Melita interface

Melita’s main control panel is composed of two main areas:

  1. The ontology (left) representing the annotations that can be
    inserted; annotations are associated to concepts and relations. A specific colour is associated to each node in the ontology (e.g. speaker is depicted in yellow).
  2. The document to be annotated (centre-right). Selecting the
    portion of text with the mouse and then clicking on the node in
    the ontology insert annotations. Inserted annotations are shown by turning the background of the annotated text portion to the colour associated to the node in the hierarchy (e.g. the
    background of the portion of text representing a speaker
    becomes yellow).
Melita interface


Intrusiveness is handled by Melita in several ways.

A button on the main interface is used to stop the system from intruding by blocking any suggestions from the learning algorithm. If the suggestion button is on, then the user will receive suggestions from the system, but not all suggestions are displayed to the user since some suggestions (especially at the start of the session) may have very low precision or recall.

To allow the user to restrict the suggestions accepted, a component having two movable knobs is displayed for each concept.
The knobs can be moved and their position is equivalent to a balance between precision and recall also called the f-measure (where 0% indicates low precision and low recall, while 100% equals high precision and high recall). The lower knob is the suggestion knob while the higher knob is the certainty knob.

Rules whose f-measure is below the suggestion knob are not displayed while rules above the suggestion knob but below the certainty knob are displayed as suggestions (in Melita suggestions are shown using a coloured border around the target concept and they must be validated by a user before they are accepted).

Rules above the certainty knob are certain to be correct and are displayed using a filled coloured square around the target concept.

intrusiveness interface


Effectiveness is achieved through the document sorting mechanism.

This approach dynamically sorts the documents after every annotation in order to find the document that best covers the unexplored areas of the domain3. Documents are rated according to the number of tags automatically found by the IE engine.

The document with the least number of tags is chosen for annotation because it is the document from which the learning algorithm can learn new rules if it is annotated.

Melita document rating


The project

Web Intelligence Group logo
Natural Language Processing Group logo

Valid XHTML 1.0!   >Valid XHTML 1.0!

Vitaveska Lanfranchi