This prototype project outlines an area of profound change in how libraries will provide access to print and digital resources. Print collections can be given new life and renewed importance by way of digital mobile technologies. Integrating new technologies with foundational library services and apparatuses such as traditional bibliographic organization lends the primary motivation to this research. As the project described in this paper moves forward, I will need to consult with campus IT departments for available Wi-Fi data sources. Lessons from these consultations will inform future results, with an eye towards offering continuing roadmaps for other practitioners. Web-based approaches to mobile application development is shown here as an accessible and interoperable design strategy.

8. References

Adomavicius, G., et al. (2005), “Incorporating contextual information in recommender systems using a multidimensional approach,“ ACM Transactions on Information Systems, Vol. 23 No. 1, pp.103-145.

Apple, Inc. (2011), “Mac OS X – Developers,” available at: (accessed April 27, 2011)

Belimpasakis, P., Selonen, P., You, Y. (2010), “A Web Service Platform for Building Interoperable Augmented Reality Solutions,” International AR standards Workshop Oct 11-12 2010, pp.1-5.

Chen, C. & Yang, Y. (2010), “An Intelligent Mobile Location-Aware Book Recommendation System with Map-Based Guidance That Enhances Problem Based Learning in Libraries,” in Z. Zeng & J. Wang (Eds.): Advances in Neural Network Research and Applications: Lecture Notes in Electrical Engineering, Vol. 67 pp.853-860.

Cranshaw, J., Toch, E., Hong, J., Kittur, A., and Sadeh, N., (2010), “Bridging the Gap between Physical Location and Online Social Networks,” In Proceedings of the 12th ACM International Conference on Ubiquitous Computing, Copenhagen, Denmark, September 2010. pp. 119-128.

Furner, J. (2002), “On Recommending,” Journal of the American Society for Information Science and Technology, Vol. 53 No. 9, pp.747-736.

Geyer-Schulz, A., Neumann, A., Thede, A. (2008), “An Architecture for Behavior-Based Library Recommender Systems,” Information Technology and Libraries, Vol. 22, No. 4, available at: (accessed April 23, 2011).

Goker, A. Myrhaug, H. and Bierig, R. (2009), “Context and Information Retrieval,” in Goker, A., & Davies, J. Information retrieval: Searching in the 21st century, Wiley, Chichester, UK, pp. 131 - 157

Hahn, J. & Morales, A., (2011), “Rapid Prototyping a collections-based mobile wayfinding application,” Journal of Academic Librarianship (37) 5 (forthcoming).

Head, A. and Isenberg, M., (2009), “Finding Context: What today’s college students say about conducting research in the digital age,” Project Information Literacy Progress Report, University of Washington’s Information School, available at: (accessed March 31, 2011)

Henricksen, K., Indulska, J., Rakotonirainy, A. (2006), “Using Context and preferences to implement self-adapting pervasive computing applications,” Software: Practice and Experience, Vol. 36 No. 11-12, pp.1307-1330

IFLA (1998), Functional Requirements for Bibliographic Records – Final Report, available at:
(accessed March 24, 2011)

Lee, H. (2000), “What is a collection?” Journal of the American Society for Information Science, Vol. 51 No. 12, pp.1106-1113.

Monnich, M. & Spiering, M. (2008), “Adding value to the Library Catalog by Implementing a Recommendation System,” D-lib Magazine, Vol. 14 No. 5/6 ,
available at: (accessed April 23, 2011).

Mellon, C. (1986), “Library anxiety: a grounded theory and its development,” College & Research Libraries, Vol. 47 No. 2, pp. 160-165.

Neilsen, J. (2011) “Mobile context is twice as difficult,” Jakob Nielsen’s Alerbox, available at: (accessed April 27, 2011)

Richardson, L. & Ruby, S. (2007), RESTful web services, O'Reilly, Sebastopol, Calif.

Skyhook (2011), “How it works > Overview,” available at: April 27, 2011);
Schamber, L. (1994), “Relevance and Information Behavior,” In Williams, M. (Ed.) Annual Review of Information Science and Technology, Vol. 29, Learned Information, Inc., Medford, NJ.
Solomon, P. (2002), “Discovering Information in Context,” In Cronin, B. (Ed.) Annual Review of Information Science and Technology, Vol. 36, Information Today, Medford, NJ.
Svenonius, E. (2000), The Intellectual Foundation of Information Organization, MIT Press, Cambridge, MA.
Zhang, J. & Kamps, J., (2010), “Search Log Analysis of User Stereotypes, Information Seeking Behavior, and Contextual Evaluation,” In IIiX '10: Proceeding of the third symposium on Information interaction in context, pp. 245-254

No comments:

Post a Comment