Experiences of QWERTY development

Teemu TOROPAINEN and Pasi OJALA Oulu, Finland

ABSTRACT good, usable QWERTY layout can be very challenging containing several compromises with the number and size of the keys. During the last decades researchers and industry have published several studies on QWERTY based text input technologies. Many As QWERTY keyboard seems to be “the standard keyboard” also studies have been made about mini-QWERTY and other reduced for mobile devices, this paper concentrates only on the existing QWERTY layouts targeted for mobile devices, but usually only QWERTY based layouts for mobile devices. This has been seen English language is used in these studies as input language. Also, justified because the understanding of the challenges that are faced there does not seem to be a wide consensus of which kind of when selecting QWERTY layout is clearly needed for developing layouts are most recommended for mobile devices. Selecting the layout that supports most languages and fits small enough space right QWERTY layout is even more challenging if we take into with usable key size. For these reasons we have not seen it consideration the number of supported languages versus the necessary to focus on the QWERTY key shapes and materials or number and the size of the QWERTY keys. software enablers like predictive text input or similar either. We have also seen that on-screen QWERTY is out of scope in relation This paper reviews existing QWERTY layouts for mobile devices to our research problem. and presents constructive critique for using them. Even some critique is presented it is notified that developing QWERTY layouts to mobile devices is always some sort of a compromise. 2. QWERTY IN MOBILE DEVICES Keeping this in mind this paper proposes methods to make the QWERTY smaller and still usable with multiple different languages. The placements of the keys in mini-QWERTY layouts usually follow those used in traditional QWERTY keyboards, except some The practical industrial case shows that neither of the tested variation required by certain languages (AZERTY used in French QWERTY layouts was perfect. However, a layout containing and QWERTZ used in German for example). All the keys that are characteristics of both layouts is usable for mobile devices. The used in traditional computer QWERTY keyboards are not needed case also shows the practical challenges which are often faced by in mobile devices, so mini-QWERTY layouts are usually mini-QWERTY developers and finally proposes how these somewhat simpler. In practice, mini-QWERTY typically consists challenges could be solved. of alphanumeric, special , modifier, Space, and Enter keys. Keywords: software, QWERTY, mini-QWERTY, text , mobile device, usability The total amount of needed alphanumeric keys depends heavily on language. [1] If a language uses another script than Latin, e.g. Cyrillic or Arabic, both Latin and the main script are printed on the to enable text input in both scripts. If there is 1. INTRODUCTION more than just one language on the layout, it means that there is less space for printing characters on physical keyboards. QWERTY keyboard is well over one century old invention. Even if it is a rather old invention it can still be called as “the standard All languages have the same set of Latin number keys available: 0 of the keyboards.”[7] It has also achieved a standard role in ISO 1 2 3 4 5 6 7 8 9. Full size QWERTY has usually independent (ISO/IEC 9995, 1994). A miniature version of QWERTY (mini- number keys but in mini-QWERTY layouts the number keys are in QWERTY) [2] has been developed for mobile devices. many cases “integrated” to alphabetic keys and is Technically, mini-QWERTY must fit into a much smaller space needed for entering them. The number keys can be placed in row than the traditional QWERTY. In practice mini-QWERTY is on the top level of the QWERTY but based on our usability test usually used with two thumbs. findings presented in chapter 4 in some cases also grid layout can be a good idea since it might help input of phone numbers for Mobile devices set several limitations on the number and the size example. Especially if one-hand usage is important, grid layout of of the keys used in mini-QWERTY. In addition, many of these numbers can be recommended. keys have often multiple functions which normally cause usability problems. Especially regarding to multimedia device development The list of special characters is more or less the same for all these challenges have been seen significant. This is perhaps languages. In practice only the prioritization of the characters may because industrial design of multimedia devices starts in very early vary: phase when only the basic dimensions of the product are known. _ % * § # | , . ; : ? ! ¡ ¿ · ´ ` ~ ^ + - ’ ” ' " @ {} [] ( ) - = In addition, as the general user requirement for the mobile devices \ / & £ $ € ¥ ¤ _ <> “as small as possible”, limits the space for keyboard, designing a

