Defining the 'Agent Type' AAG Property

[mbrush]

Building on the proposal outlined in #10, this issue outlines a set of categories for describing the high level type of agent that generated the statement expressed in an edge.

Below are brief definitions of 9 categories (note their hierarchical organization). Additional detail and insight re: the specific use and relevance in Translator can be found in the document here.

1. Manual Agent: A human agent is responsible for generating the knowledge expressed in the Statement.
2. Automated Agent: An automated agent is responsible for generating the Statement. Human contribution ends with the definition / and coding of algorithms or analysis pipelines executed by the computational agent.
3. ----Data Analysis Pipeline: An automated agent programmed to execute statistical analyses on instance data generates Statements that merely report associations/correlations between variables in the data.
4. ----Computational Model: An automated agent generates Statements (typically predictions) about the domain of discourse based on rules/logic that are encoded directly in computational algorithms (e.g. heuristic models, supervised classifiers), or rules that are learned from patterns observed in data (e.g. ML models, unsupervised classifiers).
5. ----Text Mining Agent: An automated agent is a text-mining tool that uses NLP to recognize concepts and/or relationships in text and generate Statements about these concepts with formally encoded semantics.
6. Manual Validation of Automated Agent: A human agent reviews and validates/approves the veracity of a Statement that is initially made by an automated agent.
7. Unspecified: It cannot be determined from available information if the agent is manual or automated.

Note that these categories are complemented by the 'Knowledge Level/Type' categories in #11 We had previously considered pre-composing terms from the cross-product of (relevant) Knowledge Level and Agent Type categories (e.g. 'Manual Knowledge Assertion', 'Computational Model Prediction') - but for now are splitting these into separate properties/annotations.

As we recommended for Knowledge Level tags, we would advocate for Agent Type tags to be REQUIRED on all Edges.

Examples of their application to real Translator scenarios/use cases can be found in the document here.