Universal Grammar

2021

Citations in the corpus

• ConTinTin: Continual Learning from Task Instructions (2022, background): “…learning process suggests at least two sources of alternative supervision: one is to accumulate knowledge from tasks learned in the past ( S 1 ) (Richard, 1970; Thrun and Mitchell, 1995; Chomsky, 2002); the other is to learn from natu-ral instructions ( S 2 ) describing a high-level story about…”
• Learn Continually, Generalize Rapidly: Lifelong Knowledge Accumulation for Few-shot Learning (2021, neutral): “This training scheme is in contrast with the way humans process natural language (Chomsky, 2002; Montague, 1970): humans”
• Inherent Disagreements in Human Textual Inferences (2019, background): “Entailment is arguably one of the most fundamental of language understanding tasks, with Montague himself calling entailment ‘‘the basic aim of semantics’’ (Montague, 1970).”
• A case for deep learning in semantics: Response to Pater (2019, background): “Both of the theories sketched in §2 are compositional in the technical sense originally defined by Montague (1970).”
• Compositional generalization through meta sequence-to-sequence learning (2019, background): “These abilities are due to systematic compositionality, or the algebraic capacity to understand and produce novel utterances by combining familiar primitives [5, 26].”
• UnNatural Language Inference (2020, background): “If grammatical sentences are required for sentential inference, as per a truth conditional approach (Montague, 1970), then permuted sentences should be meaningless.”
• Measuring Compositionality in Representation Learning (2019, background): “Techniques from this literature are specialized to the details of linguistic representations—particularly the algebraic structure of grammars (Montague, 1970).”
• Locality and compositionality in zero-shot learning (2019, background): “In what follows, we will consider a representation to be compositional if it can be expressed as a combination of simpler parts (Andreas, 2019; Montague, 1974).”
• The EOS Decision and Length Extrapolation (2020, background): “A core feature of the human language capacity is the ability to comprehend utterances of potentially unbounded length by understanding their constituents and how they combine (Frege, 1953; Chomsky, 1957; Montague, 1970).”
• Human few-shot learning of compositional instructions (2019, background): “This example illustrates systematic compositionality, the algebraic capacity to understand and produce an infinite number of utterances from known components (Chomsky, 1957; Montague, 1970; Fodor, 1975).”
• Compositional Generalization for Primitive Substitutions (2019, background): “Humans learn language in a flexible and efficient way by leveraging systematic compositionality , the algebraic capacity to understand and produce large amount of novel combinations from known components (Chomsky, 1957; Montague, 1970).”
• Compositional Semantic Parsing with Large Language Models (2022, background): “Compositionality is a key part of human intelligence as it allows us to understand and produce a potentially infinite number of novel combinations of known components (Chomsky, 1957; Montague, 1970; Lake et al., 2017).”
• AGQA: A Benchmark for Compositional Spatio-Temporal Reasoning (2021, background): “…indirect references of concepts, and to more compositional steps. events is reflected in the language people use to communicate with one another [8, 44], so tasks involving both vision and language comprehension, such as answering questions about visual input, can test models’ compositional…”
• Unified Visual-Semantic Embeddings: Bridging Vision and Language With Structured Meaning Representations (2019, background): “Researchers have designed various types of representations [5, 32] as well as different models [26, 50] for translating natural language sentences into structured representations.”
• How Much Do Language Models Copy From Their Training Data? Evaluating Linguistic Novelty in Text Generation Using RAVEN (2021, background): “Compositionality: Compositional generalization (combining familiar parts in novel ways) is crucial for processing both the syntax and semantics of natural language (Montague, 1970).”
• Linguistic generalization and compositionality in modern artificial neural networks (2019, background): “The focus is typically on semantic compositionality, the principle whereby the meaning of a linguistic expression is a function of the meaning of its components and the rules used to combine them [13,14].”
• Compositional Generalization and Natural Language Variation: Can a Semantic Parsing Approach Handle Both? (2020, background): “These formalisms directly embody the formal notion of the principle of compositionality as a homomorphism between the parts of the source and target structures, and operations over these parts (Montague, 1970; Janssen and Partee, 1997).”
• Compositional Generalization via Neural-Symbolic Stack Machines (2020, background): “Given a set of basic components and a few demonstrations of their combinations, a person could effectively capture the underlying compositional rules, and generalize the knowledge to novel combinations [9, 34, 27, 26].”