Annotations Database
About the Annotations DB
The Annotations Database is a Javelin-II system component that
enables retrieval, navigation and organization of a space of
Documents based on semantic information.
The Annotations DB runs on a MySQL 5.0.2-alpha database server,
and clients connect to it through a Java API built on top of
MySQL Connector/J 3.1.7.
Annotations DB Feature List
- Data model in which all updates and inserts are perfomed via
stored procedures to enforce database consistency.
- Generalized annotation model of hierarchical Tags that can express
an arbitrarily complex nested annotation structure.
- Java API: Corpus model for viewing database contents as a
Collection of attributed Documents, which are Collections of Passages,
which are Collections of attributed Tags.
- Can iterate over Documents, Passages, and Tags.
- Access to and update of docno, doctype and datetime fields in a
Document.
- Update of Tag type.
- Access to Document, Passage and Tag Attributes.
- Adding a Document to a Corpus, a Passage to a Document, or a Tag
to a Passage.
- Comparison, equality and identicality measures for Tags.
- Deletion of Documents, Passages, and Tags by destroy method.
- Document loader to rapidly incorporate Documents into the database.
- Filters that select Documents, Passages or Tags by type or other
data, and those that check attributes of these database objects.
- Fully-parameterized PassagePane for viewing an annotated passage.
Annotations DB Feature Backlog
- Robustness for multiple client access (row_locking) but not
sacrificing performance
- Fix DocumentLoader's ugly iterface.
- Batch-mode Tag insertion / Standoff loader
- Corpus model objects that know when they are stale - the basic
problem is if fields of an object are updated, holders of other
references to that table row will not know about the change.
- Corpus model; Attributes should have a Map view.
- Can higher level Tags refer to children cross-Passage? How to
encode containment relationship (iterators)?
- Higher level Tags should be relieved of the requirement that they
reference some Span if they have children.
- FilterFactory.exactMatchTagTypeAttributeFilter(String tagType,
String attributeName, String attributeValue)
Next steps for the Annotations DB
- Fix the fact that even higher level tags must be grounded in the
text.
- Batch-mode Tag insertion / Standoff loader
Annotations DB Release History
Annotations DB Viewer
- Beta version 0.34 released on 08/30/05 [.zip archive]
- Beta version 0.32 was released on 08/01/05 [.zip archive]
This point release fixes a bug in how classifier scores were displayed in the document attribute panel.
- Beta version 0.31 was released on 7/28/05 [.zip archive]
This point release adds additional tabs to the document view which show the attributes assigned
to the document and each of its passages.
- Beta version 0.3 was released on 7/27/05 [.zip archive]
- Beta version 0.1 was released on 6/27/05 [.zip archive]
Annotations DB Links
Matthew Bilotti