compound-agent 1.8.0 → 2.0.0

package/docs/research/spec_design/natural_language_formal_semantics_abuguity_in_specifications.md

@@ -1,259 +0,0 @@
-# Natural Language, Formal Semantics, and Ambiguity in Specifications
-
-*25 February 2026*
-
-## Abstract
-
-This survey examines how natural language encodes, constrains, and systematically fails to encode precise meaning, with particular attention to ambiguity in technical and requirements specifications. It integrates philosophical theories of meaning (from Wittgenstein, Austin, and Grice) with formal semantics, pragmatics, and theories of vagueness in order to articulate the mechanisms through which linguistic expressions acquire determinate interpretive content in context. This perspective treats ambiguity in specifications primarily as a linguistic and semantic–pragmatic phenomenon rather than a purely formatting or process defect. [en.wikipedia](https://en.wikipedia.org/wiki/Philosophical_Investigations)
-
-The analysis identifies several major families of approaches: use‑based theories of meaning, truth‑conditional and model‑theoretic semantics, speech act theory, Gricean and formal pragmatics, context‑sensitive and dynamic frameworks (including indexicality, presupposition, and dynamic semantics), situation semantics, accounts of vagueness, and formal techniques of semantic underspecification. These approaches display contrasting assumptions about the primary unit of analysis (sentence, utterance, context update, situation, or information state), the locus of precision, and the division of labor between semantics and pragmatics, which in turn yield different characterizations of why specifications are both inescapably natural‑language based and structurally prone to underdetermination. [plato.stanford](https://plato.stanford.edu/entries/wittgenstein/)
-
-## 1. Introduction
-
-The problem addressed in this survey concerns how natural language encodes meaning with sufficient precision to support reasoning, coordination, and technical specification, despite its pervasive context dependence, ambiguity, and vagueness. This question matters because large‑scale institutions, legal systems, and software development processes routinely rely on natural‑language documents whose interpretive stability is crucial for safety, liability, and interoperability, yet whose content remains anchored in philosophical and formal theories of meaning that were not designed primarily for engineering use cases. The analysis therefore documents how different traditions in philosophy of language and formal semantics conceptualize meaning, context, and indeterminacy, and how those conceptualizations illuminate structural sources of ambiguity long before formatting or process choices arise. [plato.stanford](https://plato.stanford.edu/entries/montague-semantics/)
-
-The scope of the survey includes analytic philosophy of language from Frege and early Wittgenstein through later ordinary‑language philosophy, truth‑conditional and model‑theoretic semantics, speech act theory, Gricean pragmatics, formal pragmatics, dynamic and situation semantics, theories of vagueness, and formal models of semantic underspecification. It excludes detailed psycholinguistic experimentation, distributional semantics, and applied natural language processing, except where these directly instantiate or challenge the theoretical frameworks under discussion. The focus is likewise on natural languages themselves rather than on domain‑specific specification languages, although requirements engineering is used as an applied case where the limitations of natural language become particularly salient. [arxiv](http://arxiv.org/pdf/2407.02820.pdf)
-
-Several key definitions structure the discussion. In plain terms, the meaning of an expression is first a pairing between linguistic form and conditions under which its use counts as correct; conceptually, this pairing becomes **semantic value** within a model, often truth conditions or context‑change potential. The act of using expressions in particular situations introduces additional information, namely **utterance meaning**, which combines semantic value with speaker intentions, background assumptions, and inference, and which belongs to **pragmatics** in the technical sense. Ambiguity is taken as a case in which a single form is compatible with multiple distinct semantic values, while vagueness refers to systematic borderline cases without sharp boundaries, and underspecification designates intentionally partial semantic representations that defer disambiguation to later stages. [faculty.washington](https://faculty.washington.edu/smcohen/453/GriceLogicDisplay.pdf)
-
-## 2. Foundations
-
