Accessing the World Wide Web By Telephone

Transcription

1 1 of 10 17/01/ :40 Accessing the World Wide Web By Telephone M. Zajicek, C. Powell School of Computing and Mathematical Sciences Oxford Brookes University, Gipsy Lane, Oxford, OX3 OBP, UK Tel: Fax: Abstract The work described in this paper was funded under the HEFCE initiative for widening access to Higher Education for disabled people, described in Zajicek & Greenhead, (1996). The problem of visually impaired people using a computer with a graphical user interface is being addressed. Current provision in terms of Braille and speech screen readers tends to be unsatisfactory, Laux et al, (1996), Petrie et al, (1993), and Stephanidis et al, (1995), and does not present the information conceptually. We have built a helper application WebChat which builds a conceptual model of World Wide Web pages and describes them using speech for visually impaired people and can be used over the telephone. This paper examines the viability of computer interaction over the telephone in the context of searching for information on the World Wide Web. Key words: Telephone; speech output; World Wide Web. 1. Introduction The World Wide Web could well emerge as a universal information source providing information normally found in books, or by attending courses, conferences etc. Although this is of great benefit to sighted people, visually impaired people become further disadvantaged as instantaneous acquisition of vast amounts of information becomes the norm. Furthermore those that do not have access to a networked computer are excluded. The requirement for up to date information and that large volumes of data should be immediately available could escalate. The cognitive demands of graphical user interfaces are considerable and require total attention. People are restricted to working where the computer is. The authors can envisage a future in which the requirements of continual update of information increases the need for individuals to update themselves at times when they are not in front of a computer. We can imagine a situation in which it might be more useful to access information on the Web via the telephone. The user might want to find something out while doing the ironing of some other physical activity without being tied down to sitting in front of a computer and giving it their undivided attention.

2 2 of 10 17/01/ :40 The work described below has developed from a project originally aimed at enabling blind users to access the information available on the World Wide Web. This was funded by the Higher Education Funding Council Executive for Widening Access to Higher Education for Disabled Students. An application called WebChat has been built which intercepts the HTML code that forms a Web page and provides a conceptually organised version of the Web page using speech. This paper describes the way in which the Web page has been conceptualised and presented using speech in a menubar format. It also discusses the results of an evaluation of the system in the context of swift orientation to Web pages to determine their potential usefulness in searching for information. 2. Conceptualisation of the Web Page 2.1 Providing easy access to parts of the page We have re-described the structure of a Web page so that it can be read out to a visually impaired user over the telephone. This involved forming a conceptual model of the information content of the page and its interrelationships and presenting it in speech form. The Web page has been presented conceptually as described in Zajicek & Powell, (1997) as a collection of textual objects (consisting of heading and accompanying paragraph) and links. WebChat provides access to conceptually different parts of the Web page using function keys and different synthesised voices to indicate conceptually different features of text such as headings, links and paragraphs of text. Table 1. Control Keys Function Key Normal Function Shift Function F1 Read document title Load a URL F2 Read all links Read links under selected heading F3 Read Headings Read headings under selected heading

3 3 of 10 17/01/ :40 F4 Read entire document Read document under selected heading F5 Read entire dictionary Read dictionary of words under selected heading F6 Select link menu F7 Select heading menu Search for a user-specified word F8 Select bookmark menu Search for interesting words The conceptually different parts or objects are ordered hierarchically within heading and sequentially down the page. The HTML code which forms the Web page allows for several headings. This feature is used by WebChat to subset the document and give the user the ability to examine only the elements of text of interest. Navigation of the Web page was achieved by easily locatable keys as shown in Tables 1. above. The Menu keys in Table 2. provide a keyboard-operated menu system see Figure 1. For example, pressing F6 selects the links menu. The first link is read out. Pressing the down cursor key causes the second link to be read out. Once the required link is located using the cursor and page keys, pressing Return will load the associated document (equivalent to clicking on an anchor in a standard web browser). Pressing escape will cancel the menu and return WebChat to its ready state. The Heading and Bookmark menus are used in the same way. Table 2. Menu Keys Key Esc Spacebar Page up Page down Home Function Cancel any menu and return to ready state, Stop any speech Pause/resume speech Go to 10th link/heading/bookmark before current position Go to 10th link/heading/bookmark after current position Go to first link/heading/bookmark

