University students develop computer software to help effective service delivery
The increasing use of smartphones in society is seeing social media platforms like Facebook, Twitter, Instagram and WhatsApp becoming the primary means of communication for many people. One of the ways people use these platforms is to vent their frustration about poor service delivery.
Four data science students have seen this as an opportunity to assist government improve service delivery, and have developed a computer software application that will help identify problems.
The students, Thabang Lebese, Nobuhle Moyo, Thokozane Mlotshwa and Sphesihle Zitha, participated in the Data Science for Impact and Decision Enablement (DSIDE) programme from December 2019 to January 2020. DSIDE, funded by the Department of Science and Innovation and hosted at the Council for Scientific and Industrial Research, targets higher education students in disciplines such as physics, mathematics and computing to support capacity building in the fast-growing field of data science.
Recruits from higher education institutions are mentored in learn-by-doing problem solving to address real-world needs identified by various stakeholders. Over the past few years, students participating in the DSIDE programme have worked on wide-ranging, data-intensive projects to help address service delivery challenges at local and provincial government level.
Lebese, Moyo, Mlotshwa and Zitha based their Project Phakiša idea on the Presidential Hotline, which was established in 2009 to allow the public to raise issues about problems with service delivery from government departments and agencies. The four are using data from Twitter to understand people's views on service delivery. This is important, as it could help affected departments implement corrective measures, thus serving citizens better and avoiding disruptive protests.
Although pre-cleaned data showed significant engagement from beyond South Africa (showing interest in the country's activities from the diaspora and others outside the country), the project used only open source data from within South Africa's borders. This was acquired from tweets addressed to particular ministers, government departments and institutions using an automated daily scraping script, a technique employed to automatically extract data and save it to a local database.
The students worked with open source packages and libraries that are widely used in the machine learning and natural language processing space. Machine learning is an application of artificial intelligence that gives systems the ability to automatically learn and improve from experience without being explicitly programmed, while natural language processing is a field of artificial intelligence in which computers analyse, understand, and derive meaning from human language in a smart and useful way for interactions between computers and human languages, in particular looking at how to program computers to process and analyse large amounts of natural language data.
Their models used supervised methods (where algorithms are fed a fully labelled dataset to predict future events) and unsupervised methods (which use information that is neither classified nor labeled), coupled with linguistic methods that are used extensively by corporations, to draw general sentiments from acquired data.
The students' observations indicate that matters raised on Twitter are not given much consideration by government. However, from remarks gleaned from their analysis, it appears that South Africans are generally unhappy about government services, and their communications with and about government need to be acknowledged.
Lebese said that data extraction and analysis was made more difficult by slang and informal language, as well as the non-standard use of emoticons. Many tweets are not made in English either, although with most of them in one of the country's official languages, this should be resolvable.
According to the data extracted, people in lower income groups are most likely to try and communicate about service delivery problems using social media, so the information obtained is skewed towards this demographic group.
Highlighting the capability of the project, Lebese maintains that it can be used to analyse and model different aspects and scenarios, for example, the implications of the COVID-19 pandemic on service delivery.
