Name Sentiment Analysis using Crowdsourcing Data
Source UN Women
Hyperlink to Source
Description In this study an existing crowdsourcing dataset is used to depict women’s political sentiments about pre-specified subjects. UNICEF Uganda’s Technology for Development (T4D) initiative has developed several crowdsourcing applications based on the RapidSMS platform, a flexible open-source framework for creating texting tools on mobile phones. Among the more notable successes is the U-Report application, a service that sends polls and alerts to its network of over 250,000 (and growing) “reporters”, mostly young people. Since its inception in 2011, well over two hundred polls have been sent out, and the application has received strong government and NGO support. Reporters can register their age and sex, and U-Report automatically geo-tags responses. The U-Report application is presently being expanded across Sub-Saharan Africa, including in Zambia, Burundi, the Democratic Republic of the Congo, Mozambique, South Sudan, and Zimbabwe. This project has the objective of utilizing this growing database to analyze women’s views on the wide range of development topics already polled, disaggregated by geographical area and age of respondent. In addition, new polls specific to issues of special concern to girls and women could be suggested to UNICEF country offices. One study found that use of extrinsic, intrinsic, and social incentives increases U-Report participation, a promising finding if greater response among particular sub-groups is desired. Although U-Report focuses on simple menu-based answers to polls, there is also the option to send short message responses to the solicited questions. Natural language processing techniques of the type already used to analyze social media feeds could be applied to this message dataset to obtain a more nuanced picture of women’s perspectives, for example to gauge intensity of sentiment. UNICEF and IBM Research are currently working on developing algorithms for this purpose, in addition to already having created protocols for auto-correction of text, keyword matching, and text classification that allow categorization of messages into particular development topics (e.g. water, health and nutrition, education, employment, etc.). In addition, because the responses are time and geo-tagged, the determinants of the U-report data – that is, what leads to variation in responses across space and through time – could be analyzed in combination with existing demographic and economic data (e.g., access to health and educational services, income and expenditure patterns, family size, etc.).
Long description
Lead Institution UN Women
Logo of lead Institution Logo institution
Partners UNICEF Uganda, IBM Research, other UNICEF country offices
Contact name Please login to see contact details
Contact email Please login to see contact details
Info link
Implementation link
Case study link
Video embed code or link to video page
Type Knowledge
Potential for reuse Low
Open source No
Open Source URL
Data Input Formats
Data Output Formats
Maintenance Requirements
Plug & Play No
Outputs compatible with indicators from Official Statistics No
Status Concept
Country use cases
Target Groups
Statistical Institute
Other Government Agencies
Civil Society Organizations
Sectors where the innovation can be applied
  • Gender
  • ,
  • Social Protection & Labour
  • Areas of technological innovation
  • Social Media Data
  • Areas of management practice innovations
    Crowdsourcing
    Use of Alternative Data Sources
    Areas of institutional process innovations
    Case Study URL with Public Sector Participation
    Case Study Description
    Picture describing Case Study
    Good practices displayed
    Created On 2014-04-12
    Updated On 2014-04-12
    Status Published
    Framework
    Build




    Comments