4 4 of 10 17/01/ :40 End Cursor up/right Go to last link/heading/bookmark Go to previous link/heading/bookmark Cursor down/left Go to next link/heading/bookmark Return Select the current menu item A graphical interface is essentially concurrent. Many items of information and options are displayed so that the user can engage with any one of them at any time. Importantly the user need only engage in those parts that interest her and ignore the rest. Speech born information is essentially sequential. We have enabled a similar concurrency to that in graphical interfaces by conceptually structuring the page to offer options such as lists of links, or lists of headings, lists of keywords found in the text, or to read out paragraphs of text etc. This is achieved by using a menubar which is accessed by speech or function key. In this way the user is able to move around the document from component to component without being restricted by its sequential nature. This enables them to decide whether the page will be useful and/or read only particular items of information. This approach is particularly useful for users wishing to move around the World Wide Web speedily, orienting themselves to pages and deciding if they will be useful in their information search. Figure 1. Selecting from a menu. 2.2 Using headings and title for orientation

5 5 of 10 17/01/ :40 The title element in HTML gives a general context for the document which contains it, and gives the user a general idea of what the document is about. One would expect this to be an important content-indicator. However, the TITLE tag is not mandatory in HTML and many documents do not contain this information. Other documents may have less-than descriptive titles, for example What s New What s Cool can be ambiguous. HTML permits a hierarchy of headings which describe the structure of the document. Top headings give general descriptions of the sections within the document, while lower headings giving more specific descriptions of subsections within a section. This is sometimes problematic since the author of the HTML document is not obliged to use the heading tags to indicate a heading; the tag is sometimes used for its visual effect only, that is, to lay out the text in a way the author finds pleasing. 2.3 Using links for orientation Anchor tags are used to provide the hypertext links to other Web pages. Unlike the tags above, these really do have only one purpose and so the meaning of their contents can be relied upon. However, usefulness of a link as an aid to determining the content of the document is difficult to determine, as anchors fall into three general classes: 1. Strongly related links, 2. Weakly related links, 3. Links back to the site home page/previous page/top of page. Only the first of these will be of any real use in determining content. For example, a site about Mickey Mouse may have a link to 'My favourite Mickey Mouse site'. The second being only weakly related may not directly name the relevant content words, for example 'My favourite Animated Characters. The third is not at all related to the users line of enquiry, for example 'Back to top of document'. 2.2 Using keywords in the page Keywords were extracted from Web pages using standard information retrieval techniques as proposed by Luhn in Shultz (1968). These are usually applied to library searches for significant words in text and are based on the premise that the frequency of word occurrence in an article furnishes a useful measurement of word significance. Common words with high frequency are referred to as noise words. These are words such as a and the etc. and are not included in the keyword list. Similarly those which have such a low frequency that they are unlikely to represent the content of the page are discarded. Table 3. below gives an example of keyword extractions.

6 6 of 10 17/01/ :40 Table 3. Examples of keyword extractions. Document title Total no. of words Keywords Document subject Thinking of getting a border collie? 1174 border collie people collies dog exercise dogs Information for prospective border collie owners Comet Hale-Bopp Home Page (JPL) Dennis' Information Retrieval Page 772 comet hale image January images vodniza 279 information indexing ir pages index search engines Information about comet Hale-Bopp General information about information retrieval and links to related sites Table 4. below shows an example of the three types of representation of the Web page using headings, anchors, and keywords. 3. Evaluation of the WebChat Menubar The evaluation was in two parts. The first was to evaluate the menubar approach as a tool for orientation to a Web page and for information retrieval, and the second was to determine which option on the menubar was perceived by the users to be most effective. Table 4. Three different representations of Web page content Document Title: Comet Hale-Bopp Home Page (JPL) Headings Anchors Keywords

