Das Web 2.0 ist u.a. auch eine weltweite Plattform für Meinungsäußerungen. Immer mehr Kunden diskutieren online über Produkte und tauschen Erfahrungen aus. Die Analyse der Online Beiträge stellt daher ein wichtiges Marktforschungsinstrument dar. Es wird ein Ansatz zur automatischen Identifikation, Aggregation und Analyse von Meinungen mittels Text Mining vorgestellt und dessen Anwendung an einem Beispiel aus der Sportartikelindustrie aufgezeigt.

The poor performance by South African pupils in The Third International Mathematics and Science Study highlighted the importance of and need for Continuing Professional Development (CPD) for South African mathematics teachers. For these teachers, the sudden, rapid and dramatic advent of the World Wide Web (WWW) and its communication conduit, the Internet, with its multimedia capabilities, interactive tools and telecommunication facilities, seems full of potential as a catalyst for significant and sustained online CPD activities. However, the Internet's usefulness for mathematics and spontaneous mathematical interaction is severely limited. Against this background, the motivation for this study was born out of two beliefs - a belief that context needs to be considered in online endeavours, particularly given the disparities that exist between disadvantaged and advantaged teachers in the South African context; and the belief that a bottoms-up approach to community formation allows space for a self-organizing system whose continual health and functioning is dependent upon local ownership and member identification. Having developed a mathematics-friendly online forum environment (ODEM) that allows teachers to include mathematical expressions in their posts, this study investigated the personal and situational tensions impacting on the use and value of this appropriate forum environment as a reflective tool in pursuit of CPD. Two groups of disadvantaged and advantaged mathematics teachers were separately provided with Personal Computers and home Internet access, thereby creating opportunities for reflection, communication with colleagues and the exchange of knowledge and ideas. Little evidence of community growth was found, while disadvantaged teachers faced more tensions than advantaged teachers in using the ODEM. Despite these differences, both groups' tensions pointed to their real (and thus forum) needs not being addressed. These forum needs are related to issues arising from their practice and the need for a channel of communication to a Subject Advisor that should actively manage these needs. A vertical relationship with the Subject Advisor is preferred over collegial interaction, over the needs to include expressions in their posts, or to reflect on their practice. Until teachers' needs are resolved, the ODEM is thus perceived to have potential value. The results furthermore informed a model that can be used by a Subject Advisor to determine teachers' tensions and needs in context, thereby ensuring appropriate online CPD strategies. / Mathematics, Science and Technology-Education / Ph. D. (Mathematics, Science and Technology-Education)

Das Web 2.0 ist u.a. auch eine weltweite Plattform für Meinungsäußerungen. Immer mehr Kunden diskutieren online über Produkte und tauschen Erfahrungen aus. Die Analyse der Online Beiträge stellt daher ein wichtiges Marktforschungsinstrument dar. Es wird ein Ansatz zur automatischen Identifikation, Aggregation und Analyse von Meinungen mittels Text Mining vorgestellt und dessen Anwendung an einem Beispiel aus der Sportartikelindustrie aufgezeigt.

[pt] Esta pesquisa de tese teve como objetivo explorar a análise de mídias sociais, especialmente as disponíveis em comunidades online de sites de redes sociais, a fim de realizar estudos sociais sobre questões de saúde. Com base em uma abordagem prática foi definido um processo para realizar esses estudos. Este processo contou com ferramentas computacionais adaptados para fornecer apoio em tarefas específicas, tais como recuperação de conteúdo, seleção e análise. Duas ferramentas que se destacam são apresentadas por causa de sua utilidade e a complexidade do processo em que a sua construção se baseou. Para o benefício da análise de comunidades online, o Mapa de Associação de Comunidades é um processo desenvolvido para apoiar especialistas em compreender os interesses dos usuários com base em suas associações dentro de suas comunidades. A outra ferramenta visa auxiliar analistas a selecionar discussões de fóruns online a serem analisados manualmente com técnicas de pesquisa qualitativa, por exemplo, análise de conteúdo e do discurso. Esta ferramenta, TorchSR, foi criada baseada em aprendizado de máquina não supervisionado, usando agrupamento hierárquico, para dar suporte na resolução do problema de seleção de conteúdo. Um estudo de caso exploratório mostra que esta ferramenta ajuda na resolução do problema. O processo proposto foi utilizado em dois estudos sobre questões relevantes de saúde (hepatite C e o abuso de drogas), que resultou em descobertas relevantes sobre saúde pública. Em conclusão, este trabalho apresenta a aplicação prática de ciência social computacional no campo da saúde, através do desenvolvimento de um processo e ferramentas utilizadas para apoiar os analistas e melhorar a sua aplicação. / [en] This research thesis is aiming to exploit valuable social media, especially those available in online communities of social network sites, in order to perform social studies about healthcare issues. Based on a practical approach, a process was defined to conduct such studies. This process relied on tailored computational tools to provide support for specific tasks such as contente retrieval, selection, and analysis. Two tools that stand out are presented because of their utility and the complexity of the process in which their development was based on. The first tool, for the benefit of online community analysis, is the Community Association Map, a process developed to support experts in understanding users’ interests based on their associations within their communities. Our second tool (TorchSR) aims to aid analysts in the selection of discussions from online forums to be manually analyzed by (qualitative) research techniques (e.g. content and discourse analysis). This task, which was defined as solving the content selection problem, was tackled with a tool based on unsupervised machine learning techniques, such as hierarchical clustering. An exploratory study case shows that TorchSR helps analysts in dealing with the problem. The proposed process was employed in two studies about relevant healthcare issues (i.e. hepatitis C and drug abuse) which resulted in interesting findings in the field of public health. In conclusion, this thesis presents a practical application of computational social science to the field of health, through development of a process and tools used to support analysts and improve its application.