Abstract

The design of usable user interfaces for digital libraries (DL) is a complex task that requires knowledge and guidelines on user-centred design. Patterns may be one way to achieve usability when designing and developing user interfaces. This paper presents a way to identify user interface patterns for digital library applications and demonstrates preliminary results.

Keywords

User interfaces, patterns, user centred design process

1 Introduction

Designing user interfaces for digital libraries is still a complex challenge; it requires knowledge about the users, their tasks, the context of use, and what is technically feasible. Developers and designers need to provide ways to structure and present information, efficient functionalities to search and browse information, and tools to work with information like readers or multimedia players [1]. The implementation of next-generation digital libraries includes annotations, multi-cultural and multi-lingual support, personalisation techniques, and recommender systems. In the future, concepts like “living” documents with comments and annotations by users and reviewers will add to the complexity of digital libraries [2].

As Blandford et al. explored in case studies, users find digital libraries hard to use, not just because the task of information seeking is difficult, but because the interfaces are difficult to use [3, 4]. Thong, Hong, and Tam conducted interviews with 397 students at the Open University of Hong Kong, and they drew the conclusion that user interface design was a significant determinant of perceived quality of digital libraries [5]. Other studies and surveys confirm these findings [6] and Arms states that “a digital library is only as good as its interface” [7]. Hence, there is a necessity to cope with this multi-dimensional complexity and to help developers to improve the usability of DLs’ user interfaces. Guidelines can help designers making sound decisions. They exist in a variety of forms – high level principles, low level detailed rules, standards or style guides. Some examples of guidelines applicable to DLs’ user interface design are:

• ISO 9241-151 [9],
• Web Content Accessibility Guidelines (WCAG) [10],
• Guidelines by the U.S. Department of Health and Human Services (HHS) [11],
• Minerva Cultural Website Quality Principles,
• Guidelines by Elsevier User Centered Design Group [12].

But the level of abstraction and significance of guidelines may be quite diverse. For example, “Ensure that the results of user searches provide the precise information being sought, and in a format that matches users’ expectations” [11]. This is quite ambiguous and can be applied only by a user interface design expert. “Provide explicit hints for kinds of reformulation other than simply adding new keywords” [13] is much more tailored to the context of use of digital libraries, but can be interpreted in different ways and does not supply detailed information how to fulfil this requirement. So as guidelines are abstract and suffer from inconsistency and ambiguity, their utility is questionable.

In order to overcome this limitation of guidelines, more and more researchers devote themselves to the development of patterns and pattern languages [15, 16 17 19 21]. Patterns present structured textual and graphical descriptions of proven solutions for recurring design problems in a specific context of use. They provide more useful and specific information as well as best practice solutions and can be a helpful reference point for those developers not being experts in user interface design. Welie points out the differences between guidelines and patterns: “In a sense, patterns represent pieces of good design that are discovered/uncovered empirically through a collective formulation and validation process, whereas guidelines usually are defined normatively by a closed group” [16].

2 Research objectives

The objective of this work is to describe a structured pattern collection providing proven user interface solutions for recurring design problems in the context of digital libraries. Usability is the criterion that defines a proven solution. That is to say, a proven solution enables typical users to achieve their goals with effectiveness, efficiency, and satisfaction. Proven solutions are identified through usability evaluation studies. In conclusion, the development of a usability evaluation framework considering the specific context of use of DLs is another objective. With regard to thorough and extensive results, the chosen approach is based on a combination of analytical and empirical methods. Furthermore, the organisation of patterns within the collection must be explored to make the collection a usable instrument for developers of DLs.

