Multi-objective optimization of mobile phone keymaps for typing messages using a word list

Abstract

Most mobile phones today offer the option of using a word list to ease the typing of short messages (SMS). When a word list is used, a word is input as a sequence of digits by pressing the key corresponding to each letter once. The word list is used to look up the word(s) that correspond to this sequence of digits. This paper describes how a mobile phone keyboard layout can be obtained that is better suited for typing such messages. Two objectives are considered: the total cost of typing, and the total cost of word clashes that occur when a certain digit sequence corresponds to two or more words in the word list. A multi-start descent algorithm is developed to obtain a Pareto set of solutions.

Introduction

For many young people, SMS (short message service) has become one of the primary modes of communication. According to GSM Association, a consortium of mobile phone operators, the more than one billion GSM users in 205 countries worldwide sent 45 billion messages in February 2004 and are estimated to send over half a trillion messages in 2004 [2]. The SMS interface however, i.e., the mobile phone keyboard, was not originally conceived for typing text, but for entering telephone numbers. The distribution of the letters over the 10 numeric keys, shown in Fig. 1, was “optimized” for easy reference.¹ This alphabetic letter order is useful for typing American-style telephone numbers, such as 1-800 FLOWERS or 1-800 SAVEAPET, in which the number of characters typed is small. Each character corresponds to the number of the key it is on, e.g., “typing” FLOWERS corresponds to dialing 3569377. On the other hand, the combination 3569377 might correspond to any other word that consists of a letter on each of the pressed keys in the same order. It is the responsibility of the telephone operator to ensure that no other “word” is issued that corresponds to the same telephone number.

Messages transmitted by SMS do not correspond to telephone numbers and ambiguities such as the ones described above have to be avoided. Two systems have been developed to form words on a mobile phone keyboard. In the first generation of message typing, the first letter on a key corresponds to a single press, the second letter corresponds to a double press, and so on. This requires the user to press key 7 four times if an s is required and is obviously a very inefficient way to type even the smallest of texts.

The next generation of mobile phone keyboards, and the main interest of this paper, uses a word list to distinguish between words. Words are formed by pressing the key corresponding to each letter of the word once, thereby transforming the word into a sequence of digits. The system then searches the word list for words that correspond to this digit sequence and—if at least one word is found—displays the most frequent word. If the user intends a different word than the one found by the system, a special key (indicated with ∗) can be used to switch between this word and the other words that correspond to the digit sequence. For example, to type the word “suppose”, the user would type keys 7877673. However, the word “purpose” also corresponds to this key combination, and is more frequent. Therefore, the system will show the latter word and the user is required to use the special key to switch words. An example of such a technology is AOL’s T9^® Text Input [9].

Given the fact that the letters on mobile phone keyboards are no longer just used as memory aids, i.e., to remember telephone numbers, but instead are used to type relatively long messages, the question arises whether the ergonomic qualities of the keyboard cannot be improved. Some attempts have been made in the past to improve the physical design of the keyboard, but these do not seem to be able to rival the standard design in terms of popularity. In this paper, we use the default physical design (i.e., the 4 × 3 matrix shown in Fig. 1), but develop a method to find a better placement of the letters on the keyboard (we will refer to such a placement as keymap). We will focus exclusively on the second-generation input systems, that use a word list.

The keymaps obtained in this paper are intended only for typing messages in the English language. For different languages, however, the same approach can be used with a different word list. The keymaps that result from the same analysis using, e.g., a French corpus are most likely not the same as those obtained for English. This does not constitute a serious problem as the “physical” keymap is usually a small piece of plastic that can be removed and replaced, i.e., nationalization of mobile phone keyboards is not an insurmountable problem. This does not mean that a user will have to switch keymaps when typing a message in a different language. The English language keymap can still be used to type—say—French messages using a French word list, it will only be suboptimal for this language.

A second note is that an “optimized” keymap can still be used to type American-style telephone numbers. In this case, the input system can translate the word typed into a telephone number by remembering which letters are on which keys in the standard keymap. That is, letter ‘A’ should be transformed into digit 2, whatever key it is on.