7 7 of 10 17/01/ :40 - Comet Hale Bopp home page -Comet Hyakutake -comet - The great comet of 1997? -Archive for comet -hale -Countdown 65 days to perihelion Hale-Bopp images -image - News Flash -Information on Comet -January -Recent Comet Hale-Bopp images Hale-Bopp for -images non-astronomers -vodniza -Central Bureau for Astronomical Telegrams -Comet Hale-Bopp Update -New Hale-Bopp Companion images proven to be fakes -Comet Hale-Bopp and the Alleged Anomalous Object 3.1 Evaluation of the menubar facility Inevitably the time taken by the user to orientate themselves to the information content of a Web site was significantly more than the time taken using visual scanning. Therefore evaluation of effectiveness was based purely on information gathered and perceived orientation rather than time taken. Two aspects of Web page presentation were evaluated using six user subjects for each aspect. (1) Web site orientation - Users were given a goal to accomplish in terms of finding information. They were given a set of Web pages to investigate for possible usefulness in achieving their goal. They were questioned not on what the page contained but rather on how clear it was to them that the page was useful or not. Subjects were asked general questions designed to determine the extent of their conceptual model of the information available on the pages. It was found that they had built a conceptual model of the information content of the page comparable to that of a sighted user after scanning. (2) Information retrieval - Users were required to use the site as an information source and were given the task of retrieving specific information.

8 8 of 10 17/01/ :40 It was found that the design of the Web site and length of sentences were major factors in the level of success in orientation and information retrieval. This was mainly due to the uninteresting nature of synthesised speech. 3.2 Evaluation of options on the menubar The effectiveness of the browsing facility provided by the menubar was evaluated in terms of recall, and precision. It is assumed that the more effective the system the more it will satisfy the user. It is also assumed that precision and recall are sufficient for the measurement of effectiveness as described by van Rijsbergen (1979) The user s were asked to rate the effectiveness of headings, links, and keywords as indicators of the contents of a page. Table 5. below shows that headings were perceived as being the most useful with keywords coming a close second and links last. Table 5 Scores for different representations Total Score Percentage of available marks Headings Anchors Keywords It is recognised that the evaluation methodologies outlined above use a somewhat contrived experimental environment. A more valuable evaluation program is planned which will use in-built machine monitoring to see how users actually work with WebChat over a time period in the field. 4. Implications for Telecommunications The main challenge in using speech for information retrieval or information presentation is its inherent sequential nature. Users can become impatient as they are forced to listen to output that they are not interested in. The aim of this work is to reduce this major obstacle to the use of speech output by providing a menu bar for jumping around the web page. The danger with this form of partition is that users can loose their way within the information elements. Users will build a conceptual model of the information to guide them. The results of evaluation show that users are able to retain a robust conceptual model of the page and to dip in to the information they require as and when they need it.

9 9 of 10 17/01/ :40 If we can find a way of providing concurrency of information which is transparent to the user and embodies the power of the concurrency inherent in graphical user interfaces then access to computers by telephone will open up a new information flow channel. People are being bombarded with visual images and although these are very powerful the audio channel might provide more lasting retention. There is evidence that people retain audio information such as songs etc more readily than visual information. 5. Conclusion These initial experiments have shown that there is potential for the presentation of Web pages in audio form. Although orientation and information retrieval took longer the models built up of the information available seemed comparable with that those achieved by visual scanning. It is intended that system monitoring will be used to see how these different representation are used over a time period, and in the field. Under these conditions users would have the opportunity to become more familiar with the different representations, have time to develop strategies, and have real goals in using the Web. Working under more real conditions might yield different results. Different algorithms for selecting keywords will also be investigated. In addition the problem generated by the increasing use of multimedia objects on Web pages is being addressed within the project by experimentation with the use of different notation techniques to represent the real information embedded within them. References Laux, L., McNally, P., Paciello, M., Vanderheim, G., 1996, Designing the World Wide Web for People with Disabilities: A User centred Design Approach, Proceedings of ASSETS 96, The Second Annual ACM Conference on Assistive Technologies, ACM, New York Petrie, H., Fabrizi, P., Homatas, G., 1993, Report on user requirements for the accessibility of graphical user interfaces by blind people, Report to the CEC for TIDE Pilot Action Project 103: Graphical User Interfaces for Blind Persons, London, Royal National Institute for the Blind. Schultz, C. K.; 1968; H. P. Luhn: Pioneer of Information Science - Selected Works; Macmillan; London; Pg Stephanidis, C., Savidis, A., Acoumianakis, D., 1995, Tools for User Interfaces For All,

10 10 of 10 17/01/ :40 Proceedings Second TIDE Congress, Amsterdam, IOS Press van Rijsbergen C. J.; 1979; Information Retrieval; Butterworths; London; 2, 17. Zajicek M., Powell C.; 1997; Building a conceptual model of the World Wide Web for visually impaired users; Proc. Ergonomics Society 1997 Annual Conference; Grantham. Back to publication list.