Modifier keys (Shift and Alt Gr for example) are typically used in Table 1. QWERTY layout used in Hwang & Lee study. combination with some alphanumeric characters to input upper case characters, special characters and accented characters. Also EQW TFY OP dead keys are considered as modifier keys in this paper because they behave in the same manner. Typically, full size QWERTY has more modifier keys than mini-QWERTY and in many cases it ADZ RGV ILJ also has more operating system specific keys. For example some keys like Enter, Backspace and Shift are often made bigger in full SCX HUB NMK size QWERTY for achieving better usability.

Related work Clarkson et al. (2005) used separate mini-QWERTY devices that were attached to computer in their study. Therefore their results do

not match the real mobile device environment. However, they did Green et al. (2004) have built a specialized keyboard for text entry set some kind of performance benchmark for future studies that maps four rows of a standard keyboard onto the home row, regarding input speed of mini-QWERTY devices. with different characters encoded via modifier keys and multi-tap input. Use of this keyboard relies on lexicon-based disambiguation. This kind of design has two benefits: it limits physical space requirements and capitalizes on user knowledge of the standard QWERTY keyboard layout. The resulting “stick” keyboard is between 15% and 25% of the size of a standard keyboard. In practice in a preliminary empirical study, users have reached half of their normal typing speed using lexicon-based disambiguation (22.5 wpm) and a reasonable but lower speed with multi-tap input (10.4 wpm) with only a few minutes of practicing.

Figure 2. The keyboards used in Clarkson et al. (2005) study.

Rennie and Cockburn (2005) did a comparative analysis and empirical evaluation of three alternative techniques for helping Figure 1. The stick keyboard used in Green et al. users input non-standard alphabets using a standard keyboard. In study. particular they investigated whether their VKPLUS (Visual Keyboard Plus) user interface, which displays both the physical Green et. al (2004) point out that each software or hardware key labels and the new key bindings, improves text entry rates innovation in text input technology must address tradeoffs between over Microsoft's Visual Keyboard. The third technique, included several issues: input speed, accuracy, physical form factor, for baseline comparison, used sticky-labels placed over the learning time and cost. Text-entry input technologies trade off physical keyboard. these factors in different ways.

As result of their studies Green et. al (2004) came to the conclusion that stick keyboard will not replace devices that already require a numerical keypad, such as cell phones, or those that must be held in one hand for use.

Hwang & Lee (2005) compared 3x4 QWERTY layout to traditional ABC layout in their study that was started from the hypothesis that a user's skill in a QWERTY keyboard may be Figure 3. VKPLUS: Both QWERTY characters and the transferred to a 3x4 keypad environment. They carried out two symbols to be typed in are visible on each key. experiments to find supporting evidence. The results from the first experiment suggested that the spatial memory about the QWERTY The results of Rennie and Cockburn (2005) show that VKPLUS keyboard does help the user locate correct keys. An important significantly improves performance over Microsoft's system, but implication of this was that a keypad layout that follows that of the the stickered keyboard outperformed both systems. However, it is QWERTY keyboard may be easier to learn for computer users. not a practical solution in many cases, as the stickers would Their second experiment compared the efficiency of an ABC adversely affect normal typing, it would be hard to change layout and a QWERTY-like layout and the results were clearly in alphabets, and it is not very portable. favor of the latter one.

Clarkson et al. (2007) came to the conclusion that Analytic models small the keys can be from a usability point of view and a lot keys like the two-thumb model can be powerful HCI (Human Computer crammed into a small space will also look very crowded and hard Interaction) tools: predictive statements about prospective designs to read. Both of these methods must be used to achieve good have great utility. Comparing mini-QWERTY design alternatives, compromise between usability and the number of keys available. for example, this can be done at very early stages of the design process using these models. Here is a standard 62 key Finnish/Swedish QWERTY keyboard:

As conclusion it seems that none of the earlier studies concentrate on how to design a mini-QWERTY that effectively supports a wide variety of languages. Rennie and Cockburn (2005) pointed out that even standard QWERTY keyboards do not support multiple languages well without software aid. In addition most the earlier studies have been done using just English language as an example. This can be a problem as English is one of the easiest languages to support since it has fairly limited number of characters when compared to many other languages. Figure 4. Finnish/Swedish QWERTY layout for Microsoft Windows operating system [8].

