In order to provide optimal access to the data we have elicited, the Afranaph Database was designed to permit online search and manipulation of the data stored in the extensive AQ responses. The database permits detailed comparisons of fine-grained distinctions across all of the languages we have data for.

The software for this database server was created by Floris van Vugt (programmer) and Alexis Dimitriadis, for the Berlin-Utrecht Reciprocals Survey [LINK?]. The system is designed to house information about languages in a flexible and highly configurable fashion, which has allowed it to be adapted to the needs of several other linguistic database projects.

The database houses three major kinds of “entities”: Languages,  anaphoric markers, and sentences. A marker or sentence is always particular to one language; a language may have any number of anaphoric markers or (of course) sentences. You can browse the database or use the search form to find languages, anaphoric markers or sentences of interest.

Each of these entities is described by a considerable number of properties: Identifying properties such as name or language family, the text, gloss and translation of sentences, and of course the linguistic analysis and classification of the markers and sentences. When viewing a page, attributes for which there is no data are not displayed; thus some pages contain more properties than others. Anaphoric marker properties can include a link to an example sentence, while sentences are typically linked to the anaphoric marker they contain (if they contain one; for various reasons it can be useful to include non-anaphoric sentences in the database). The database also allows comments to be added to any property; again, these are only shown if a comment was actually entered.

The system is unusual, among databases of this type, in that most questions are not hard-coded into the definition of the database: they are maintained as a separate set of tables that list the questions applying to each entity, and the type of answer(s) that will be accepted. This is what allows the database to be reconfigured for different research projects without much programming. There are no facilities for summarizing or statistics at this time. You can find a presentation of the design principles of the software here:  http://languagelink.let.uu.nl/burs/docs/burs-design.pdf

Our database includes all of the sentences that are recorded in our AQ responses and sometimes a few more, it includes most of the commentary in the AQ responses, and it includes a great deal of categorization of the data based on criteria we at the Afranaph project thought would be interesting to researchers interested in anaphora and perhaps other syntactic and semantic domains. For example, sentences can be sorted for the kinds of anaphoric markers that are in them, the nature of the verbs in the clause, the complexity of the clause, the nature of the antecedents for pronouns or anaphors in the sentence, the relation of the main verb in the clause to any subordinate clause and so forth. Properties of anaphoric markers are sorted into types, as are languages, by historical relations and word order type, and so forth, but about all these properties and how they can be accessed, we leave for the next section.

What is important to mention here is that we have designed the database to provide the optimal access to the data we have collected, and so that interested linguists could pursue research agendas that use our database as a testing ground for theories and, more specifically, the analytic generalizations that theories generate. We have tried to remain as theory neutral as our own generative grammar-oriented prejudices permit, and we recognize that our neutrality should be suspect. In fact, this is one reason that we provide both the static AQ responses as well as the database, so that property attributions that some may find suspect, or errors that we make even in applying our own system, can be checked, discovered, refuted or improved by those who are willing to delve closely into the data we have collected. Hopefully, we have managed to remain consistent enough in our property attributions so that search and comparison functions that the database permits will provide pointed and accurate information. We are counting on our users to tell us when we fail.

At present, we only enter data into the database when an AQ response is complete or close enough to complete and we such that we do not expect the data to change radically on the basis of further elicitations in the near future.  Whenever there is a significant change in what is in the database, that is, when the data of a language is updated or changed in some way, or if a new language is added to the database, we will announce the change in our News and Events page. If you have used the database to establish a generalization and you are returning to it after a period of time to explore a related question, we recommend checking the News and Events page announcements to see what has been updated since your last visit.

* For optimal viewing of the African Anaphora Database, we recommend using Firefox as your browser.  Please click here to download Firefox.