The Reference Engine Developer is an environment that supports (a) cross-linguistic research about reference phenomena, (b) corpus annotation and (c) the configuration and evaluation of reference resolution engines. All of this can be carried out by end users without the intervention of developers.
The Reference Engine Developer was originally created to support research and development about so-called "broad" referring expressions such as it, this, that, which can refer not only to noun phrases but also to entire spans of text.
The main screen of the interface is shown in the gallery.
The main interface is divided horizontally into 4 sections:
- In the top section, texts are input and various types of expressions (the referring expression in question, its actual sponsor, 3 candidate sponsors that are not the actual sponsor, and chains of reference) are highlighted in different colors.
- In the second section, ontological mappings (i.e., semantic interpretations) of certain text elements are provided.
- The third section contains an inventory of multiple-choice questions, divided into simpler tasks for students (e.g., indicating gender and number) and more complex tasks for superusers (e.g., indicating case roles and certain syntactic parsing information).
- The bottom section permits users to toggle between "layers" of text annotation: one layer compares the referring expression in question with its actual sponsor and the other layers compare it with 3 selected non-sponsors. Only 3 non-sponsors are selected (rather than potentially dozens) due to time constraints for annotation.
For details about this environment and the research results of the related research, see the related publications listed below.