3. DESIGNING QWERTY LAYOUT FOR MOBILE As an example, we can make it smaller by first reducing the DEVICE number of keys. Firstly, we can remove the first row except the Backspace key and the (red). Numbers and special characters can be accessed by using modifier keys and/or special In English and some other languages that have fairly limited character menu. Secondly, we can remove operating system character set there is not much need for character prioritization as specific keys. (Win Keys, Menu). Thirdly, we can remove all important characters fit nicely on any kind of QWERTY layout. duplicates (one of the Ctrl and Shift keys). We can also make However, this is not the case with most languages of the world so Shift, Ctrl, Alt, Alt Gr, Space, Enter and Backspace keys smaller. some kind of prioritization of the characters is clearly needed. However, it should be notified that a developer should always Typically the used prioritization may be for example following: consider very carefully how small he makes these keys, because it is also important to differentiate these very often used keys from 1) Lower case alphabets, the most used punctuation marks the other keys. So in practice it is highly recommended that (comma, period and apostrophe for example). developer leaves these keys bigger than the other keys or at least 2) Upper case alphabets, numbers and less used alphabets. somehow makes them visible and usable. 3) Less used punctuation marks. 4) Less used special characters. It is also possible to remove the keys that are not essential in the 5) Rarely used characters. mobile devices (Tab) or if the functionality of the key can be implemented by other means (). Less used symbol keys In practice there can be more or less above kind of priority levels can be also removed. They can be accessed by using modifier keys depending on the language a designer is working with and the and/or special character menu instead. purpose of the device the QWERTY is planned for. The first priority characters should, however, always have direct access The importance of dead keys depends heavily on the language from the QWERTY. variant. If a designer is planning to use the same layout for multiple languages it is recommended that he leaves space for at Many mobile devices on the market have implemented some kind least one dead key. In language variants that do not need dead key of a special character menu that can be opened to the screen of the the free space can be used for accented or special characters, for device by using one or two modifier keys. Normally, such menu example. reduces the need of complicated key combinations where two or more modifier keys are needed to input some character. It also A simplified example of a reduced QWERTY could look as reduces the needed printings on the keyboard. Reducing the following: complex key combinations is very important since mobile devices are used in many cases with thumbs only. The last priority characters can be placed on a special character menu if such is available in the device.

There are two methods to make the QWERTY smaller: reducing the number of keys and downscaling. Reducing the number of keys means that we simply remove the keys that are not necessarily needed in mobile device altogether from the Figure 5. Reduced Finnish/Swedish QWERTY layout. QWERTY. Downscaling means that the regular QWERTY is just squeezed down to smaller size. There is a limit, however, how

“Alt Gr” is in many mobile devices on the market replaced with The tests were recorded with video camera and log of the written shorter “Fn” term or with an icon. is often replaced with texts and writing speed was also recorded with separate test key named Sym, Chr or similar. That key can be usually used for software. In this test we were more concerned about the user scrolling through accented characters (e.g. á, ë) by holding it down comments and input errors rather than the actual writing speed and pressing character key (e.g. a, e) repeatedly. It can be usually because the speed is affected a lot by the used shapes and the also used for opening special character menu or list to the screen materials of the keys and we were not concentrating on those in of the device. (Figure 5) this study. We had a native speaking moderator for each language and two observers that were sitting in separate room and writing After reducing the number of keys it is possible to downscale the down notes of the test. layout to fit the available space in the device. How small it is possible to make these keys depends on the shape and the material English (the “feel”) of them. As result the keys should be big enough, however, so that all the printings on keys are still readable. There We had 12 English speaking test users, eight men, four women. can be up to four printings per key in some languages, especially if They were from 19 to 45 years old. Six of them had previous there is a Latin and some other script, for example Arabic, printed experience of using mini-QWERTY of some kind. Ten users were on the same keyboard. right handed and two left handed.