-Foundational work by Frege, Russell, and early Wittgenstein framed meaning in terms of logical form, reference, and truth conditions, establishing the background against which later theories either elaborated or rejected the representational picture. In this tradition, the central explanatory tool is a model that assigns denotations to expressions and evaluates sentences as true or false relative to structures, thereby making precision a matter of mapping linguistic form to logically regimented content. Wittgenstein’s later writings resist this approach by emphasizing language as a collection of rule‑governed activities (“language games”) whose meanings are inseparable from their uses within forms of life, thereby shifting attention from abstract representation to the situated practices through which expressions acquire function. [fr.scribd](https://fr.scribd.com/doc/159832648/Irene-Heim-and-Angelika-Kratzer-Semantics-in-Generative-Grammar)
-
-Truth‑conditional semantics, particularly in the Montagovian paradigm, formalizes the Fregean idea that complex meanings are composed functionally from parts, using tools from typed lambda‑calculus and model theory to derive semantic values compositionally from syntactic structure. This framework enables systematic treatment of quantification, intensionality, and higher‑order constructions, but presupposes that the key notion of meaning is a stable mapping from expressions and contexts to truth conditions in possible worlds. In parallel, speech act theory and Gricean pragmatics argue that many central features of communication concern what speakers do by speaking, and what they convey beyond literal content through implicature, so any account of precision must accommodate illocutionary force and cooperative reasoning. [en.wikipedia](https://en.wikipedia.org/wiki/How_to_Do_Things_with_Words)
-
-The semantics–pragmatics interface is therefore foundational, since it partitions labor between context‑independent encoding and context‑sensitive inference in generating interpretable content. Dynamic semantics reconceives semantic value not as static truth conditions but as instructions for updating information states, aligning meaning with context change and thus with the evolution of conversational common ground. Situation semantics instead evaluates meanings relative to partial situations rather than entire possible worlds, which aims to better model context dependence, perception reports, and attitude ascriptions by treating information flow and partiality as intrinsic to meaning. Theories of vagueness and underspecification further complicate this picture by showing that indeterminacy is often not an accidental defect, but a systematic feature of how natural language categories, thresholds, and scope relations are encoded. [plato.stanford](https://plato.stanford.edu/entries/dynamic-semantics/)
-
-## 3. Taxonomy of Approaches
-
-The following table provides a classification of major approaches discussed in §4, organized by their primary unit of analysis and their characteristic view of how meaning is encoded.
-
-| Approach | Primary unit of analysis | Core mechanism for meaning | Locus of context sensitivity | Typical treatment of ambiguity |
-|---------|--------------------------|-----------------------------|-------------------------------|--------------------------------|
-| Use‑based theories of meaning (Wittgensteinian) | Language games, practices | Rules of use within forms of life | Practices and rule‑following | Emergent from multiplicity of uses, often tolerated |
-| Truth‑conditional model‑theoretic semantics (Frege–Montague) | Sentences in models | Compositional mapping to truth conditions | Parameters (world, time, assignment) | Resolved by enriching syntax, logical form, or models |
-| Speech act theory | Speech acts (illocutionary force) | Conventional rules linking forms to acts | Social institutions, felicity conditions | Ambiguity partly in act type, partly in propositional content |
-| Gricean and formal pragmatics | Utterances in cooperative interaction | Conversational maxims and pragmatic inference | Assumptions about rational cooperation | Ambiguity reduced by implicature and scalar reasoning |
-| Context, indexicals, and presupposition | Indexical expressions, presupposition triggers | Character–content distinction, presuppositional constraints | Context parameters, common ground | Ambiguity traced to shifting parameters and accommodation |
-| Dynamic semantics and discourse representation | Information states, discourse contexts | Context‑change potentials, update functions | Update rules over common ground | Ambiguity modeled as branching updates or unresolved discourse referents |
-| Situation semantics | Situations and infons | Relations among discourse, resource, and described situations | Choice of situation and information constraints | Ambiguity as multiple supporting situations or infon structures |
-| Vagueness theories | Vague predicates and categories | Supervaluation, epistemicism, or related logics | Context‑dependent standards, epistemic limits | Ambiguity and borderline cases as structural vagueness |
-| Semantic underspecification frameworks | Underspecified semantic representations | Compact structures encoding multiple readings | Constraints resolved by additional information | Ambiguity maintained until constraints select specific readings |
-
-Each row in this taxonomy corresponds to a distinct subsection in §4, and every approach analyzed there appears as a leaf node here, satisfying the classification requirement.
