| layout | main |
|---|---|
| title | PNG Pilot Project |
| css | aikuma |
{% include aikuma/navigation.html %}
This project grew out of earlier work in PNG that used digital voice recorders for large scale audio collection. Olympus donated 100 voice recorders to the project. For details please see the BOLD:PNG Project Website.
We visited the village of Imikori, in the Usarufa language area, from 15-18 February 2012. We trained 10 people to use the Android phones for recording and sharing stories.
On our first day in the village, we organised a large mumu (ground oven) and fed about 60 people of all ages. Then we set up a bench and invited people to tell stories to the assembled crowd. We asked that the children not be sent away, as had been suggested. There was no shortage of storytellers, and their stories were well received, sometimes with minor corrections from the audience.
We used this setting to communicate key aspects of our work: documenting traditional knowledge of the kind that can be shared publicly, demonstrating interest in the local language in front of a younger generation that has stopped learning it, and making continual use of recording devices.
On this occasion we were formally welcomed to the village, and elders spoke about the importance of recording their language for the benefit of future generations.
Each participant signed on to the system, speaking their name, typing it, and permitting their photo to be taken. User identities were automatically propagated to all the phones, so that any individual could use any phone for recording their speech.
We explained this process in English to one person. Then he explained it in Usarufa to the others. Participants helped each other to identify the location of the microphone, to use the touchscreen keyboard, and to take a photograph. We were surprised how quickly people were able to use the touchscreens, even though they had not experienced this technology before.
Once registered, individuals went off on their own and started to record stories. (Given that the software prototype is at an early stage, we had not disabled the touchscreen during recording, and so people could not hold the phone to their face in the usual way. This did not seem to compromise the naturalness of the recording activity.)
In some cases, it was easiest if one person held the phone to record another. This woman told the story of her life, and used discourse styles that were not apparent in the speech of younger people.
(Hear this recording translated in the oral annotation)
The content of each phone was automatically shared with the other phones, enabling people to listen to each other's stories. An experimental +1 feature enabled people to indicate whether they liked a story. Stories were ranked by popularity and recency.
We tested two mobile versions for respeaking, or oral annotation. This involves two activities, careful respeaking, in which the original phrase is repeated carefully and slowly, and oral translation, in which the original phrase is translated into English.
In the approach pictured here, the user must press and hold the play button to hear a recorded original, then release it to stop playback. When ready, the user must press and hold the record button to supply an oral annotation. This is linked to the corresponding part of the original recording, to support later work in producing a written transcription and translation.
A second approach we tested was a speech controlled interface, in which playback and recording was solely controlled by voice. This approach is not pictured, since it is simply a person holding a phone, as one would in a phone call.
For our test subject, both approaches proved to be feasible from a human computer interface standpoint.
The setup consisted of seven Huawei U8510-1 mobile phones with Android version 2.3.3 and the latest version of our BOLD application prototype. The phones were networked using an ASUS NT-16P wireless router with FTP file storage on a 4GB USB memory stick. This memory stick was used to synchronize recordings and user accounts between phones via wireless.
We used a 70Ah lead-acid battery and 230V inverter to power the router and the phones. The router drew 1A, while the phones drew a total of 3.5A when recharging. The car battery proved enough to power the whole setup for three days (2 major experiments of three hours each) including two full charges of a MacBook Pro.
This configuration was necessary for working off-grid. The software just requires the phones to have (occasional) broadband access for synchronizing data to a central server.
Programming for Android during the night to get it ready by the next morning.
It all would not have been possible without the generous donations of time, effort and lodging by the Usarufa community, particularly the family of Waks Mark.
We are grateful to Matt Taylor of SIL PNG for logistical support, and to Andrew Bird for catering and video work.