In our project we used 8(w) mm x 6(h) mm as a guideline for the The users liked the idea of having a familiar looking QWERTY target size of an individual key. It was not easy to follow that present in layout A but they thought that the size of the keys was guideline, however, since the available space for the keyboard was better on layout B. After the test the key size of both layouts was only about 85(w) mm x 25(h) mm. This kind of limited space is criticized, but nine users preferred the layout B because it had very common in pocket-size mobile devices. Two different layouts bigger keys which were easier to read. Most of the users preferred were defined, referred as layout A and layout B from hereafter. to use the QWERTY with two thumbs by holding the device with Layout A was a more conservative, PC-like, layout, but it had both hands. The locking functionality of Shift and Fn keys was not smaller keys. All keys in it were the same size except Space key clear to some of the users and some of them wished that the which was the size of three keys. It included modifier keys Shift, modifier keys were separated from the other keys by using color Fn, Chr and Ctrl and 31 character input keys. All in all it had 38 coding or similar. keys. The layout B was more adventurous, having six keys less, but the keys were bigger than in layout A, except the Space key Some of the users did not find out that Chr key could be used for which was smaller and located on the bottom right side of the opening the special character menu. One of the test sentenced layout. Keys in the layout B were closer to the set target size, required opening the menu because all of the special characters exceeding it in height. needed in the test were not available on QWERTY. It must be made clear to the users with suitable icon or abbreviation what the modifier keys mean. The universally used icon for is familiar to most of the people, but for example the “Chr” text used 4. TESTING THE LAYOUTS in the test layout is not clear to all. It also depends on language how these abbreviations are understood, so it might be a good idea The usability of layouts A and B were tested with English, French to localize them. Also, using an icon might be a good idea, but and Arabic languages with a total of 32 native speaking test users. good icons that are easily understood by all can be hard to come Mainly we were concerned whether these mini-QWERTY (or by too (this could be a good subject for further study). mini-AZERTY in the case of French) layouts were easy to use and if there was something preventing from using them effectively. Some of the users preferred that the numbers were placed in the Most of the users had previous experience of using mini- grid layout instead of the row to enable easier number input QWERTY on mobile device. especially now that the was required to enter numbers. The users also wanted to be able to enter the most commonly used The users were interviewed shortly before and after the test. punctuation marks without having to use modifier keys. Before the test we asked the first impressions of the QWERTY. All users used both layouts but every other user started with layout The big handed users had some difficulty in typing, especially B. After the test we asked what was good and bad about the with layout A. The most common error was that the users pressed layouts and proposals for improvement. Before the actual test the two or more keys at the same time. Some of the users thought that users were given two practice sentences to write. In the actual test it was too hard to reach the small space key at the bottom row in the users wrote 7 test sentences and after each sentence the users middle of the keyboard and preferred layout B that had it on the were asked how hard it was write and why it was hard or easy. At right side of the device. Most of the users found the Shift, Enter the end of the test the users were asked if they are right or left and Backspace keys easily. handed and their primary hand was drawn on paper to be measured afterwards. None of the users had very small hands. Each of them It is important that graphics printed on the keys are clear. It is also had from average to large size hands. important that modifier key and the secondary characters that can be inserted by using it are printed with color that is clearly different from the main color used in the primary characters. Not Arabic too many colors should be mixed however. We had 9 Arabic speaking test users, five men, four women. They French were from 21 to 31 years old. Five of them had previous experience of using mini-QWERTY of some kind. All of them We had 11 French speaking test users, seven men, four women. were right handed. The layouts used in Arabic tests were different They were from 24 to 45 years old. All of them had previous in that they had both Latin and Arabic printed on the same experience of using mini-AZERTY of some kind. Nine users were keyboard (i.e. more printings per key). Switching between input right handed and two left handed. AZERTY layouts were used in modes was done with a combination of two modifier keys. In test also. In this test the accented characters were emphasized; Arabic tests we were especially interested if from right to left there was much more need for the modifier keys than in the writing used in Arabic had any effect on the usability of the English layout test. We assumed that this fact will make the layouts and what the users thought about the switching between writing more difficult. We also wanted to know if the users were two input modes. Many of the test sentences required using both using the accented characters in the first place when using a Arabic and Latin input. mobile device. Also, we were seeking data on the character prioritization and layout since not all of the accented characters Most of the test users assumed before the test that the more PC- used in the French language fit to the keyboard. like layout A would be easier to use. After the test five users preferred the bigger keys in the layout B, however, because they The first impressions of the keyboard were similar to English were easier to read, but even in that layout the keys could have language users. Similarity with PC keyboard was liked in layout A been bigger. Small keyboard gives the impression that the but many users also commented on the size of the keys in favor of QWERTY is very crowded and hard to read. That slowed down the layout B. Most of the users felt that the mini-AZERTY type the writing with most test users. Seven users were confused keyboard is nice to have in a mobile device but not mandatory. because similar kind of characters were printed on the same key After the test six users preferred the layout B because of the bigger (there are no upper case characters in Arabic, so Shift can be used keys. It felt easier to use. for typing a different character).

