Designing an IPA keyboard for the iPad

Phonetics will always hold a special place in my heart. It’s what first got me interested in Linguistics. Making exotic sounds and representing them with even more exotic characters—it was love at first phoneme.

Anyway, those exotic characters I fell in love with of course make up the International Phonetic Alphabet, or IPA. IPA is a standard way to represent the spectrum of human speech across virtually every language with a sophisticated level of nuance. It’s as essential to Linguistics, both academically and culturally, as the periodic table is to Chemistry.

Yet despite its immense significance, there is an utter lack of hardware or software to support its transcription onto computers. For anyone not lucky enough to be fluent in LaTeX, it’s an arduous process of ALT- codes and copy-pasta. This is not just a problem for phonologists. I believe that this technological disconnect is the single greatest barrier to the growth of Linguistics as a field.

At the very least, this is an interesting problem for someone interested in technology, design, and linguistics. As luck would have it, the problem just got a whole lot easier with the release of the iPad.

You Can’t Spell iPad Without IPA

For the better part of the last century, our lives have been controlled by the push of a button—physical interfaces resulting in electromechanical output. With the iPhone, iPad, and similar devices, we are seeing a transition into a new paradigm of touch screen interfaces, wherein the physical interface becomes virtual, able to dynamically adapt as needed to fit any context.

In these heady times, I though it’d be cool to employ some Blue Sky Solutioneering™, and see what’s possible. But first, let’s set some design goals to make sure we’re heading in the right direction:

• Efficient: An input method’s sole function is to be an optimal method of transcribing thought into representation. A new IPA interface would be successful if and only if it is able to offer an improvement over competing alternatives. Though the bar is set pretty low in this respect, it should be at least comparable to the speed one could achieve writing in, say, Kirshenbaum, or ASCII-IPA.

• Intuitive / Familiar: When one refers to something as being “intuitive”, they usually mean “familiar”. Thus, an ideal interface should capture the best organizing principles for IPA and translate them into an interface with ideas borrowed from various input methods for other orthographies. As ubiquitous as IPA charts are in understanding phonology, there is much to be glanced from earlier traditions, such as the foundational work done by Jakobson, Trubetzkoy, and others from the Prague Circle, or Chomsky and Halle’s seminal “The Sound Pattern of English”.

• Completeness: Given the dozens of symbols and diacritics in its repertoire, representing IPA in its entirety becomes quite a daunting goal. In an ideal interface, one would have access to the full range of the alphabet. However, since usability is the main priority, certain pragmatic affordances might have to be made.

Inspiration

Keeping these goals in mind, I set out to create some initial designs to iterate upon. I drew heavily on the international keyboards available on the iPad. Here are some concepts I found to be informative, and how I translated them into an interface for IPA.

Romanized Japanese Input

Compared to English, Japanese has a fairly constrained phonetic inventory. Using the Hepburn romanization system, one can transcribe Japanese syllables by typing how it would be spelled in English. For instance, ma becomes ま tsu turns into つ, and a is written as あ. Kanji has many homophones, so for example, あお could be written as 青, 碧, 蒼, and 襖. An IME, or input method editor, suggests characters that correspond to a particular phonetic input, from which a user picks what they meant. In practice, IMEs are orders of magnitude faster than having to write characters by hand. Even in Japan, where keyboards have a direct one-to-one mapping of kana onto the keyboard, most people prefer this romanized method of input.

One could imagine a similar method of text expansion being used for IPA input. For example, if a user types in th, the system could recommend th, θ and tʰ. Taken a step further, suggestions could also include voiced variants (so ð in addition to θ), or in the case of vowels, any related monophthongs or diphthongs. There are several standard ASCII ↔ IPA schemes that could be included as well, such as X-SAMPA.

Chinese Handwriting Input

While phonetic input using Kana for Japanese or Pinyin for Chinese is an effective means of transcribing text, there are some cases in which handwritten input is desirable (for instance, characters that you don’t know how to pronounce). Based on the position of your strokes, a handwriting interface like the one shown above can present you with possible matches. Similar to the romanized Japanese interface, a rough input is refined to an exact match using suggestions.

Borrowing from this concept, I designed a more intuitive system for transcribing vowels. Unlike consonants, which are relatively distinct from one another, vowels are nebulous, flowing one into another. Although the same method of ASCII expansion described above would allow for vowel transcription as well, it’s often hard enough to figure out which vowel is the best fit (let alone what its ASCII representation is).

As such, I employed that familiar trapezoid structure, which spatially maps vowels to regions of the mouth. Just touch the general region of articulation, and select the desired vowel from the list of nearest neighbors provided on the right. As you drag your finger around, the suggestions would dynamically update accordingly. Not only is this a convenient interface, but it serves as a useful reference to enforce the relationships between the vowels and their place of articulation.

Long Touch Gestures

One of the interface idioms used on the iPad is to display a contextual menu when you tap and hold certain objects for a second or two. Pictured above above, I touch and hold down on the a key to reveal a collection of related characters and diacritical variants.

With so many ways to ornament characters, this method of hiding complexity is nearly essential to keeping the interface clean and functional. For instance, by tapping and holding t, we could expect to see tʰ, tʷ, and tˠ, along with all of its voiced and other coarticulated variants.

Scratching the Surface

Of course, this is only the beginning in thinking about the potential for touch-screen IPA interfaces. I just wanted to get these ideas out there, so I could get some initial feedback.

If you have any ideas for how to organize or present the alphabet, or how to improve the UI, please put them in the comments! I’ll continue to keep iterating on these ideas, and hopefully get a working prototype out soon.