Automatic Creation of User Profiles for Achieving. Personal Web Accessibility. Markel Vigo, Amaia Aizpurua, Myriam Arrue, and Julio Abascal. University of the ...
Automatic Creation of User Profiles for Achieving Personal Web Accessibility Markel Vigo, Amaia Aizpurua, Myriam Arrue, and Julio Abascal University of the Basque Country, Informatika Fakultatea, Manuel Lardizabal 1, 20018 Donostia, Spain {markel,scpaiaga,myriam,julio}@si.ehu.es
Abstract. Automatic capture of the user’s interaction environment for useradapted interaction and evaluation purposes is an unexplored area in the Web Accessibility research field. This paper presents an application that collects user data regarding assistive technologies (be either software or hardware) in an unobtrusive way. As a result, CC/PP based profiles are created so that interoperability between components such as evaluation engines or server-side content adaptors can be attained. The implications that versioning issues and the potential user group of a given assistive technology have on the guidelines to apply are also remarked. The major benefit of this approach is that users can perform their tasks avoiding distractions while interacting with the World Wide Web.
1 Introduction Adaptive user interfaces, intelligent help systems and explanation systems require maintaining user models to represent the characteristics of the user [1]. To this end, collecting information about the user and their environment is a key practice in user modelling related scenarios. In the first stages of this task the main objective is to collect data about the user and their behaviour so that preliminary assumptions can be made. User characteristics (language, abilities, etc.), preferences (settings in an application), behaviour (navigational styles), goals, a priori knowledge on a subject and likings are gathered to create a first user profile or assign a stereotype. Knowledge about the user can be obtained explicitly by asking questions or filling in forms while implicitly obtained knowledge, retrieved by monitoring user interaction, can be used to polish the first user profile or assign a profile that fits better with user characteristics. Accessibility is about accommodating people with disabilities. Therefore, the research field of universal access to the World Wide Web (WWW) can take advantage of user-modelling practices in order to provide user-adapted interaction mechanisms for people with disabilities. In this sense, Stephanidis [2] presents several developments that prove the feasibility of this statement. In this paper, a software component that collects data about the user and their access environment is presented. According to Sloan et al. [3], user characteristics and technological requirements are part of the “context of use” which has to be captured in order to obtain acceptable accessibility levels. Data gathering thus focuses on system features as well as assistive technologies. The later are necessary as they give clues about the user. For instance, if a determined screen reader is detected it can be inferred that the user is blind or visually impaired. K. Miesenberger et al. (Eds.): ICCHP 2008, LNCS 5105, pp. 354–361, 2008. © Springer-Verlag Berlin Heidelberg 2008
Automatic Creation of User Profiles for Achieving Personal Web Accessibility
355
The fact that data are collected unobtrusively in order to not to cause the distraction of the user is one of the strong points of this contribution. Furthermore, after inferring some knowledge, hardware, software and user characteristics are automatically encapsulated in a profile that follows W3C CC/PP recommendation [4]. These profiles are the basis to achieve Personal Web Accessibility and can be useful in the scenarios described in Section 1.2. 1.1 Personal Web Accessibility A website meeting accessibility guidelines such as Web Content Accessibility Guidelines 1.0 [5] does not ensure it is accessible [6] whilst in some cases it is not necessary to satisfy these guidelines for certain user groups to obtain an acceptable accessibility level. Amongst others, there are some reasons for such contradiction: guidelines are vaguely specified and therefore their application is ambiguous, some guidelines contain unresolved references to the user’s delivery context, or guidelines are incomplete. In order to overcome this situation, Personal Web Accessibility provides user-tailored accessibility resources by adapting web accessibility guidelines to the requirements of the user and their delivery context. A preliminary sketch of a system that tackled personal web accessibility challenges was presented by Arrue et al [7]. Later, this system evolved into a more comprehensive framework which was extended in order to handle specific issues arisen by assistive technologies and access devices such as mobile devices [8]. Theoretical considerations were specified and the server-side evaluation logic was implemented. The framework is capable of managing different guideline sets and WCAG 1.0 and Mobile Web Best Practices, MWBP 1.0 [9] were implemented in the aforementioned work. Due to the release of the mobileOK Basic tests [10], which are the technical recommendations to partially implement the MWBP 1.0, the system has been enhanced especially in the case of mobile devices’ access to the WWW [11]. Aiming at a holistic approach of the framework, this paper presents the client-side components that capture the user environment constraints. 1.2 Application Scenarios The software component presented in this paper has two major purposes in the context of Personal Web Accessibility: • User-tailored accessibility evaluations: adaptive navigation support [12] can be provided by means of local and global guidance indicating the most accessible links according to a profile. In addition, information retrieval systems can consider web accessibility as a ranking criterion as demonstrated by Arrue et al. [13]. Moreover, the usage of profiles makes possible the development of websites for specific user groups. Once profiles are created they are stored in a remote database so that developers can make use of them for evaluation purposes. For such scenario, evaluation tools should interoperate with profiles obtaining profile-tailored evaluation reports. It would be the foundations to achieve universal access from an alternative path. This may sound contradictory with the established design-for-all paradigm. Yet, other authors support this idea coined as the design-for-one paradigm [14].
