|
Aditi Sharma, an expert on voice user interface design
|
Lwazi becomes a reality: the Sterkspruit community development worker and a community member during the pilot
|
HLT experts converge at the National HLT Network Day held in January 2010
|
South Africans now have the opportunity to use a mother language of their choice when interacting with technology, thanks to the research of Aditi Sharma. Her work in the human language technology (HLT) research group of the CSIR Meraka Institute, supports the use and promotion of mother language. The use of mother language is endorsed by UNESCO (United Nations Educational, Scientific and Cultural Organization). This has resulted in the annual celebration of International Mother Language Day on 21 February to promote linguistic and cultural diversity and multilingualism.
South Africa is a multilingual country. Besides the 11 officially recognised languages , scores of others languages – African, European, Asian and more – are spoken here.
Aditi’s expertise lies in voice user interface design for the developing world. “By designing the voice user interface for different languages, we ensure that people feel comfortable using interactive systems,” she explains. “It needs to be simple and fast though – we don’t want them to abandon the call!”
Voice user interfaces (VUI) are not common in the developing world. Aditi’s research considers all aspects relating to human factors, such as culture, literacy and socio-economic circumstances, as well as structural aspects, such as the call flow, the content and quality of the feedback that the user will hear when using a VUI application.
The dialogue or user interface design model developed by Aditi and IBM colleagues during her 2008 stint as an intern at the IBM Research Labs of the T. J. Watson Center in Yorktown, NY, USA, consists of three components: Get input; error-recovery and play results/output.
The quality and appropriateness of translations used in voice user interfaces is of great importance in these systems. “There is no simple one-to-one translation,” she notes. “Localisation or adaptation is important to ensure that the culture associated with each local language is captured in translation, persona and user interface metaphors.” This aspect known as enculturation, tailors the functionalities and interactive aspects to the cultural values and needs of the user population. It may at times result in the use of longer or more explanatory translations.
To date, voice user interface research has been used mainly in the CSIR’s OpenPhone and Lwazi projects. OpenPhone is a pilot telephone-based HIV/Aids community-oriented VUI service that makes use of language technologies to address informational needs of caregivers of children with HIV. Plans are afoot to expand this type of service further into Africa and beyond in 2010.
Lwazi, a project funded by the Department of Arts and Culture, is a telephone-based, speech-driven information service for South Africa. The project provides South Africans with an opportunity to access government information and services in any of South Africa's eleven official languages using either landline telephones or mobile telephones, free of charge.
Voice user interfaces have several applications, such as in telephone-based systems (e.g. call centres and directory assistance services), or audio-search and access-control systems.
Aditi’s group is currently exploring future options and additional domains of application as part of a follow-up project known as Lwazi II. Lwazi II aims to improve on the human language technologies developed during Lwazi and implement these in applications that will have an impact on the lives of South African citizens.
In addition to her niche role as a researcher in voice user interfaces in the developing world, Aditi has also been the lead researcher in a technology audit conducted on the local HLT landscape – an audit which was the first of its kind and scale in South Africa. This was done under the auspices of the National HLT Network (NHN), an initiative supported by the Department of Science and Technology which commenced in 2005.
The HLT landscape in South Africa is made up of a relatively young R&D community of universities,science councils and a handful of companies in the private sector. The audit was motivated by the fact that numerous local HLT R&D activities have been taking place with many instances of international recognition for the quality and impact of work. South Africa has invested a reasonable amount of capital into the development of HLT. However, in contrast, the potential for impact that HLT has, as an enabling technology across various application domains, is not aligned with the current scale of HLT activities in the country.
Aditi points out, “Despite a number of efforts by government and the R&D community, South Africa has not yet capitalised on the opportunities of HLT to create a thriving HLT industry. One of the key challenges is the perceived fragmentation of the R&D activities in this domain. This problem is further exacerbated by the fact that HLT development in South Africa needs to occur across the eleven official languages.” These languages are regarded as resource-scarce languages, adding further to the challenges facing HLT researchers in South Africa.
The aim of her research was twofold: Firstly, to find out what exists in terms of HLT language resources and the current state of these resources. This information enables the broader South African HLT community to create technology road-maps and plan for future research and development (R&D).
Secondly, the NHN HLT technology audit was the topic of Aditi’s MEng degree in Technology Management, under the supervision of Professor Gerhard van Huyssteen, research group leader of the HLT research group at the CSIR, and Professor M.W. Pretorius, chair of the Graduate School of Technology Management at the University of Pretoria
The findings of her research reveal that a number of HLT language resources and applications are available in SA, but are of a rather basic and exploratory nature and much investment still needs to be made in the development of these resources. The comparative language analysis indicates that HLT development is relatively more advanced in languages such as Afrikaans, South African English, and that African languages lag behind, though isiZulu, isiXhosa, Setswana, Sesotho and Sepedi have more HLT activity compared to other minority languages. Several recommendations have also been made for accelerating HLT development in SA.
During 2009, this audit already guided the structure and initiatives of the Department of Arts and Culture’s National Centre for Human Language Technology, and has been used to inform priorities for future HLT funding.
As one of the outcomes of the audit, Aditi and her colleagues will make available the information gathered on South African language and technology resources through the means of a database with a web-based front-end to provide the HLT community with access to the audit information.
“By designing the voice user interface for different languages, we ensure that people feel comfortable using interactive systems”
|
With her Master’s degree now completed – she graduates with distinction in April 2010 – Aditi envisages working on other types of applications to make VUI services more usable and accessible for the developing world.
With less pressure on immediate academic deliverables, she looks forward to spending time with her husband, Hemant Grover, a childhood friend who became her husband a year ago.
|