3 Relevant Research

The idea of applying structured patterns was developed at the end of the 1970s by Christopher Alexander. He defined patterns as three-part rule expressing a relation between a certain context, a problem, and a solution [14]. As Alexander introduced patterns as tools for engineers and architects designing buildings, towns, and other urban structures, in recent years his ideas have been transferred to design pattern in software engineering [15] and human computer interaction (HCI) [16]. The pattern language by Jenifer Tidwell is a collection of generic user interface design patterns that covers several kinds of applications, including Web or mobile applications. Examples are Global Navigation or Input Hints patterns [17, 18]. Other catalogues aim to guide design towards usable applications in the domain of e-commerce [19, 20] or interactive TV [21, 22]. Van Dyne, Landay, and Hong provide numerous patterns for a variety of Web site genres [23]. The pattern catalogue proposed by the Minerva Group for the design of usable cultural Web applications is inspired by the work of Welie [24] and contains patterns in the categories

• Make Contents Clear (e.g. Home Page or Site Map patterns),
• Present the Contents (e.g. Page Layout pattern),
• Navigating the Site (e.g. Clear Reference Points or Contextual Navigation patterns),
• Doing a Search (e.g. Search Page or Advanced Search patterns) and
• Interact with the Users (e.g. Form, Login or Newsletter patterns) [25].

Design patterns are proven solutions to recurring design problems in a specific context, and have been successfully applied in several domains [15, 16, 18, 19, 22, 24]. Well formulated, they carry a significant amount of reusable design knowledge, provide a thorough understanding of context of use and encapsulate concrete examples for applying to different kinds of applications.

To better select and understand patterns, they need to be organised. A categorisation gives structure to pattern collection. Welie [24] organizes his Web pattern languages according to both the design process and user interface structuring elements (such as Page Types, Feedback or Basic Interaction). Mahemoff proposes the following organizing principles:

• tasks,
• entire systems,
• user-interface elements, and
• users [26].

Another approach in pattern research is the concept of model-based design patterns. Central to model-based design development is the use of models at different levels to describe the relevant aspects of user interface design and the mapping between these models (e.g. user, task, business object, dialogue and presentation model [27]). The aim is to render user interfaces automatically from model descriptions. Sinnig et al. explored the roles of patterns in model-based design, suggested patterns using UML and Concurrent Task Tree notations and drew a pattern classification [26, [29]. But because of lack of tools and technology support for mapping and rendering the model descriptions automatically, developers and industry never adopted the model-based approach. Javahery et al. [28] investigate a pattern-based design process. The process is based on capturing user information and requirements (persona) as well as context of use information in form of context variables. These values are used to select design patterns and to leverage proven design solutions directly into conceptual models or prototypes of user interfaces.

There is no commonly agreed-upon format for patterns, different formats contain different elements. Most formats include at least the following four parts [26]:

• Name: A meaningful or catchy name to identify the patterns and that relates to the problem or the solution.
• Context: The situation(s) where the pattern is relevant for determining when this pattern should be applied.
• Forces: The forces present, which may constrain or suggest alternative solutions. When these forces are in tension with one another, the problem is harder to solve and a compromise may be necessary.
• Solution: A solution which resolves, as far as possible, the various forces.

Further elements are suggested to make patterns more usable for developers. In particular, patterns should focus on solutions that improve the usability of the system. Therefore, examples and a description of known uses are applicable elements. To easily access the knowledge concerning a particular problem, each pattern can reference other patterns that solve sub-problems related to the same problem or to the solution. Table 1 shows a very simple pattern that illustrates what a user interface design pattern may look like [24].

Table 1. Example pattern from Welie’s Interaction Design Pattern library [24].

There are many digital libraries, making design patterns highly relevant to reinvest design knowledge from a project to another. As far as I know there are no specific patterns for the design of DLs’ user interfaces. But one might say that all these existing Web design patterns from e.g. Welie or Tidwell are applicable to user interface design of digital libraries. Though these pattern languages provide design knowledge for Web design, they do not consider the specific use context of DLs. They are not written bearing in mind the special requirements and tasks of DL users, do not present specific design solutions found in digital libraries, and are not close enough to the DL domain. The above example states that “search functionality is not the main way to access the information, it has a secondary role”. This may be the case with a company Web site but the statement is not true in the context of digital library usage where search is the core function of the system. Design of DL Web sites is hardly comparable with the design of company or e-commerce Web sites.