-
-## 4. Analysis
-
-### 4.1 Use‑Based Theories of Meaning
-
-**Theory & mechanism.** Use‑based theories, exemplified by Wittgenstein’s later work, treat meaning as rooted in patterns of use within “language games” rather than in mental representations or referential links to objects. The causal channel runs from communal practices, rule‑following behaviors, and forms of life to the normative standards that make particular word uses correct or incorrect, with meaning emerging as a stabilization of these practices over time. On this view, attempts to assign fully precise meanings independently of use appear misguided, because expressions acquire significance only against backgrounds of techniques, activities, and shared expectations. Ambiguity is thus not simply a failure of encoding, but a reflection of the multiplicity of roles a single expression may play across different games, which may or may not be regimented in technical sublanguages. [philosopheasy](https://www.philosopheasy.com/p/deep-dive-ludwig-wittgensteins-language)
-
-**Literature evidence.** The *Philosophical Investigations* articulates the slogan that “the meaning of a word is its use in the language,” and systematically criticizes the Augustinian picture of language as primarily naming objects. Secondary work on language games and private language arguments documents how Wittgenstein’s approach shifts attention toward the social, rule‑governed aspects of language, emphasizing context, practice, and shared criteria. Recent scholarship further investigates the distinction between primary and secondary uses, arguing that even ostensibly descriptive vocabulary participates in a variety of expressive and normative functions. This literature collectively demonstrates that many ordinary and even technical terms resist reduction to single model‑theoretic denotations without erasing important use‑sensitive structure. [suslj.sljol](http://suslj.sljol.info/articles/10.4038/suslj.v12i1.7025/galley/5421/download/)
-
-**Implementations & benchmarks.** While use‑based theories are not typically implemented as computational systems, they have influenced discursive approaches in philosophy of science, social theory, and qualitative methodologies, where context and practice are treated as primary data. In semantics and pragmatics, use‑based insights motivate frameworks that encode norms and inferential roles, such as inferentialist semantics, as well as discourse‑oriented models that emphasize conversational history. Evaluative benchmarks here are largely qualitative, involving the capacity to account for meaning shifts across disciplinary contexts, metaphors, and evolving technical jargon, rather than quantitative prediction of experimental judgments. [pmc.ncbi.nlm.nih](https://pmc.ncbi.nlm.nih.gov/articles/PMC8450915/)
-
-**Strengths & limitations.** Use‑based theories illuminate how meaning depends on background practices, which clarifies why specifications written in the same language may be read differently across organizational subcultures that instantiate distinct language games. They capture the plasticity and context sensitivity of natural language meaning, but offer limited formal tools for deriving precise conditions for truth or for systematically resolving ambiguity beyond appeal to practice. This limitation becomes salient when one needs explicit criteria for correctness that can be encoded in verification tools or formal specifications, where purely use‑theoretic descriptions often require supplementation by model‑theoretic or proof‑theoretic machinery. Consequently, this approach appears most informative at a diagnostic and conceptual level, while leaving many computational and engineering questions underdetermined. [idsl1.phil-fak.uni-koeln](https://idsl1.phil-fak.uni-koeln.de/current-research-in-the-semantics-pragmatics-interface.html)
-
-### 4.2 Truth‑Conditional and Model‑Theoretic Semantics
-
-**Theory & mechanism.** Truth‑conditional semantics explains meaning by associating each sentence with conditions under which it would be true in a model, thereby framing understanding as knowledge of those conditions. Montague’s program extends this perspective to natural language, deploying typed lambda‑calculus and model theory to derive denotations compositionally from syntactic structure, thereby making precision depend on correctly specified types, domains, and interpretation functions. The causal mechanism is that syntactic combination operations trigger corresponding semantic function applications, guaranteeing that small structural differences induce distinct truth conditions, which promises fine‑grained control over ambiguity once underlying logical forms are fixed. [mbrenndoerfer](https://mbrenndoerfer.com/writing/montague-semantics-formal-compositional-natural-language-understanding)
-
