[Paper Review] A cookbook of translating English to Xapi

Ladislau Bölöni|arXiv (Cornell University)|Mar 31, 2013

Linguistic Variation and Morphology3 citations

TL;DR

This paper presents a systematic methodology for translating natural language narratives into Xapi, a simplified, semantics-driven 'pidgin' language designed to interface with the Xapagy cognitive architecture. By mapping English syntactic and semantic structures—especially in folk physics and folk psychology—into Xapi's instance-based, scene-oriented representation using tools like identity relations, change events, and coincidence links, the authors demonstrate that complex narratives can be consistently and logically encoded, enabling narrative reasoning through autobiographical memory and shadowing mechanisms.

ABSTRACT

The Xapagy cognitive architecture had been designed to perform narrative reasoning: to model and mimic the activities performed by humans when witnessing, reading, recalling, narrating and talking about stories. Xapagy communicates with the outside world using Xapi, a simplified, "pidgin" language which is strongly tied to the internal representation model (instances, scenes and verb instances) and reasoning techniques (shadows and headless shadows). While not fully a semantic equivalent of natural language, Xapi can represent a wide range of complex stories. We illustrate the representation technique used in Xapi through examples taken from folk physics, folk psychology as well as some more unusual literary examples. We argue that while the Xapi model represents a conceptual shift from the English representation, the mapping is logical and consistent, and a trained knowledge engineer can translate between English and Xapi at near-native speed.

Motivation & Objective

• To address the challenge of mapping complex natural language narratives into a formal, computationally tractable representation suitable for narrative reasoning.
• To develop a systematic, rule-based translation framework (a 'cookbook') for converting English into Xapi, a simplified language aligned with Xapagy’s internal cognitive structures.
• To enable knowledge engineers to perform near-native speed translation between English and Xapi by establishing consistent, logical mappings across semantic domains.
• To demonstrate the feasibility of representing folk physics and folk psychology concepts in Xapi using scene-based reasoning, identity chains, and attribute changes.
• To identify and address translation limitations, particularly for self-referential or stylistically nuanced language that lacks direct semantic equivalence in Xapi.

Proposed method

• Using Xapi’s simple syntax—subject-verb-object, subject-verb, subject-verb-adjective, and subject-communication-verb-scene-quote constructions—to represent narrative events as verb instances.
• Employing instances with attributes (e.g., 'is-a', 'owns', 'is-far-away') to model entities and their properties, with identity relations linking multiple instances to a single real-world referent.
• Applying change events (e.g., 'changes / mirage') to model evolving perceptions, such as in the Lawrence of Arabia scene, where a distant speck transforms into a recognized figure.
• Using coincidence links to associate observations with interpretations (e.g., 'utters text' linked to 'thus greets') without requiring simultaneous creation, enabling inference and summarization.
• Leveraging scenes and scene relations to model context shifts, such as in reported speech or hypothetical reasoning, where quotes are evaluated in alternate scenes.
• Utilizing shadows and headless shadows to support reasoning: shadows for memory recall, headless shadows for prediction or detection of missing narrative elements.

Experimental results

Research questions

• RQ1 How can complex narrative structures in English, especially those involving perception shifts and identity evolution, be consistently mapped into Xapi’s instance-based representation?
• RQ2 What representational tools in Xapi (e.g., identity chains, change events, coincidence links) are necessary and sufficient to model folk psychological and folk physical reasoning?
• RQ3 To what extent can narrative reasoning be achieved through Xapi-encoded autobiographical memory, and what are the limits of semantic fidelity in translation?
• RQ4 How can ritualized or conventionalized language (e.g., greetings) be interpreted in Xapi without deep syntactic parsing?
• RQ5 What are the inherent limitations of Xapi in translating self-referential or stylistically nuanced sentences from English?

Key findings

• The translation of narrative snippets—such as the Iliad combat exchange—into Xapi is straightforward when actions are sequential and entities are clearly defined.
• The use of identity chains allows the system to maintain continuity of perception across multiple interpretations of the same entity, as demonstrated in the Lawrence of Arabia scene where a speck evolves into a Beduin.
• Change events in Xapi effectively model perceptual evolution, such as the transformation from 'speck of dust' to 'mirage' to 'man', preserving narrative coherence.
• Coincidence links enable the system to associate observations (e.g., utterances) with interpretations (e.g., 'greets') even when the interpretations are generated later, supporting inference and summarization.
• The system can summarize complex exchanges—such as a mutual greeting—into a single summary verb instance (e.g., 'in-summary meet-and-greet-each-other'), enhancing narrative abstraction.
• Self-referential or stylistically nuanced sentences (e.g., 'This sentence has five words') cannot be semantically preserved in Xapi, highlighting inherent translation limitations.