4 Chosen approach

Figure 1 shows the proposed approach to identify and describe the user interface patterns for DLs. The first step is to analyse and describe a generic design space for DLs as well as to outline the specific context of use. The objective is to show how digital libraries differ from other kinds of Web sites, like, for example, commerce Websites, and to identify interface issues in general that are typical for DLs. The results should guide the identification and organisation of DL patterns. A careful literature review will identify typical user roles as well as their goals and tasks. Furthermore, we will analyse existing digital library applications collecting typical DL spaces, pages, functions and user interface elements (preliminary finding are stated later in this paper).

Work within steps two and three focuses on comprehensive pattern identification according to the previously proposed design space. Wesson and Cowley [36] recommend two ways to discover new patterns. Identification, study, and generalisation of existing examples of good practice are the most common way of pattern finding. If it is difficult to find sufficient (or any) examples of good practice, patterns may be identified through prototyping, evaluation, and redesign. I tend to go both ways: I will conduct a broad usability study (more than 50 Dls), and I will design prototypes in case of not finding usable solutions. My usability evaluation framework combines analytical and empirical methods like heuristic evaluation and usability tests. First, several digital libraries will be analysed and evaluated by experts using heuristic evaluation. Web guidelines like ISO 9241-151 [9], Web Content Accessibility Guidelines (WCAG) [10] and guidelines by U.S. Department of Health and Human Services (HHS) [11], as well as context specific guidelines like those presented in [13], serve as heuristics. The aim is to identify user interfaces observing fundamental usability guidelines. Experts in usability and user interface design will inspect the compliance with the heuristics while running through specified use cases. For example, analysing an advanced search of a digital library system, the expert can ascertain if the search form follows general design guidelines (e.g. “Place (automatically) a blinking cursor at the beginning of the first data entry field when a data entry form is displayed on a page.” [11]).

The solutions with less or no breach of heuristics will be evaluated further in usability tests. Usability tests with typical users focus predominantly on identification of usable DL functions. Users will run through given tasks within specified use cases. In case of an advanced search the effectiveness and efficiency of the search dialog is evaluated by observing the users performing their tasks. If they can reach their goals in an acceptable time, without help to go through the search dialog and with a high extent of satisfaction, the function will be classified as highly usable. This overall process seems to be efficient because it is a user- and expert-oriented evaluation approach.

Finally the usability evaluation results are translated into patterns. The pattern format will be determined according to the needs and requirements of DLs’ developers like librarians, Web developer or user interface designer. The last step is to organize patterns in a way that supports the work of developers.

5 Preliminary findings

5.1 Design Space of DLs

Spanning the design space of digital libraries is bringing together information about users, their tasks and the digital library system. The DL design space is intended to give an explanation of a general context of use and to give guidance in development and design of specific DL systems. The dimensions of the design space should be generic in the sense that it should be possible to consider them during a specific development process. Furthermore, the dimensions represent the patterns’ content in general. Digital libraries, of course, will differ from one another: they vary in content and services offered to the users, in language and cultural aspects, in technology, and of course the user community will differ in various characteristics. But nevertheless, there are generic needs and requirements common for most DLs’ users and there are user interface fragments recurring over and over again (e.g. search forms). I identified design dimensions at six levels of granularity: users, tasks, spaces, pages, functions, and basic user interface elements.

Users
Knowing the user is the fundamental requirement for user-centred design [30]. According to the MINERVA Handbook for quality in cultural Websites a user is “a professional person or not, a specialist or not, who casually or with specific aims, occasionally or systematically uses the Cultural Web Application. User identity is extremely variable, depending on cultural profile, aspirations for cultural growth, professional aims, and even momentary curiosity” [25]. Developers need to acknowledge individual differences in library system and domain knowledge. Users may be novices or experts. They are frequent or occasional users with different access rights to data. Characteristics considered to be typical for all users of digital libraries and assumed to have impact on design decisions are presented in Table 2. The objective is to model user patterns in a quite general and application-independent way.

