NOMOS Ontology Documentation
This document contains information on the NOMOS
annotation ontologies used at CSLI. The
ontology is the NOMOS default core ontology. The other ontologies are extentions of this which provide schema definitions for specific corpora and datasets. See the NOMOS manual
for background information on how ontologies are used in NOMOS as annotation schemata.
ontology is designed as a base ontology which contains the most generic upper-level concepts for representing discourse data. The classes and slots in this ontology are generally inherited from all other ontologies. It also serves as the ontology for the NOMOS tool. All the classes required by NOMOS (not including plugins) are in this ontology.
NOMOS comes packaged with a core ontology (called
) which establishes essential assumptions about the data which are required for the system to work at the most basic level. Specifically, NOMOS must know about highly abstract things such as events, entities, and relations, as well as discourses (called meetings) and recordings. These are the core concepts on which NOMOS is built which are not configurable. Users may then embellish this ontology with terms that suit their annotation task (and may then customize NOMOS to handle those terms in special ways if desired). The following shows the core NOMOS class hierarchy.
Additionally, the following binary relations are defined for the classes (they are inherited from their superclasses as well):
NOMOS-alt-id : NOMOS-Thing -> String (an alternate ID for the object)
NOMOS-comment : NOMOS-Thing -> String (a comment about the object)
NOMOS-description : NOMOS-Thing -> String (the description of the object)
NOMOS-file : NOMOS-Record -> String (the file location containing the recording)
NOMOS-object : NOMOS-Relation -> NOMOS-Thing (the subject of a relation)
NOMOS-subject : NOMOS-Relation -> NOMOS-Thing (the object of a relation)
NOMOS-time-begins : NOMOS-Event -> Time-Instant (the start time of an event)
NOMOS-time-ends : NOMOS-Event -> Time-Instant (the end time of an event)
There are some other classes and slots defined in the
OPI, but these are only used by certain datasets we have created. They are not used by the core NOMOS architecture and comprehension is not necessary for basic use of NOMOS.
ontology is summarized in the following diagram:
Understanding the Diagrams
Part of the documentation of the ontologies is done using a diagram. In the diagrams, a graph is drawn which exemplifies how the individuals (objects) in the dataset are structurally related to one another. The following representations are used:
- a box represents an individual
- the box label represents the asserted class of the individual
- the box color represents the base superclass of the asserted class
- an arrow represents a property of individual from which the arrow originates
- an arrow label is the property name
For the box colors, the following applies:
- yellow =
- green =
- blue =
- dark grey =
NOMOS-Thing (unspecified or most generic type)
Additionally, rounded boxes
are used to depict literals. A large magenta box
signifies how datasets may be distribution into multiple models.
This ontology inherits from the
ontology since it is designed to represent the annotations in the MRDA corpus, a corpus of annotations on the ICSI corpus. The ontology includes concepts such as utterances, dialogue acts, adjacency pairs, and word-level segmentations of utterances.
This ontology describes hierarchical topic segmentations of meetings as well as shallow annotation of action item decisions.