-**Literature evidence.** The Stanford Encyclopedia entry on Montague semantics outlines the development of “English as a Formal Language” and “Universal Grammar,” emphasizing compositionality and the tight alignment between syntactic categories and semantic types. Heim and Kratzer’s textbook *Semantics in Generative Grammar* presents a widely adopted implementation of this architecture within generative syntax, systematically articulating how phrase structure trees map to semantic types and denotations. Secondary expositions describe how generalized quantifier theory, intensional semantics, and the treatment of determiners, tense, and modality grew out of this framework. [plato.stanford](https://plato.stanford.edu/entries/montague-semantics/)
-
-**Implementations & benchmarks.** Montague‑style semantics has been implemented in numerous computational grammars and semantic parsers, including typed feature‑structure grammars and categorial grammars that attach lambda‑terms to syntactic rules. Evaluation typically proceeds via coverage of linguistic phenomena (such as quantifier scope, anaphora, and intensionality) and comparative adequacy against native speaker judgments or corpora, rather than raw predictive accuracy metrics. In formal verification and specification, similar logical tools support the translation of natural‑language like requirements into temporal logics or higher‑order logics, where the precision of the resulting system hinges on the choice of logical vocabulary and type structure. [aclanthology](https://aclanthology.org/P08-1026.pdf)
-
-**Strengths & limitations.** Model‑theoretic semantics provides powerful tools for precision, since once expressions are mapped to logical forms in a well‑specified language, their truth conditions become fully explicit, which strongly constrains interpretive variation. However, the mapping from natural‑language surface forms to logical forms itself can be highly ambiguous, particularly for quantifier scope, anaphora, and context‑dependent expressions, so precision is effectively displaced rather than fully achieved. Moreover, static truth‑conditional frameworks struggle with phenomena where meaning appears inherently dynamic or procedural, such as discourse referents, presupposition projection, and certain types of context‑sensitive updates, which motivates dynamic extensions discussed in §4.6. Finally, the focus on idealized, context‑abstracted sentences can obscure the ways in which institutional and pragmatic factors shape real‑world interpretation of specifications. [linguistically.substack](https://linguistically.substack.com/p/irene-heim-file-change-semantics)
-
-### 4.3 Speech Act Theory
-
-**Theory & mechanism.** Speech act theory, originating with Austin and developed by Searle and others, distinguishes between locutionary acts (producing meaningful utterances), illocutionary acts (performing actions like promising or ordering), and perlocutionary effects (producing consequences like persuading). The central claim is that many utterances do not merely describe states of affairs but constitute actions governed by social and institutional rules, so their meaning includes the act type performed under appropriate felicity conditions. The causal mechanism links conventional procedures, speaker intentions, and contextual factors to determine whether an utterance counts as, for example, a valid promise or an effective specification of requirements. [en.wikipedia](https://en.wikipedia.org/wiki/How_to_Do_Things_with_Words)
-
-**Literature evidence.** Austin’s *How to Do Things with Words* lays out the performative–constative distinction and subsequently replaces it with the tripartite classification of speech acts, stressing how formulaic utterances like “I hereby declare” enact institutional changes. Subsequent work in philosophy and linguistics extends and systematizes this framework, differentiating between direct and indirect speech acts, and analyzing institutional contexts such as legal discourse and contractual language. Discussions in legal theory and requirements engineering highlight how speech act classification affects the interpretation of obligations, permissions, and declarative specifications in formal and informal documents. [cs.uwaterloo](https://cs.uwaterloo.ca/~dberry/natural.language.html)
-
-**Implementations & benchmarks.** Computational treatments of speech acts appear in dialogue systems, where utterances are tagged with illocutionary force labels (questions, requests, confirmations) that guide system responses; these tags instantiate simplified versions of Austinian categories. In legal and requirements analysis, formalisms sometimes encode performative verbs and deontic operators to capture normative force, though benchmarks tend to be task specific, such as consistency checking or contract analysis, rather than standardized across domains. There are also formal systems of input–output logics and normative systems that partially operationalize speech act intuitions for reasoning under obligations and permissions. [coli.uni-saarland](https://www.coli.uni-saarland.de/courses/semantics-17/lectures/ST08-Dynamic.pdf)
