1) Stanford University HCI Group
2) IBM India Research Laboratory
3) Development Support Center (Gujarat, India)
4) UC Berkeley School of Information
SUMMARY:
This paper presents a study comparing speech and dialed input voice user interfaces for farmers in Gujurat, India. India has over 480 million illiterate people who need an access to information. For some of them spoken language is the only way for them to obtain information. They could possibly obtain information anytime using a phone that gets information from a recording. Of course this is restricted only to people who have access to a phone.
In this study the authors performed a study on 45 participants. All of them were people living in rural India. 87% of the participants had never used a computer. 73% of the participants had less than an 8th grade education. The participants were both male and female. Due to some traveling constraints 7 of the women had to be observed within their own residence.
The authors tested two methods of interaction with the user. The first is called Voice User Interface (VUI). VUI is where the application will get user input by them speaking a word. Most of the questions in this study were yes-no questions. The authors managed to get a 2% error rate by using just 15 voices to compare to. Previous work shows that users prefer VUI for non-linear tasks. An example of this would be checking voicemails in a certain order.
The second method the authors used was called Dual-tone multi-frequency (DTMF). This works like voicemail. The user is presented with several options and then hits the corresponding number on their phone. Previous work shows that users prefer DTMF when they are doing linear tasks. An example of this would be listening to all your voicemails in order.
The prototype that the authors built was called Avaaj Otalo. It is a Gujarati language application that allows farmers to access agricultural information over the phone. The farmers could perform any of three tasks: listen to announcements, listen to radio archives, and post questions. These are in order of increased difficulty.
The overall completion rate was 74% for DTMF and 61% for VUI. The third task was the most difficult because it involved categorizing a question, recording the question, and recording your name and location. The completion rates for task 1 were 48% and 42%. The completion rate for task 2 was 81% and 67%. The completion rate for task 3 was 86% and 54%. Each time DTMF outperformed VUI. However, there was no significant difference in user satisfaction. Over 80% in both groups said they found it easy to access information on the system and 75% said they would use such an application. Most farmers said an ability to access such information anytime would dramatically change their lives.
DISCUSSION:
Although most people don't realize it, they use DTMF whenever they check their voicemail. This seems like such a simple task but this could be a useful application for farmers in India. I am not surprised DTMF outperformed performed VUI because I never use VUI on my phone. One fault in this work is that for VUI the questions were mostly yes/no questions. I believe a true VUI should accept any word as input. If this were the case error rates would probably be much higher.
Its hard to imagine but a product like this could change the lives of millions. We tend to take it for granted that we have such quick access to knowledge on the internet. Many people are still living in a developing world. Technologies such as DTMF that we take for granted could improve the lives of millions of people on the other side of the world.
No comments:
Post a Comment