We noticed that it is important to have a clear contrast between the The moderator had to help the test users in many cases because the colors used in the keyboard. It is very important that the users purpose of the modifier keys and the key combinations were not connect the modifier keys to the characters that can be typed by clear. Layouts should follow PC keyboard in how the modifier using the modifier keys. Also locking functionalities of the keys are placed. Based on the comments by the users changing of modifier keys should be clearly indicated to the users. Several the input should be very easy. In the test layouts it was users had problems because they had some lock on, but did not found to be too hard for most users. It can be made easier if PC- notice it. Using LED indication could be one solution for this like method is used. Input mode should not have any effect on how problem. special characters are typed. The users wanted to use the same method whether using Latin or Arabic input mode. The users wondered why some of the accented characters were printed on the keyboards but some were not. They also expected Arabic character placement was better in layout A compared to that all of the accented characters would be printed on the same layout B. In the layout A most of the basic characters had direct area on the keyboard and could be entered in the same manner. access and also the placement of the characters was better, more Our test layouts did not have a dead key and some users missed it. familiar. In addition, keys were less crowded. Some users had Most of the users also expected to find all of the accented difficulty in understanding which character is the direct access characters also from the special character menu. Most importantly, character in each key. Better color coding could have been used. there should be a clear indication to the user how to enter accented Some of the different forms of Arabic characters can be used by characters. Most users felt that it is important to use proper combining two basic characters, so it would not be necessary to language especially when writing other than SMS messages so have those forms occupying separate keys. they wanted easy way to enter accented characters too. Only a couple of users were not using accents but they only wrote SMS Summary of the test results messages with their mobile device. Neither one of the layouts was satisfactory enough, but most users Key combinations are always needed in the cases when there are preferred the layout B after using it in the tests just because it had less keys available than there are alphabets in the language. These bigger keys. Almost all of the users felt that it makes sense to key combinations are hard to indicate to the users clearly. Most of follow the familiar PC like QWERTY layout also in mobile the test users were looking for the combinations through trial and devices. error. When they found one combination, they tried the same logic to find other combinations too. Because of the difficult key If it is not possible to have direct access for all characters used in combinations some test users missed the help of some kind of the language the users could still find the characters if they were predictive text input method. printed on the familiar spot on the keyboard even if modifier key was needed to input the character. Characters printed on the

wrong place were not easily found even if the users had used them Some kind of compromise between how many accented characters multiple times during the test session. and how many special characters are printed on the keyboard must be done. The most common special characters should be also The layout of Arabic characters was cumbersome in both layouts. visible on the keyboard and the most common punctuation marks However, it was somewhat clearer in the layout A because it had should have direct access. Prioritization helps with the selection of more keys and more familiar placement of the keys. Input characters to be printed on the keyboard. Some languages have language change was too difficult with both layouts. Changing the such a vast number of characters that all of them can not be printed language is important in Arabic countries since they are mixing on the keyboard in any case and always require using of modifier Arabic and for example English words as they write. Typing the keys and some kind of software enablers that help with building or accented characters was also too cumbersome with both layouts. choosing the right character.