-
-**Strengths & limitations.** Speech act theory exposes how specifications operate not only as descriptions but as performative acts that create commitments, allocate responsibilities, and trigger institutional procedures, thereby revealing a dimension of meaning invisible to pure truth‑conditional analysis. However, the theory typically relies on informal lists of felicity conditions and illocutionary types, which complicates attempts at formalization and cross‑linguistic generalization, especially when indirect speech acts or mixed performative–descriptive sentences are involved. Moreover, its focus on act types often leaves the internal semantic content of utterances to be treated by other frameworks, so speech act theory alone does not fully resolve ambiguity in what precisely is being required or guaranteed in a specification. [cambridge](https://www.cambridge.org/core/books/cambridge-handbook-of-formal-semantics/semanticspragmatics-interface/38D0793106DB47D5467E6C3434A3213F)
-
-### 4.4 Gricean and Formal Pragmatics
-
-**Theory & mechanism.** Gricean pragmatics explains aspects of meaning beyond literal content by positing a Cooperative Principle and associated conversational maxims (quantity, quality, relation, and manner), from which hearers infer **conversational implicatures** that enrich or adjust what is said. The causal channel runs from assumptions about rational, cooperative behavior and shared goals to inferences that bridge gaps in what is explicitly encoded, thereby generating additional communicated content without altering conventional truth conditions. Formal pragmatics extends this perspective using game theory, decision theory, and probabilistic models, which treat utterance production and interpretation as strategic reasoning under uncertainty about speaker goals, costs, and priors. [lawandlogic](https://lawandlogic.org/wp-content/uploads/2018/07/grice1975logic-and-conversation.pdf)
-
-**Literature evidence.** Grice’s “Logic and Conversation” provides the canonical statement of the Cooperative Principle and maxims, illustrating how utterances like “Some of the students passed” typically implicate that not all passed, given expectations about informativeness. The semantics–pragmatics interface literature, surveyed in the Cambridge Handbook of Formal Semantics, documents debates about whether scalar implicatures are derived pragmatically, encoded semantically, or even reflected in syntax, showing how different architectures relocate the boundary between encoded and inferred content. This body of work demonstrates that many intuitions about what is communicated depend sensitively on contextual expectations and processing considerations, which affects the apparent precision of natural‑language assertions. [faculty.washington](https://faculty.washington.edu/smcohen/453/GriceLogicDisplay.pdf)
-
-**Implementations & benchmarks.** Game‑theoretic and Bayesian models of pragmatics (including so‑called Rational Speech Act frameworks) have been implemented to predict human judgments about scalar implicatures, referential choices, and other pragmatic phenomena, often fitted to experimental data. Benchmarks typically compare model predictions to human responses in controlled tasks, measuring rates of implicature endorsement, reading times, or acceptability, and have revealed that pragmatic enrichment is both context‑sensitive and subject to processing constraints. In applied domains, Gricean reasoning informs guidelines for drafting clear contracts and specifications, although such applications remain informal and rely on human interpreters rather than automated systems. [idsl1.phil-fak.uni-koeln](https://idsl1.phil-fak.uni-koeln.de/current-research-in-the-semantics-pragmatics-interface.html)
-
-**Strengths & limitations.** Gricean and formal pragmatic models elegantly capture how speakers systematically convey more than they literally say, explaining patterns of under‑ and over‑specification, scalar terms, and indirectness, which are pervasive in natural‑language specifications. However, reliance on idealized rationality and shared knowledge assumptions can obscure power imbalances, misaligned incentives, or bounded rationality effects that often characterize real‑world organizational communication. Furthermore, the division between semantic content and pragmatic enrichment remains contested, with some empirical findings suggesting that certain inferences are so automatic that they may be encoded in semantics or syntax, thereby complicating attempts to localize ambiguity to a purely pragmatic component. [lawandlogic](https://lawandlogic.org/wp-content/uploads/2018/07/grice1975logic-and-conversation.pdf)
