A Tense and Aspect Calculus

Diana SantosCOLING 1992

TL;DR

This paper introduces a formal calculus for representing tense and aspect in natural language using first-order logic. By developing a rich ontology of verb aspects and a two-step translation process, it enables robust reasoning about temporal relations, bridging the gap between linguistic parsing and logical inference. This work is crucial for building more sophisticated natural language understanding systems.

What Problem Does This Paper Address?

Understanding the temporal nuances of natural language is a significant challenge in computational linguistics. Tense (when an event occurs relative to the speaker's timeline) and aspect (the internal temporal structure of an event, e.g., whether it's ongoing, completed, or instantaneous) are fundamental to meaning. Existing approaches often struggle to formally capture the complex interplay between lexical verb meanings and grammatical markers that convey these temporal properties. This paper aims to provide a unified framework that can represent and reason about these phenomena, moving beyond simple temporal ordering to encompass the rich semantic information conveyed by aspect.

Methods

The authors propose a two-step translation process into first-order predicate calculus.

  1. Aspectual Classes: The lexicon is populated with "aspectual classes" that describe the temporal patterns of verbs. These are not just simple states or events but a more nuanced ontology, distinguishing between basic patterns and complex combinations.
  2. Grammatical Mechanisms: Grammatical elements (tenses, aspectualizers, adverbs, etc.) are translated into implication rules. These rules operate on the aspectual classes, allowing for generalizations and the expression of temporal meaning derived from syntax, not just lexicon.

Key features of their method include:

  • Integrated Parsing and Reasoning: The semantic representation and inference are not separated, allowing for immediate reasoning from parsed text.
  • Non-monotonic Reasoning: The system allows for defeasible information, meaning assumptions can be made in the absence of contradictory evidence, reflecting how humans often process language.
  • Rich Aspectual Ontology: They distinguish between various aspectual classes, including states, activities, accomplishments, and achievements, and also introduce "acquisitions" and "series" to capture more complex verb behaviors.
  • Contrastive Analysis: The framework is motivated by comparing English and Portuguese, highlighting the importance of cross-linguistic data for theory evaluation.

Datasets

Not discussed in the paper. The paper focuses on developing a formal calculus and illustrating it with examples, rather than empirical evaluation on large datasets.

Key Insights and Ideas

  • Aspect is More Than Just Verbs: The paper argues that grammatical mechanisms (syntax, tense) can introduce significant aspectual properties not present in the lexical meaning of verbs alone. This challenges approaches that solely focus on lexical aspect.
  • Vagueness as Logical OR: Linguistic vagueness, where a word or phrase can have multiple interpretations, is elegantly modeled using the logical OR operator (∨). This allows for multiple aspectual properties to be associated with a single lexical item, with context selecting the intended meaning.
  • Redundancy as Assertional Reinforcement: Similar information brought in by different linguistic devices is handled naturally. Instead of conflicting or requiring complex ordering, redundant information simply reinforces the assertion, leading to an elegant treatment of redundancy.
  • Unified Treatment of Temporal Phenomena: The calculus provides a unified way to handle diverse temporal phenomena, from simple past tenses to complex aspectual constructions like the progressive and perfect tenses, and even adverbials like "lately" and "just."

Key Results

The paper presents a formal calculus and demonstrates its application through numerous examples, showing how it can derive the meanings of various temporal constructions in English and Portuguese. For instance, it shows how "I have run" implies completion, and how the Portuguese "Perfeito" can translate English present perfects. It also provides formal definitions for aspectual classes and grammatical mechanisms, illustrating their use with examples like the progressive and simple present tenses in both languages. While no quantitative results are presented, the qualitative success lies in the formalization and the ability to derive nuanced temporal meanings.

Limitations and Future Work

  • Lack of Implementation and Testing: The authors acknowledge that the investigation is in its early stages, with no implementation or testing results available yet.
  • Syntax-Semantics Interface: While the paper notes the importance of syntax in disambiguating interpretations, a precise formulation for integrating syntactic information into the reasoning process is still needed.
  • Non-monotonic Reasoning Details: Further investigation is required to fully connect the proposed calculus with non-monotonic reasoning aspects.
  • Extensibility: The authors express confidence that the system is easy to extend by adding new implications or complex syntax imports, suggesting future work in expanding the ontology and handling more linguistic phenomena.

Summary updated

BibTeX
@inproceedings{santos1992tense,
  author = {Diana Santos},
  title = {{A Tense and Aspect Calculus}},
  booktitle = {International Conference on Computational Linguistics},
  year = {1992},
}
Figure