We did not notice any big difference in the usage of mini- It is a good idea to retain the PC-like layout of the keys. Having QWERTY between right and left handed users. The size of the alphabets in wrong positions was seen as a bigger problem than hands was more significant. Big handed users could reach all parts having to use modifier keys to type them. Implementing too of the QWERTY easily but they also easily pressed wrong key or complicated key combinations is not a good idea. It is better to more than one key at the same time (bigger problem with layout implement a software enabler like special character menu instead A). Familiar placement of the Space key in the middle of the that can be launched easily by using some modifier key for bottom row of the QWERTY was generally liked more than the example (the problem of clearly indicating the purpose of such Space key on the right side on the layout B. keys still remains).

Based on the test findings it is essential to have some kind of Even if neither of the tested QWERTY layouts was perfect, a predictive text input method when using mini-QWERTY combination of the two would be a usable solution for a mobile especially with languages that require using a lot of different device, the size and readability of the keys being most important. accented characters or alphabets since in many cases all of them The usability test results verified that the individual keys should be can not be printed on the keyboard and require using complicated at least the size of the 8(w) mm x 6(h) mm guideline that we used key combinations or special character menu for input. The users in this project to retain good usability and readability of the keys get easily frustrated when trying to find the different key although the shape and the material of the keys are important combinations. They even rather use special character menu or factors as well, especially if aiming for good writing speed. similar (if they can first find how to open it). Testing the QWERTY layouts with real native speaking users is essential. The most important finding from these tests was that it is really important to test the QWERTY layouts with real native speaking users. A large number of the requirements set by different 6. REFERENCES languages are easily overlooked because of the cultural differences. 1. S. Ager, Omniglot: Writing Systems and Languages of the World. http://www.omniglot.com/writing/, 2008 2. E. Clarkson, J. Clawson, K. Lyons and T Starner, An 5. CONCLUSION Empirical Study Of Typing Rates on Mini-QWERTY Keybords. ACM Press, 2005. Designing a mini-QWERTY is always a compromise between the 3. E. Clarkson, K. Lyons, J. Clawson and T. Starner, Revisiting size of the keys and the amount of keys. Two methods to shrink And Validating a Model of Two-Thumb Text Entry. ACM Press, the PC QWERTY were presented: reducing the number of keys 2007. and downscaling. If the keys are small, more keys can be fitted to 4. L. Rennie and A. Cockburn, Aiding Text Entry of Foreign the keyboard and there is less need for the modifier keys but the Alphabets With Visual Keyboard Plus. Australian Computer small keys are harder to read and use and that increases the amount Society, inc., 2005. of input errors and reduces input speed making the text input 5. S. Hwang and G. Lee, QWERTY-Like 3x4 Keypad Layouts slower than when using modifier keys. Accented characters are in for Mobile Phone. ACM Press, 2005. most cases typed by using modifier keys (even in PC keyboards) 6. N. Green, J. Kruger, C. Faldu and R. st. Amant, A Reduced but some languages have widely used accented characters that QWERTY Keyboard for Mobile Text Entry. ACM Press, 2004. should have direct access from the QWERTY. Prioritization can 7. J. Noyes, QWERTY-The Immortal Keyboard. Computing & be used also with accented characters. The most common ones Control Engineering Journal Volume 9, Issue 3, 1998, pp. 117- should have direct access or adding the accent should be made 122. easy (by using dead keys for example). The less used ones can be 8. Wikipedia: http://en.wikipedia.org/wiki/Keyboard_layout, printed on the keyboard with color clearly different from the main 2008 color used to indicate that modifier key (indicated naturally with the same color) must be used.