Skip to content

Inference Module

Jump to bottom
LuotongCheng edited this page Oct 3, 2024 · 2 revisions

Inference Module

Overview

The Inference Module within the GOLEM framework is designed to handle the complexities of observational data in narrative and fiction. Observations are formalized as statements consisting of a subject, predicate, and object. The module not only captures these statements but also models their provenance, focusing on how inferences are made. By documenting the methods and sources behind each statement, the module reflects the multiplicity of interpretations that can arise from the same source—acknowledging that different approaches may lead to different conclusions.

Similar to the approaches used by MiMoTextBase and Wikidata, which prioritize provenance and multi-perspectival data, the Inference Module captures multiple layers of scholarly analysis and interpretation. In MiMoText, statements from secondary literature (such as literary histories and scholarly articles) are enriched by linking them to metadata and critical interpretations (Schöch et al., 2022). This process is similar to how the Inference Module connects methods and sources to the statements within the GOLEM framework. Wikidata similarly handles knowledge representation by ensuring that each statement about an entity (e.g., an author or literary work) includes detailed provenance information and reflects the perspectives that shaped it.

By modeling the relationships between statements, methods, and sources, the Inference Module enhances both transparency and traceability. It clearly maps the path from source and method to final conclusions, helping users navigate the complexities of literary interpretation and research (Carroll, 2015).

Scope

The Inference Module models any GOLEM statement using the crm:E13_Attribute_Assignment class, which captures the attribution of properties to subjects. The subject of the statement is linked using crm:P140_assigned_attribute_to, and the object or value is linked using crm:P141_assigned. The module also includes the sources and methods used to generate the statement. It also allows one crm:E13_Attribute_Assignment to serve as the source or premise for another.

Ontology (diagram)

Classes

  • crm:E13_Attribute_Assignment: According to CIDOC CRM, this class represents the act of assigning a property to an object or asserting a relation between concepts (Bekiari et al., 2024). crm:E13_Attribute_Assignment allows for detailed modeling of attribution by specifying the subject and object of the attribution, capturing the type of property being attributed (crm:E55_Type), as well as linking statements to their sources and methods.

  • crm:E55_Type: Represents the kind of property being assigned in the inference. It is the type of the assigned attributes in crm:E13_Attribute_Assignment.

Properties

  • crm:P140_assigned_attribute_to: Links the crm:E13_Attribute_Assignment to the subject of the statement.

  • crm:P141_assigned: Links the crm:E13_Attribute_Assignment to the object or value being assigned in the statement-making process.

  • crm:P177_assigned_property_of_type: Indicates the type of property or relationship being asserted in the attribution process. In CRMsci, the class CRMsci:S9 Property Type is a subclass of crm:E55_Type. In GOLEM, this property type can be customized to fit the context of the inference, specifying categories such as "topic," "character," or other relevant attributes.

  • crm:P16_used_specific_object: Links to the source used for both methods.

  • crm:P32_used_general_technique: Links to the method applied for inference.

Example 1: Topic Inference (diagram)

In this example, we model the assignment of the topic "friendship" to the work Harry Potter and the Philosopher's Stone using two methods: human reading and topic modelling. Both methods lead to the same conclusion.

Statement

  • Subject: Harry Potter and the Philosopher's Stone
  • Object: "Friendship"
  • Property Type: "Topic" (indicates the type of assigned attribute)
  • Property Type: "crm:P129_is_about" (indicates the type of relationship between assigned attributes and the subject)

Methods

  1. Human Reading
  2. Topic Modelling

Source

  • Source: The text of Harry Potter and the Philosopher's Stone

Inference Process

  • crm:P140_assigned_attribute_to: Links the work to the crm:E13_Attribute_Assignment.
  • crm:P141_assigned: Links the topic "friendship" to the crm:E13_Attribute_Assignment.
  • crm:P177_assigned_property_of_type: Indicates the type of property.
  • crm:P16_used_specific_object: Links to the source (the text of Harry Potter and the Philosopher's Stone).
  • crm:P32_used_general_technique: Links to the techniques applied for inference (human reading or topic modelling).

Example 2: Relationship Inference in Harry Potter and the Deathly Hallows (diagram)

In this example, we model the inference of a romantic relationship between Ron Weasley and Hermione Granger, based on an event described in Harry Potter and the Deathly Hallows. The relationship is inferred from the event: "Running at Ron, she [Hermione] flung them around his neck and kissed him full on the mouth."

Sources

  • Source of event inference: Text of Harry Potter and the Deathly Hallows
  • Source of relationship inference: Event inference

Inference Process

This example involves two layers of inferences, where the relationship inference (relationship assignment) uses the event inference (event assignment) as its premise:

1. Event Assignment (Premise)

  • crm:E13_Attribute_Assignment: Ron and Hermione are assigned as participants of the event (the kiss), with the text serving as the direct source.
  • crm:P140_assigned_attribute_to: Links the attribute assignment to the event.
  • crm:P141_assigned: Links the attribute assignment to Ron and Hermione.
  • crm:P16_used_specific_object: The passage in Harry Potter and the Deathly Hallows is used as the source for the event assignment.

2. Relationship Assignment (Based on Event Assignment)

  • crm:E13_Attribute_Assignment: Infers the romantic relationship between Ron and Hermione, using the event assignment as its source.
  • The event assignment (Hermione kissing Ron) serves as the premise that justifies the conclusion that a romantic relationship exists between the two characters.
  • crm:P140_assigned_attribute_to: Links the attribute assignment to the relationship of "Romantic Love".
  • crm:P141_assigned: Links the attribute assignment to the characters.
  • crm:P16_used_specific_object: The event assignment (Hermione kissing Ron) is used as the source for this relationship assignment, linking the two layers of inference.

Advantages

By employing crm:E13_Attribute_Assignment, the module allows for the representation of attributions and argumentation without necessarily tying them to beliefs, which are mental states that raise issues concerning intersubjective access (Sanfilippo et al., 2024). The crm:E13_Attribute_Assignment, as described in the CIDOC CRM, models the action of making attributions, offering a more flexible, neutral representation of interpretational claims across various scholarly frameworks.

Extension Potentiality

Our inference module by utilizing E13 has the potential to combine MITE ontology as an extension. MITE provides a more detailed classification of observation types, including: (i) basic observation (BasicObs), (ii) argumentation observation (ArgumentObs), and (iii) source observation (SourceObs). While our model mainly focuses on basic observations, the MITE framework expands this by offering a richer structure for representing different types of scholarly observations, making it suitable for more complex scenarios involving debates.

Reference

Bekiari, C., Bruseker, G., Canning, E., Doerr, M., Michon, P., Ore, C.-E., Stead, S., & Velios, A. (2024, October). Conceptual reference model (CIDOC CRM), version 7.3 (Tech. rep.). CIDOC CRM-SIG.

Carroll, N. (2015). Interpretation. In The Routledge companion to philosophy of literature (pp. 302–312). Routledge.

Doerr, M., Kritsotaki, A., Rousakis, Y., Hiebel, G., & Theodoridou, M. (2023, October). CRMsci: The scientific observation model (Tech. rep.). CIDOC CRM-SIG.

Sanfilippo, E. M., & Ferrario, R. (2024). D3.1—Observations modeling: State of the art.

Schöch, C., Hinzmann, M., Röttgermann, J., Dietz, K., & Klee, A. (2022). Smart modelling for literary history. International Journal of Humanities and Arts Computing, 16(1), 78–93.

Clone this wiki locally