/ 5 June 2015

Multilingual software improves access to data

Access to information in your home language is an important empowering tool, especially when it comes to the range of services provided by the government.

In order to facilitate the creation of documents in the government, the department of arts and culture has invested in the development of translation tools for South African languages through the Autshumato project. In some aspects the project is on par with, or outperforms, international machine translation systems.

“When you consider the limited human and financial resources we have in South Africa compared to the United States and Europe, we are doing well,” says Wildrich Fourie, project leader of the team that has developed the Autshumato range of translation tools.

The project team, based at the Centre for Text Technology (CTexT) of the North-West University, Potchefstroom Campus, named the project in Autshumato’s honour. 

This was the Khoisan name of the country’s first documented interpreter, known as Harry the Strandloper by the Dutch settlers who arrived at the Cape of Good Hope in 1652.

Promoting language parity

Autshumato is one of several human language technology projects that the department of arts and culture has been funding since 2007 to promote the use of all 11 official languages.

The Use of Official Languages Act, which promotes parity and the esteem of all the official languages, requires all government departments to give people access to information in their own language. “The problem is that most government documents, including legislation, gazettes and forms, are prepared in English and have to be translated into the other languages,” says Fourie.

The sheer volume of documents that must be translated by translation units in government departments leads to backlogs and long turnaround times. Human translators just cannot cope.

To assist government translators — as well as the many translators and editors working in the private sector — the Autshumato team has developed three main types of translation tools: machine translation systems, an integrated translation environment, and terminology management systems.

Speed, quality and consistency

Customised for South African circumstances, these systems can dramatically reduce the time it takes to do translations, says Dr Martin Puttkammer, head of CTexT.

“We conducted an informal experiment involving professional translators with experience in using Autshumato translation tools. Documents were first translated without using Autshumato ITE, and they were then asked to translate similar documents with the ITE. The translators achieved an average translation speed increase of 30% with the Autshumato ITE.”

Just as important as translation speed is the consistency and quality of machine translation. Again, Autshumato does not disappoint. “A translation is deemed acceptable when it is difficult to discern whether it was translated by a human or a machine,” says Puttkammer.

As with all their products, Autshumato is based on extensive research conducted at the university as well as internationally. “Interest in developing resources for so-called resource-scarce languages (languages for which few digital resources exist) has increased significantly over the last couple of years. 

“We are privileged to be involved with international initiatives on improving technologies for our South African languages.”

The Autshumato range of tools is being adopted by a growing number of translators in the government domain, including Parliament. The CTexT team supports them by running a helpdesk service and holding training workshops, such as the February 2014 and May 2015 workshops held in Cape Town for translators in Parliament.