what is potentially relevant in each document.”43 These issues would then have to be contained within the database design. It is therefore important to see if other data sources will share their data, and to determine how best to relate this information to the central violations database. This will facilitate a broader understanding of a range of structural and thematic considerations and allow an assessment of violations at a local, as well as regional and national context. The creation of a human rights violation relational database is a practical option on the basis of existing primary data sources in Zimbabwe, which can in turn be complemented by alternative data sources that would set out amongst other things a chronology of political, social, and economic events. This will provide a very rich base from which to analyse key patterns of violence. The process will, however, require a detailed and rigorous examination of existing data sources, which will in turn require the engagement of competent personnel and effective management to ensure the quality and consistency of data interpretation. What is required to create the database? The creation of an information management system requires the effective organisation and analyses of data. There are four main steps in an information management system: (a) (b) (c) (d) Collection of data Data processing – classification and coding Database representation Generating analytical reports Collection of data In most large-scale human rights data projects, the information is collected on the basis of a developed protocol, which in turn facilitates the consistency of the coding process. What is suggested in the current context is somewhat different and certainly more complex — in that we will be assessing multiple sources of data that already exist. This is likely to be considerably more time-consuming and will require more hands-on management and ‘quality control’ to ensure the data is interpreted consistently. Data processing – classification and coding “Data processing moves information from its partially structured, partly-coded state in a questionnaire (statement/narrative/case file) into a fully structured, fully-coded state in which it can be entered directly into a computer.”44 Data processing is often very time consuming and requires skilled staff to do it. Narrative statements will be coded according to controlled vocabularies – i.e. an agreed list of types of violation. It will be necessary to assess a sample of existing statements to draw up this agreed list, as well as assessing existing violation categories already utilised by some organisations. The level of detail of violations captured will depend on what has been recorded. It is important, however, that each category in the controlled vocabulary exclude the other categories 43 Who Did What to Whom? Planning and Implementing a Large Scale Human Rights Data Project Patrick Ball, American Association for the Advancement of Science, 1996, p. 31. 44 ibid., p. 32. 36