About

Kyle Rawlins is an Associate Professor in the Cognitive Science Department at Johns Hopkins University, with a background that includes a PhD in Linguistics from UC Santa Cruz and a BA/BS in Linguistics and Computer Science from UMass Amherst. His research interests encompass semantics, pragmatics, computational modeling, philosophy of language, computational semantics, experimental linguistics, crowdsourcing, and syntax. Rawlins's empirical work focuses on interrogatives/questions, question-answer discourse, clause-embedding, conditionals, adverbs and modification, lexical semantics, and definiteness. He has contributed to numerous publications in these areas, exploring topics such as anaphoric variability, indefinite readings, asymmetries in definite descriptions, evidential meaning, and multi-sentence argument linking, among others. Rawlins regularly teaches courses on theoretical and computational semantics, mathematical foundations of cognitive science, and language & advertising, including graduate seminars on semantics and pragmatics. He has also served as a faculty advisor for the JHU Go Club and engages in hobbies such as music creation and contributing to open source projects like the game DCSS.

Research topics

• Natural Language Processing
• Computer Science
• Linguistics
• Philosophy
• Artificial Intelligence
• Epistemology
• Geology
• Programming language
• History
• Library science
• Art history

Selected publications

• Analyzing naturally-sourced Questions Under Discussion

  Experiments in Linguistic Meaning · 2025-01-24

  articleOpen accessSenior author

  The Question Under Discussion (QUD) framework of discourse has been a highly influential theoretical device in many accounts of various pragmatic phenomena, yet there has been comparatively little work assessing the extent to which the QUD can be reliably inferred from naturalistic contexts. In this paper, we focus primarily on measuring the variability across individuals in QUD inference, while also verifying other related, commonly held assumptions about QUD theory. To this end, we collect QUDs from many theoretically naive subjects tasked with processing a radio interview utterance by utterance. We consider various analyses designed to address the problem of measuring question similarity. Overall, we find that there exists moderate variability among subjects, consistent with possibly the insufficiency of context in determining QUD, or possibly also the simultaneous coexistence of multiple valid QUDs. To more adequately tease apart these possibilities, we also propose additional analyses for addressing the issue of question identity.

  PublisherOA PDFDOI

• Asking (non-)canonical questions

  Proceedings from Semantics and Linguistic Theory · 2025-02-12

  articleOpen access1st authorCorresponding

  Questions are classically taken to be requests for information, while acknowledging a wide variety of ‘non-canonical’ questions that do not have this function (e.g. rhetorical questions, exam questions, etc). A standard current approach is to take the request-for-information view as an analytical starting point and then weaken it for the counterexamples. This paper proposes an alternative view of questioning that encompasses many of these counterexamples directly: to ask a question is to open coordination on the public resolution of an issue. This coordination-centric view, I argue, accounts for much of the landscape of both canonical and non-canonical questions, while generalizing much previous work related to Questions Under Discussion in discourse.

  PublisherOA PDFDOI

• Deriving the evidence asymmetry in positive polar questions

  Proceedings from Semantics and Linguistic Theory · 2024-01-19

  articleOpen access1st authorCorresponding

  This paper explores a famous puzzle about English positive polar questions introduced by Buring and Gunlogson 2000: while in many contexts they seem to indicate nothing whatsoever about what the speaker takes for granted or thinks likely, in contexts that provide evidence against the content proposition of the question, they are infelicitous. This pattern, which I term the "evidence asymmetry", has been particularly troubling for standard accounts of polar questions that treat the positive and negative answers on par with each other. However, given that polar questions are felicitous in neutral contexts, it doesn't have an easy solution: polar questions in general don't seem to place constraints on evidence or context. I propose that polar questions have a fairly weak presupposition requiring just the content alternative to be possible (but say nothing about its negation), and (building on Trinh 2014) that this together with Maximize Presupposition-based reasoning about competitor questions (specifically"or not" alternative questions) can derive the evidence asymmetry. This account does not require the covert evidential marker of Trinh 2014, and essentially proposes that the evidence asymmetry follows from norms for English polar questions.

  PublisherOA PDFDOI

• Perspectival biscuits

  Proceedings from Semantics and Linguistic Theory · 2024-01-19

  articleOpen accessSenior author

  This paper describes a novel class of biscuit conditional, the 'perspectival biscuit', which arises when an if-clause containing a generic pronoun (e.g., generic you) is used to shift perspective for the interpretation of a perspective-sensitive item in the consequent: e.g., fixing the directionality of behind in "If you're at the door, the cat is behind the desk." This sentence is like a biscuit conditional in that it entails a fully-specified, propositionally stable consequent describing the spatial configuration of cat and desk, but this reading vanishes in favor of a conditional dependence reading when the antecedent contains any non-generic DP, a prediction that is not straightforwardly accounted for by existing theories of biscuit conditionals. An analysis is given demonstrating that biscuithood for perspectival biscuits arises due to generic quantification exclusively over individuals, not worlds.

