Nothing Special   »   [go: up one dir, main page]

WAT-SL 2.0 Example Project

This is a short readme to help you understanding how WAT-SL 2.0 works. All file paths in this document are giving as starting from the server directory, which is the directory in which you run the Java JAR file.

Annotation

The annotation interface is available at wat/annotate. Annotator credentials to access the annotation interface are defined in the wat.conf. Example annotator login: "Charlie Brown" (password "Peanuts").

Administration

The administration interface is available at wat/admin. The password to access the administration interface is defined in the wat.conf. For this example project, the login password is "a". The interface currently allows to reload the server and write results (see description in the interface).

Code

The code of WAT-SL 2.0 and all files of this example project is available at https://github.com/webis-de/wat

Setup

This page is read from /web/index.html. Replace it appropriately. To configure the server, you have to change the wat.conf and component settings You can also add curation tasks.

The wat.conf

The main configuration file of WAT-SL is the wat.conf in the server directory. The file is used to configure the project name, which is displayed on every page, an admin password, the components and the annotator accounts. The file uses a key = value format with one key-value pair per line. The wat.conf of this project contains comments that detail every setting.

Components

WAT-SL 2.0 uses the concept of annotation components to describe jobs in the annotation interface. The only two components in WAT-SL are a simple text box (used for comments) and segment labeling. Components are defined and named in the wat.conf, where a name is given to each component (here: "segment-labeling" and "comments-box"). Each component can be configured project wide (in a /project/<component-name>.conf) and for each task (in a /tasks/<task-name>/<component-name>.conf, which overrides settings in the project wide configuration file). The project wirde configuration files of this project contain comments that detail every setting.

Most importantly, the /project/<component-name>.conf defines the labels available to the annotators, but also defines a stylesheet that overrides the default segment-labeling look by assigning different colors to the labels. The segments are read from a /tasks/<task-name>/<component-name>.txt for each task. These files contain one segment per line, where paragraphs are separated by empty lines.

Curation

Curation is implemented by creating a new task and copying the annotation record files of the annotators. When the annotators are finished, you can (but don't need to) disallow them to perform changes by removing the task from their tasks list in the wat.conf. Then copy the task directory (usually just a cp -r tasks/<task-name> tasks/curation-<task-name> on Linux) and copy the annotation files from the task states directory into the new directory (cp -r task-states/<task-name>/* tasks/curation-<task-name>). After that, add the new task curation-<task-name> to the task list of the curator. Note that WAT does not distinguish between annotators and curators, so the curator account is defined as an annotator account in the wat.conf. Finally, reload the server. For an example, user "Charlie Brown" (password "Peanuts") has one curation task assigned to him.

Imprint