A case to better fit sign languages into Wiktionary, and its prospects in Wikimedia projects

Wiktionary has done a wonderful job on documenting written languages, from major linguae francae to hinterland tongues with little record, and even recreated languages proposed by linguists who imagined what words our ancestors speak. However, there is one large gap in Wiktionary – sign languages. While major languages can have up to thousands or over a million of entries in Wiktionary, sign language only make a tiny fraction of entries. And the need to answer the gap is getting louder as awareness for deaf accessibility is gaining more attention.

It’s not that the Wiktionary community doesn’t care, but it’s just not designed for sign language corpora. Wiktionary is primarily text-based which suits perfectly for written language. However, this means that sign languages, being a visual-first communication, will naturally have a hard time to fit into the Wikitonary system.

This is what I questioned during the ESEAP Conference in Kaohsiung back in May, where I dedicated myself at the “mainstage” (as what I like to call it) to discuss this question. You can look at the presentation material here; this Diff writing can be considered as a bit more refined version of the presentation.

Before it all, we’ll look at one status quo: you can fit sign languages through text-based “transliterations” – symbols designed to transcribe how a word should be signed, and this is how Wiktionary records sign languages for the time being (here is one example done for the American Sign Language). However, this is not as intuitive as directly using visuals directly. It’s akin to learning Russian solely through Latin transliteration, or learning Chinese only through pinyin. It can work, but you will never get the pure visualization as the natives do. And at least for written languages, transliterations are typically in a script that you and I can read; for sign languages, it’s usually an entire set of symbolism that you need to learn from scratch (as the example below), meaning that you have to learn yet another “language” just to learn sign language.

As such, I wanted to do an experiment – how could I possibly fit sign language into Wiktionary in a better way?

In this writing, we will be using the English Wiktionary as our model. Other Wiktionaries have their own systems, so if you want to implement sign language into your Wiktionary, adjust accordingly.

First challenge: Page title

The first challenge, and perhaps the worst, is to determine what title the page for sign languages should have, because obviously, we cannot put pictures as titles. The most simple approach is to put something like “dog” for “dog” entries in sign languages.

In surface instinct, one may prefer to put American and British signs under the English word, or the Malaysian one under the Malay name (like my first example in my Kaohsiung presentation). However, sign languages have zero correspondence to any particular written language. American Sign Language (ASL) cannot be corresponded to English, and Malaysian Sign Language (BIM) cannot be corresponded to Malay, and same to all other sign languages. The sign of cat in ASL cannot be linked to “cat” and the BIM one cannot be simply tethered to “kucing” (“cat” in Malay).

As such, the best solution we can come up with here is putting all signs of a single concept under a single page. ASL, BIM and all others end up under the “cat” page for example. The language of the word will be based on the wiki’s language, and not a presumed written language tied to a sign language.

We can say the first challenge is solved here, but there is something we can improve here. When we apply this, sign language material will share the same page with written language material, which has a different nature in Wiktionary’s context.

We take the “cat” page example again, but now with sign language entries together with written langauges. All of our “cat” signs mean the animal, but written “cat” in doesn’t necessarily mean cat the animal as in English; in Malay, it means paint. Basically we’re mixing two natures in a page. Although not objectively wrong, but it’s good for us to give a separate place for both nature. For that, we cue in a new section of discussion.

A new namespace?

In the English Wiktionary, lexemes are added in three namespaces, differentiated by the nature of the languages:

• The default main namespace for most entries
• The “Reconstructed” namespace for proto-language entries
• The “Appendix” namespace for special-case languages (e.g. fictional and reconstructed languages)

As such, a nice way to separate this two natures is to dedicate a separate namespace for sign languages – a “Sign” namespace, perhaps. As such, all signs for “cat” can be listed under the page “Sign:cat”, as example, in the English Wiktionary. In the Malay version, that’ll be under the “kucing” page in its “Isyarat” namespace.

This is, I think, is the best way to tackle the page problem. No more worry over the sign-to-written dilemma as we put on a “neutral” solution based on individual Wiktionary version. And I think this format is far better than a title of, for example, “OpenB@Chin-PalmBack-OpenB@CenterChesthigh-PalmUp OpenB@Palm-PalmUp-OpenB@CenterChesthigh-PalmUp” which simply means “good” in ASL. That’ll definitely ease searching, whether from within Wiktionary, search engines or perhaps AI searching.

Challenge two: Entry content

The content of the entries can be written as usual like with written languages, but we need to do some adjustments. Ideally for me, a sign language entry should have the following:

• The visual of doing the signing – this should be de facto mandatory
• A textual instruction of doing the signing
• General info over the entry (e.g. type of speech, language)
• If possible, transliterations – this is the one I mentioned before; but instead, should become an additional info instead of being the main character, similar to how non-Latin lemmas have Latin transliteration, but as guidance.
• Etymology, references and further reading, if available

The “textual instruction” shall replace the typical “definition(s)” part in written languages. Perhaps the “textual instruction” will be a challenge if someone’s having a hard time to visualize signing into text, but other than that, this should be fine.

The final result

After tackling those, now we can finally build our model. Here is how my model worked as of now, you can look at this page here.

Compare for example, the page for “Wiktionary” in ASL as below. This is a bit similar to my proposal, other than the need for a new namespace and title format, and also the lack of visual material. Also, the separation of sign language “production” and the definition itself can also be discussed.

The future of sign language in Wikimedia projects

Sign language information is currently under a huge gap in Wikimedia projects, but also gives us a big field to work on. In Wikipedia, many sign language articles still doesn’t have any visual material, which is especially ironic on a material centered around visual-based communication.

Meanwhile, Wikibooks can be a good platform to construct “books” for sign language learning, which can be a big boost for sign languages with limited material online. I propose here an example of what I did in the Malay Wikibooks.

And the most important part is the very core: visuals of sign language in Commons, which is currently really lacking. There are only 42 categories under the “Sign language by language” parent category, in which many of them only have few material, typically only visualizations of a sign language in general, and not individual signs. You can refer to my works under the “Malaysian Sign Language” category.

We, me and you, can start with Commons – signs for the most basic stuff: letters, numbers, phrases, basic verbs. This depends on each sign language, major sign languages has extensive signing information. As a start, you could start with ASL where we have various websites such as PocketSign and HandSpeak – find signs that still hasn’t available in Commons and try uploading yourselves doing the signs. This is how I typically do for BIM. The pictures doesn’t have to be really that formal or professional; I simply do my signs from my bedroom as you see in the picture above.

You could then move to local sign languages in your community, although depending on existing material availability, this may need some research from yourself. For my work in BIM, this means crosschecking with signing in live news and various applications fragmented throughout the web. Then, put one example picture into Wikipedia. If your community has Wikibooks, write a beginners’ book like the one I provided above. Feel free to contact me if you do your part so I could incorporate your work into my projects as well.

Eventually, this is not just a “fill in the gaps” work, but an exciting innovation on Wikimedia projects. A free knowledge on sign language means helping communities to become more inclusive – connecting aging old folks with the newer generations, and defogging services to the public – and together, making the unhearable no longer be unheard.