9 March 2013, 2.50pm AEST
Androids in Amazonia: recording an endangered language
Associate Professor in Computer Science at University of Melbourne
Steven Bird receives funding from the Australian Research Council and the US National Science Foundation.
The University of Melbourne
Provides funding as a Founding Partner of The Conversation.
Augustine, one of the few remaining speakers of Tembé, recording a story using an Android phone. Steven Bird
The village of Akazu’yw lies in the rainforest, a day’s drive from the state capital of Belém, deep in the Brazilian Amazon.
Last week I went to Akazu’yw, carrying a dozen Android phones with a specialised app for recording speech. My students and I had developed the software in Melbourne. How effectively would it work in Amazonia?
The plan was to record and translate the stories of the surviving speakers of Tembé, a critically endangered language of the Amazon, with just 150 speakers left.
We wanted to test our mobile phone app, an inexpensive and scalable method for capturing the world’s vanishing languages while there is still time.
Travelling with me were Katie Gelbart and Isaac McAlister. Katie had been exploring South America for six months, and Isaac had been on holiday with his ornithologist uncle in Manaus, further up the Amazon river.
By chance, both arrived in Belém a few days before I did, got wind of my project, and asked to come along for the ride.
So it was that we found ourselves clambering onto a bus and heading out into the unknown, wondering what adventure awaited us.
Heading out of range
As we settled into the journey, I described the project to Katie and Isaac, and demonstrated the app. Aikuma lets you record stories and translate them phrase by phrase. Since only audio is involved, there’s no requirement for the user to be literate.
In fact, most endangered languages are not even written.
Aikuma also supports sharing: people can listen to recordings that were made on other phones, and “+1” any stories they like. Such social media is run-of-the-mill to digital natives. But the Tembé village is out of range of the internet.
This trip was to be an experiment in bringing digital culture to “analogue natives”.
After the five-hour journey we arrived in the town of Paragominas, where we would leave the paved road. A man called Emidio welcomed us with a broad smile. We towered over him; Emidio couldn’t have been more than five feet tall. I liked him instantly.
He was the champion of the Tembé language. I would have been Emidio had I come from his village. Or was this just a fanciful projection, an attempt to create something familiar that I could identify with?
After stocking up on provisions, we climbed into a 4WD truck and set off into the rainforest, bouncing along the rough road, crossing over the digital divide.
Three hours later, hot and exhausted, we arrived in the sleepy village and pulled up in front of the chief’s house, having no clue what to expect. We were welcomed, chairs were arranged in a circle, and the chief sat, arms crossed, enquiring:
Have you brought the internet?
Had I brought the internet? I had been forewarned that the chief had high expectations, but the question shocked me. We were 120km from the nearest town. I struggled to explain. Bringing the internet was impossible.
Now the chief was confused. Recovering, I demonstrated our phone app, and recorded the chief offering Tembé greetings. He didn’t know any Tembé stories, he said, then repeated: What about the internet?
Things were off to a bad start.
The next day, the village was taken up with preparations for an all-night coming-of-age ceremony. We agreed to make a video of the event.
Everyone, including us, was tattooed with the Tembé pattern of snake tracks and fish scales using juice from the fruit of the peach palm tree. I was pleased to be sporting this tangible sign of local acceptance, though later, after meeting Puluta, I had some misgivings.
Dance and music at the all night Tembé festival. Steven Bird
Click to enlarge
We asked people to talk about the festival in the Tembé language. We handed out phones, and they put them to their ears and began talking. It didn’t seem to matter that no-one was listening on the other end.
It all went smoothly. This first small success delivered disproportionately large encouragement. We had fluent speakers who could record texts!
Augustine was a wizened virtual archive; Elias was reluctant yet obliging when pressed; Emidio, as the the village organiser, was hard to pin down.
Puluta was proud of his Indian identity and sceptical about our role.
This was a start. But who else could speak Tembé? I made several passes through the village at different times of day, making enquiries, without making any new discoveries. We heard some women spoke Tembé, but none of them would interact with us.
A short presentation by Steven about his visit to the Tembé village.
Struggling against the flow
After a couple of days we had recordings from six fluent speakers: stories about the way things used to be, stories about peoples’ lives, accounts of the festival, instructions on how to make a traditional wooden bowl, and so on. I kept wondering where to find more speakers, but there weren’t any.
Meanwhile, there were technical issues. In between Tembé recording sessions, I took time out to troubleshoot the fileserver on our Wi-Fi router.
With the 30°C heat and 90% humidity, and being so far removed from the office, it was simply too difficult to get into the headspace of computer network diagnosis.
