Observations Workshop 2007 Session 1

This is version 9. It is not the current version, and thus it cannot be edited.
[Back to current version]   [Restore this version]

Session 1: Observational data "in the wild"

Breakout-Group Products

• A consensus definition of "observation"

Background

It is likely that workshop participants will have different perspectives on what constitutes observational data. This session will consider the various definitions in use for the terms 'observation' and 'observational data' to ascertain the breadth of the concepts that need to be captured/represented in a formal model. The objective is to clarify the various issues and underlying assumptions surrounding these terms, while forming a working consensus definition for 'observation' that will be used throughout the workshop.

Discussion Topics

What is meant by 'observation' and 'observational data'? What are the different types of observational data that should be considered in a standard data model?

Possible issues:

• What kinds of data are collected and represented?
• What are the ways that data is collected (e.g., sensors, field, derived)?
• What are the formats (e.g., spreadsheets, text files, binary formats, etc.) used to store and exchange data?
• Are observations stand-alone entities (with all necessary context)?
• Are observations dependent on other observations

Session notes and results

SOME MORE SPECIFIC ISSUES TO DISCUSS

1) Can (or should) we derive a core set of terms that completely enable the generic description of "scientific observations"? General enough to enable description of any type of scientific observation (how does "scientific" constrain our definition of an observation, if at all?) Should we scope observation to only pertain to some specific types of interests? Is there a mechanism for specializing/refining descriptions of observation to make "the framework" useful for data integration?

Are there ways to "add" domain-specific extension ontologies to the core?

2) Can we establish synonymies and other mappings among terms in the various Observation Models presented so far? clarify subsumption (more/less general) and other (part_of, constituted_of) relationships among terms necessary "components" of an observation; necessary & sufficient?

3) Need unambiguous definitions of Keywords* (currently massive overloading)

Observation Entities concrete/theoretical; thing'ish vs process'ish Attributes/characteristics/features/traits Procedure/method Instrument Measurement Units Measurement standards Specimen/record/collection Time/duration Location--point, others Context/composition/complex types Transformations/identity Essential (necessary) features or components

4) Is the assertion of some specific entity type (rock, crow) best accomplished as a special form of measurement (attribute=NAME) Identifying vs non-identifying names presence/absence measurements-- are these (ontologically) different from say, measuring mass of a specimen? names-- are names a special characteristic associated with an entity (a "name" is classifying the entity as some thing) counts (is a property of set of individuals)-- does this create problems with the model?

5) Do "where" and "when" pertain globally and consistently to any type of observation of interest, or are these also types of observation

6) Is there ability within the Observation model, to inter-relate observations to "construct" composite or emergent entities, processes or phenomena?

7) How can ontology/model be applied? Must data be ingested into framework implementing the model, or can model(s) be applied to data (e.g., via annotation)

8) Is there possibility for multiple interpretations of a given observation? Is this ever necessary or desirable?

9) How indicate that parts of tuple are dependent/independent e.g. enriching context can be optional, but high dependency requires association of two observations (

10) How rich must be set of associations/relations among observations. How specialized can these get? (wolf fighting mountain lion)

11) Is it important to be able to differentiate individual instances vs. generic assertions in a data set ("a wolf fighting a mountain lion" vs "wolves fight mountain lions")