Table 2. Types of users and typical usage characteristics.

Information tasks
The information task dimension determines what the user typically wants to do. Tasks that users will perform play a vital role in user-centred design. They give developers insight into the functionality that should be provided and how it will be used. Each task can be broken down further into subtasks [32]. For example Find something can occur as Search or Browse. Although a ‘typical’ digital library cannot be said to exist, there are normally a number of common tasks a user wants to perform like Searching or Browsing, Save items or Send references by email. Developers can use task patterns to identify and discuss services and functions that a system should support. Therefore it is also important to show the inter-relationship among task patterns and how tasks fit to function patterns. A “Save items” pattern, for example, might propose that the DL system offers functions like Download, Export Metadata, Purchase item, and so on.

Spaces
Spaces are a structural approach to conceptual design of digital libraries. Examples are the broad information space of DLs accessible for public or the individual workspace available only for registered users (myLibrary). There are also referenced spaces provided by other organisations, where the user can save result lists or references (e.g. a users’ account within social bookmarking services like http://delicious.com/).

Pages types
The level of pages constitutes all pages typically occurring in DLs (e.g. “Home page”, “Browse”, “Result list”, or “About us”).

Functions and user interface elements
Functions are defined as an aid or ‘tool’ the user can use in order to perform a task (e.g. search within search results or citation alerts), and are mostly composed of user interface elements as forms, buttons or hyperlinks. In the following, a pyramidal structure is chosen to organize the design space (and upcoming patterns) – from the details to the general (see figure 2).

5.2 Tasks and functions

In spring 2009 a study was performed to identify typical library functions and to analyse how they support what users want to do (goals and tasks) [31]. A task is what people need to do to achieve their goals. The goals are what users expect, want and need from the system and what they would like to do with the system [32]. Thus, we conducted two studies: in order to identify typical goals and tasks several studies have been evaluated in a literature review. This analytical kind of approach is considered to be sufficient, because a number of findings have been mentioned over the last years and the results are adequately published. Furthermore, 25 existing digital library applications have been analysed collecting typical DL functions. As result a number of typical DL functions has been identified and matched to user goals and tasks (see Appendix 1).

Goals and tasks of DL users
Within the last decade there have been several studies concerning with the description of users’ information needs, requirements, and habits and their expectations on how to perform their tasks in digital libraries. Paepcke [33] suggests a broader image of what users should be able to do with digital libraries, "not just search, but four other aspects of user tasks." Within a series of semi-structured interviews he identified main activities that can be partitioned into five categories:

1. locating and selecting among relevant sources,
2. retrieving information from them,
3. interpreting what was retrieved,
4. managing the filtered-out information locally, and
5. sharing results with others.

Another approach is presented by DELOS network gathering data in several case studies, a literature study and questionnaire survey [34]. They are presenting a Digital Library Lifecycle containing the activities of administrators and users. The user activities are:

1. Discover: Search and find the appropriate digital library for the topics of interest
2. Access: Gain access to the digital library registering personal information and preferences
3. Use: Search or browse the available material in the digital library concerning personal interests and take advantage of the appropriate resources
4. Review/Comment: Provide feedback to the other users and administrators of the digital library regarding the value of the available resources
5. Personalise: Personalise the use of the digital library creating an appropriate plan of use that serves user’s purposes

Goals (activities) of digital library users are presented by Gonçalves et al. [35] within "Taxonomy of digital libraries terms" This formal model is mixing activities of administrators and users. The users’ high-level goals are

1. Abstracting,
2. Creating,
3. Disseminating,
4. Organizing,
5. Personalizing,
6. Requesting/Selecting.

A comparison of these three approaches shows that there are some correspondences between the meanings of the different goals although the wording is not the same (Table 3).