356
M. Vigo et al.
• Adapting web content according to a profile: server-side services are capable of adapting web content to device specifications and user preferences. Generally, the data exchange format is CC/PP [15]. 1.3 Related Work Since web content adaptation for people with disabilities is itself a discipline within web accessibility, this section just provides a view on the approaches that have been followed in the universal access field as far as collecting facts and assumptions about the user is concerned. Preliminary attempts to merge the knowledge and techniques between usermodelling and universal access field were carried out in the AVANTI project. The AVANTI web browser is capable of tailoring web content to several user groups: mobility impaired people, visually impaired users etc. Assumptions about the user are made based on questionnaires or are retrieved from smart cards [16]. Later, Velasco et al. [17] state how people with disabilities can benefit from user-modelling techniques and present a prototype which seems that obtains data about the user interacting with forms. Hanson et al. [18] extend the Mozilla Web browser with several functionalities that allow the user to adapt web content to their needs and it is stated that the user can set their preferences using a logon screen that interfaces a remote DB. The remaining of the paper is organized as follows: Section 2 describes the requirements to build the application, its architecture, the output obtained following the CC/PP grammar as well as the implications that assistive technologies have in providing user-tailored accessibility resources. Finally in Section 3, conclusions are drawn and future work is devised.
2 Context of Use Detector The main objective of the Context of Use Detector (CUD) is to collect data about the access environment or delivery context, be assistive technologies features or accessing computer characteristics. As the application is a component within a more general framework [8] the following requirements are met: • Data are unobtrusively collected in order to avoid user’s distraction. • Information about hardware and software features of the computer as well as assistive technologies is gathered. • The application should be multiplatform and currently runs in Windows and Unixbased environments. Multiplatform issue is solved using Java programming language. However, accessing to the hardware, software and assistive technologies’ features requires different strategies in each platform. • The output of the application should be formatted following standards so that interoperability between tools can be achieved. This component should produce a CC/PP-based profile that interoperates with other components such as evaluation tools, user-agents or server-side content adaptation services. 2.1 Architecture of the Context of Use Detector As mentioned in the previous section one of the objectives is to obtain interoperability among different components. All components in the Context of Use Detector (CUD)
Automatic Creation of User Profiles for Achieving Personal Web Accessibility
357
can be considered as autonomous applications and incorporation of new developments is planned in order to enhance current capabilities. Numerous components are identified in the CUD: • A remote database where all information regarding assistive technologies is stored. Data are manually maintained by the administrator. For such task, a web interface is provided to the person in charge so that updating, editing or removing entries becomes straightforward. The administrator creates a record for each assistive technology introducing data such as name, vendor, version, description, modality (whether is input, output or both), etc. • A database that mirrors the remote database is located at the user’s computer. • The Assistive Technologies Detector compares key values in the database with their values in the registry. When a match is found data from both sources (registry and database) are included in the profile. • The Access Device Features Detector retrieves information concerning hardware and software characteristics of the computer that is being used. • Automatically created profiles are remotely stored in the Profiles Repository. The abovementioned components, the information flow among them and how profiles can be used by different roles is illustrated in Figure 1. The process of the automatic creation of user profiles is described as follows: 1. When the system is initialized in the user’s computer (be a laptop, PDA or workstation) it queries the remote database in order to ascertain whether updates have been made. If keeping the coherence between the databases is required, new data are downloaded to the local database. 2. The Assistive Technologies Detector queries the local database to check whether a particular assistive technology is installed (it includes both hardware such as Braille displays, and software such as screen readers, screen magnifiers etc.). The objective of the queries is to compare database entries regarding available assistive technologies with key values in the System Registry. When a match is found, information regarding the assistive technology is obtained from both information sources (database and Registry) as the Registry tends to store complementary information. 3. Information regarding access platform, operative system, CPU, RAM memory size, keyboard type, screen resolution etc. of the computer is retrieved using a specific system API. 4. Data obtained in step 3 and step 4 are put in a CC/PP file which is subsequently uploaded to the profile server. 5. User-agents (mainly browsers) provide developers frameworks to extend their capabilities. For instance, the Mozilla Web browser uses XUL technology [http://www.mozilla.org/projects/xul/] for such task. User profiles could be used in the application scenarios described in Section 1.2 if specific plug-ins are developed. 6. Developers can take advantage of the stored profiles to create user-group-focused websites. Evaluation tools should be able of “understand” CC/PP profiles.
358
M. Vigo et al.
Fig. 1. A overview of the architecture: components, relationships as well as the roles that take part in the system
2.2 Output and Implications on the Evaluations CC/PP provides a framework to model hardware, software and user agent characteristics. However, attributes are not defined and third parties should extend new vocabularies considering that reuse is encouraged. In this case, a new vocabulary has been defined in order to express the detected characteristics. New words in this vocabulary can be identified as they are preceded by the access namespace. An example of the output for a Linux computer with several assistive technologies installed on it can be found at the Appendix. Some excerpts have been omitted due to lack of space. Since it is possible to infer the user’s disability from the detected assistive technologies, one of the main tasks in the Personal Web Accessibility framework is to retrieve from a heterogeneous guidelines repository those guidelines that impact on a specific user group. This is a partial step because guidelines tend to capture “typical” needs of their respective group members as individual needs may deviate considerably. Therefore, other considerations such as assistive technologies’ release version are of major importance. Different releases of assistive technologies have new features that make some accessibility barriers obsolete (positive dependencies) while users of older versions of these technologies may face accessibility problems even if guidelines are met (negative dependencies) [8]. Some examples for JAWS screen reader with regard to WCAG 1.0 are the following:
Automatic Creation of User Profiles for Achieving Personal Web Accessibility
359
• Positive dependencies: a subset of WCAG guidelines (6.3, 6.4, 8.1, 9.3) propose some good practices regarding scripting for attaining device-independence and lack of support reasons. For such reasons, OnClick attribute had to be carefully used. From the version 5 onwards it does not pose a problem anymore. • Negative dependencies: an acronym is identified and correctly coded following the recommendations in WCAG 4.2 guideline. Yet, users of older versions (