  PublisherOA PDFDOI

• Zero and Few-shot Semantic Parsing with Ambiguous Inputs

  arXiv (Cornell University) · 2023-06-01 · 2 citations

  preprintOpen access

  Despite the frequent challenges posed by ambiguity when representing meaning via natural language, it is often ignored or deliberately removed in tasks mapping language to formally-designed representations, which generally assume a one-to-one mapping between linguistic and formal representations. We attempt to address this shortcoming by introducing AmP, a framework, dataset, and challenge for translating ambiguous natural language to formal representations like logic and code. We define templates and generate data for five well-documented linguistic ambiguities. Using AmP, we investigate how several few-shot text-to-code systems handle ambiguity, introducing three new metrics. We find that large pre-trained models perform poorly at capturing the distribution of possible meanings without deliberate instruction. However, models are able to capture the distribution well when ambiguity is attested in their inputs. These results motivate a call for including ambiguity explicitly in datasets and promote considering the distribution of possible outputs when evaluating systems. Data and code: https://github.com/esteng/ambiguous_parsing

  PublisherOA PDFDOI

• Semantic Incorporation in English Singular Indefinites

  Journal of Semantics · 2023-04-13

  articleSenior author

  Abstract In this paper, we introduce a class of exceptionally narrow-scoping singular indefinites in English (e.g., “Sam drove a car for several years before switching to a truck”), which pattern more closely with what have been termed “weak definites” in the literature (e.g., Poesio, 1994; Carlson et al., 2006) than with regular indefinites. While the existence of such exceptional “weak” indefinites has been previously anticipated by Klein et al. (2013), the category is difficult to distinguish from simple narrow-scoped singular indefinites in most contexts. Here, we argue that there is one environment where weak singular indefinites can be distinctively identified: namely, when they appear with for-adverbials. We sketch a concrete implementation of a semantic incorporation-based account for such nominals, bringing them analytically in line with incorporation analyses of weak definites, building closely on the ideas in Dayal (2011). We further briefly discuss how the proposed analysis adjudicates between two competing analyses for for-adverbials, one which assumes that for encodes a universal quantifier (e.g., Deo & Piñango, 2011) and another which takes for to be non-quantificational (e.g., Champollion, 2013), in favor of the latter view. We close by considering some remaining issues surrounding semantically incorporated DPs in English: specifically, how weak (in) definites relate to other nominals that receive covarying interpretations across contexts—such as bare plurals on the one hand (which we do not take to be semantically incorporated) and bare singulars on the other.

  PublisherDOI

• LOME: Large Ontology Multilingual Extraction

  arXiv (Cornell University) · 2021-01-28

  preprintOpen access

  We present LOME, a system for performing multilingual information extraction. Given a text document as input, our core system identifies spans of textual entity and event mentions with a FrameNet (Baker et al., 1998) parser. It subsequently performs coreference resolution, fine-grained entity typing, and temporal relation prediction between events. By doing so, the system constructs an event and entity focused knowledge graph. We can further apply third-party modules for other types of annotation, like relation extraction. Our (multilingual) first-party modules either outperform or are competitive with the (monolingual) state-of-the-art. We achieve this through the use of multilingual encoders like XLM-R (Conneau et al., 2020) and leveraging multilingual training data. LOME is available as a Docker container on Docker Hub. In addition, a lightweight version of the system is accessible as a web demo.

  PublisherOA PDFDOI

• Asymmetries between uniqueness and familiarity in the semantics of definite descriptions

  Proceedings from Semantics and Linguistic Theory · 2021-03-02 · 1 citations

  articleOpen access

  In over a century of research into the English definite article "the", two main theoretical factors have been identified as relevant to its meaning: namely, (i) uniqueness and (ii) familiarity. The identification of these factors has led to an extensive debate in semantics about which of them is more fundamental to the meaning of "the". In this paper, we contribute to this debate by introducing novel data obtained through two controlled psycholinguistic experiments. We manipulated uniqueness and familiarity of potential referents, examining how these factors affect the comprehension and production of English definite descriptions. The behavioral results reveal an asymmetry between these two factors, with familiarity being a weaker cue than uniqueness – a pattern that is unexpected under any existing theory of definiteness. We close with a discussion of possible extensions to existing theories in light of this result, as well as avenues for future work.

  PublisherDOI

• About 'What About': Topicality at the Semantics-Pragmatics Interface

  2021 · 1 citations

  Senior authorCorresponding
  • Computer Science
  • Natural Language Processing
  • Computer Science
  Publisher

• LOME: Large Ontology Multilingual Extraction

  2021 · 24 citations

  • Computer Science
  • Natural Language Processing
  • Artificial Intelligence

  Patrick Xia, Guanghui Qin, Siddharth Vashishtha, Yunmo Chen, Tongfei Chen, Chandler May, Craig Harman, Kyle Rawlins, Aaron Steven White, Benjamin Van Durme. Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: System Demonstrations. 2021.

  PublisherOA PDFDOI

Frequent coauthors

• Benjamin Van Durme

  21 shared

• Aaron Steven White

  University of Rochester

  19 shared

• Rachel Rudinger

  10 shared

• Patrick Xia

  6 shared

• Elias Stengel-Eskin

  5 shared

• Siddharth Vashishtha

  5 shared

• Sadhwi Srinivas

  William & Mary

  5 shared

• Craig Harman

  4 shared

Labs

• Semantics LabPI