So I abandoned the plan to synchronise the phones. The audio sharing and social networking experiment would have to wait. Disappointed, I consoled myself with the fact we didn’t have enough fluent speakers to do the experiment in any case.
Katie, Emidio, Isaac, and Elias, discussing the work during a break. Steven Bird
Click to enlarge
I exercised off my frustrations in the wide brown river at the bottom of the village and – in an apt metaphor for slow fieldwork – drained my energies pushing against the flow while making no observable progress.
Yet the river was cool and refreshing and it carried away the sweaty grime and filled me with calmness and energy and resolution. We needed to get the stories interpreted.
Getting at the meaning
The next step was to translate the stories, again using the phones. I wandered around the village with a pocketful of unsynchronised phones, each with its own collection of recordings.
Where was Elias? We had agreed to meet at 3pm but he didn’t arrive, and now at 3:30 I couldn’t find him. Who else should I try?
Perhaps Puluta would be back from fishing. Sure enough, he was, and Elias with him. I sat and listened to conversational Tembé for the first time. This was the full, real, everyday language.
During a lull in the conversation I asked if I could record the two men speaking. They agreed, or so I thought, but then Elias excused himself and departed. More confusion. Fortunately, Puluta agreed to stay and interpret stories for me.
I was delighted to see how easy it was to translate stories using our simple thumb-controlled app: hold down the left thumb to play a phrase of the original recording, and hold down the right thumb to record the translation. Over the next couple of days I managed to get every story translated at least once; some twice.
Puluta using the phone to talk about who the “real Indians” were – not the white researchers with their Tembé tattoos.
Once, as we relaxed in our hammocks in the cool evening breeze, Katie asked how much work it would be to transcribe our recordings.
We had already seen that Tembé does not have a standard writing system that people can use; they only know how to write in the national language, Portuguese.
That’s how it is for most of the world’s endangered languages, I explained. Linguists can transcribe speech using the International Phonetic Alphabet, but that takes up to an hour for every minute of recording.
Tánia, a Tembé girl, being raised to speak only Portuguese. Steven Bird
Click to enlarge
There is ultimately no way to avoid this pernicious transcription bottleneck. But our mobile-phone software is designed to at least postpone the problem.
We collect and archive language recordings now while the speakers are still alive. That’s all. We have the whole of the future to transcribe and process the recordings.
After four days, it was clear we had done all we could, so we made copies of all our recordings to leave behind, and planned our departure the next morning. A truck would be ready at 7am.
I was relieved that we had reached the end of our stay without falling ill or being bitten by the cobras that come out at night.
But the strangeness, the incomprehensibility, of so many things in the village now weighed heavily on us. And the most dangerous leg of our journey was still to come.
We were ready at 7am, but the truck wasn’t moving until a group had arrived from down-river. They were already on their way, we were informed, and would be with us in an hour. The day wore on, and with each passing hour, creative new reasons were offered for the delay.
They finally arrived at 3pm and loaded the truck full, inside and out. My protests fell on deaf ears. We squeezed ourselves and our gear into the corners of the tray. Somehow Emidio squeezed himself in as well. Then, as we drove off, the skies opened.
This quickly became the most hair-raising journey of my life, as the driver pushed to get back to Paragominas by nightfall, skating over bumpy, muddy roads. He had lost a passenger off the back of the tray on our way in, fortunately with only minor injuries, so we held on extra tight for four difficult hours. The rain kept coming, falling heavy and sharp on our faces.
It was the closest thing to being submerged. In a moment of comic genius, Isaac taught Emidio to sing You are my Sunshine, and his rendition got most of the vowels wrong and drove us into fits of hysterical laughter.
Somehow we survived those four long hours. Finally in Paragominas, we changed into dry clothes and collapsed on the midnight bus back to Belém, arriving at 6am exhausted but happy. A short 48 hours later we would be on our next trip.
Safe and sound
My electronics were safely dry inside waterproof containers buried deep in my luggage, so we still had functioning equipment. My detailed notes about our work were dry.
And there were our precious recordings, a collection of stories with their translations, a linguistic snapshot from one of the 100 languages of Amazonia.
To close with the words of Augustine Tembé, recorded after the coming-of-age ceremony:
We finished the singing night. We sang songs of monkeys, music of birds, of the genipapo fruit …
We sang the music of the onca, the little sarapo fish. Also the music of the tatu, which is the one that has a shield called tatu tatu tatu. We sang genipapera also we sang the music of the little frog. All the songs of animals we sang this night, even the music of the monkeys.
We sang the music of the little bird called nambuzim … We all sang, sang all the songs of the animals of the forest. And we sang as well the music of one of those little cabo that have a very long tail, and the music of the tucano bird. We finished to sing on that night.
All of us sang that night. This is all.