Table 3. Comparison of users’ goals and tasks in literature review

The goals described in Paepcke are supplemented with the goal “Personalize” in the DELOS approach. However, the description of using information resources in DELOS (“Use”) is insufficient. The goal “Creating” is only considered by Gonçalves but is becoming more and more prominent in times of Web2.0 or Library2.0. Finally, bringing together these different views the following goals and tasks of DL users are defined as follows:

• Information Retrieval: Search or browse the available material in the digital library concerning personal interests.
• Interpretation: Cogitate, reflect, examine and organise the results regarding the quantity and quality (relevance) to decide if the results are suitable for analysis in detail or if a new search has to be run.
• Extraction: Export, save, bookmark, print or purchase the results to get permanent access to the information outside the digital library and recommend results to other people that might be interested in the same topic.
• Personalisation: Tailor information, functions and DLs’ User Interface to match the unique and specific needs of an individual or a community.
• Collaboration: Work together in networks to provide feedback to the other users and administrators of the digital library regarding the value of the available resources

In the following the results from Web-analysis of 25 digital libraries are presented regarding the question how the goals and tasks are ‘implemented’ in DL functions.

Typical DL functions
The following digital libraries have been explored (Table 4). The selected DLs are well-established systems, many of them existing for more than 5 years. Some of them are publicly funded (e.g. Europeana), others are in the private sector or services offered by publishers (e.g. Science Direct). The collections offer a great many of records between 10.000 and 150 million items and consist mostly of text documents. But there are some DLs like Digital Mechanism and Gear Library (DMG-Lib) or Internet Archive offering also pictures, audio or video content. Most of the DLs were accessed via Athens Login provided by Ilmenau University of Technology. If personal registration was offered, this possibility was used to gain access to almost all functions. Appendix 1 shows the results of the Web analysis. The identified functions are assigned to the users’ goals and tasks mentioned before.

Table 4. DLs explored in Web-analysis

5.3 Organizing patterns for digital libraries

So far, as mentioned above, patterns are going to be classified along the hierarchical structure of the design space (see Figure 2), following a quite traditional way of designing applications: from the general to the specific. Moreover, patterns should refer to other patterns because they usually do not describe isolated solutions. For example, if a task pattern contains search functionalities, references should be available to other patterns in which search functions appear as well. The search function is in turn related to User Interface Elements Patterns, to give just one example, because search patterns usually contain forms or buttons. Therefore, I propose additionally a cross referenced organisation of patterns just considering that some patterns can ‘relate to’ other patterns and that more detailed information can be found in ‘refined’ patterns (see Figure 3).

Figure 3. Cross referenced organisation of patterns for DLs

A process oriented approach of organisation is shown is figure 4: patterns are organised the way the phases of development process succeed. The process is based on the Usability Engineering Lifecycle [8].

Figure 4. User centred design process oriented patterns organisation for DLs

Keeping in mind that patterns are generic solutions to recurring problems [14], they can support the requirements analysis by showing problem areas of different kinds that need to be treated. But nevertheless, it is important to do specific requirements analysis because a “generic” user seldom exists. While designers work on conceptual and detailed user interface design, the different user interface design patterns can guide them through prototyping. These patterns describe the user interface without ambiguity though avoiding technical detail, and intuitively enough so that any design stakeholder can read and at least slightly modify the description. Finally, patterns aim at supporting designers in the evaluation of design alternatives, be it formative or summative using analytical or empirical evaluation methods.

6 Conclusion and Future research

Patterns can be a useful reference point for those developers not being experts in user interface design or usability, but they neither eliminate nor substitute for involving users in the development process. Users can be integrated at several stages of a digital library design project:

• to support requirements analysis and specification,
• to organise the information,
• to study graphical aspects and even
• to evaluate the usability of the system.

My future work focuses on the comprehensive pattern identification according to the design space proposed above. Therefore the next step is to prepare for usability evaluation studies combining analytical and empirical methods like heuristic evaluation and usability testing.

Acknowledgments

I thank my supervisors Prof. Dr. Heidi Krömker (Ilmenau University of Technology) for her support. The revision of this paper was supported by Prof. Dr. Dagobert Soergel (College of Information Studies, University of Maryland).

References

Appendix 1