-
-### 4.5 Context, Indexicals, and Presupposition
-
-**Theory & mechanism.** The theory of indexicals and demonstratives, developed notably by Kaplan, distinguishes between **character** (a rule from contexts to contents) and **content** (the proposition expressed in a given context), thereby making context a parameter of semantic evaluation rather than merely a pragmatic backdrop. The causal channel here is that linguistic rules, together with contextual features like speaker, time, and place, determine referents for expressions such as “I,” “here,” and “now,” which then feed into truth‑conditional evaluation. In parallel, theories of presupposition and common ground, such as Stalnaker’s, treat certain background assumptions as shared information that constrains what can be felicitously asserted, and that can be updated dynamically through accommodation when mismatches arise. [plato.stanford](https://plato.stanford.edu/entries/indexicals/)
-
-**Literature evidence.** The Stanford Encyclopedia entry on indexicals summarizes Kaplan’s claim that indexicals are “devices of direct reference,” whose content in context is fixed directly by contextual parameters and demonstrations rather than by descriptive senses. Stalnaker’s work on common ground articulates a model in which presuppositions correspond to propositions already in the conversational common belief set, and where utterances function as proposals to update that common ground, subject to acceptance or rejection. Dynamic semantics overviews emphasize how presupposition triggers, such as “stop” or definite descriptions, impose definedness conditions on context updates, thereby connecting presupposition theory to context‑change potentials discussed in §4.6. [semantics.uchicago](https://semantics.uchicago.edu/kennedy/classes/f07/pragmatics/stalnaker02.pdf)
-
-**Implementations & benchmarks.** Formal treatments of indexicals are standard in compositional semantics, with lexical entries parameterized by context indices and implemented in many semantic parsers and computational grammars. Presupposition projection and accommodation have inspired algorithms in natural‑language generation and interpretation that attempt to track discourse status of referents and background assumptions, though robust, large‑scale systems remain challenging. Benchmarks involve capturing intuitions about projection in complex embeddings (such as conditionals and questions), and modeling how readers or listeners adjust common ground when confronted with presupposition failures. [semprag](http://semprag.org/article/download/sp.4.3/article-pdf)
-
-**Strengths & limitations.** Context‑sensitive semantics for indexicals and presuppositions clarifies how specifications that contain temporal, personal, or situational references can be interpreted relative to particular deployment or contract contexts, formalizing aspects of context that might otherwise remain implicit. However, specifying which contextual parameters are relevant, and how they should be set in multi‑party or multi‑jurisdictional environments, remains nontrivial, and formal frameworks often idealize context as a well‑defined tuple rather than a complex institutional arrangement. Presupposition accommodation, while capturing real conversational behavior, also reveals how readers can silently repair or reinterpret documents, potentially masking inconsistencies or gaps in specifications rather than forcing explicit resolution. [en.wikipedia](https://en.wikipedia.org/wiki/Situation_semantics)
-
-### 4.6 Dynamic Semantics and Discourse Representation
-
-**Theory & mechanism.** Dynamic semantics reconceives the meaning of an expression as its potential to change a context, represented as an information state or file, rather than as a static set of truth conditions. Heim’s File Change Semantics, for example, models discourse participants as maintaining “files” on discourse referents, with sentences serving as instructions to create, update, or access these files, thereby explaining phenomena like anaphora and indefinites in terms of context updates. The causal mechanism runs from syntactic structure through dynamic interpretation rules to transformations of the information state, making ambiguity correspond to divergent possible update paths rather than solely to multiple static denotations. [en.wikipedia](https://en.wikipedia.org/wiki/Dynamic_semantics)
-
-**Literature evidence.** The Stanford Encyclopedia article on dynamic semantics surveys frameworks in which context is treated as a parameter that grows over time, including File Change Semantics and Discourse Representation Theory, and emphasizes their shared “context‑change potential” perspective. Heim’s work introduces the file metaphor and formalizes novelty and familiarity conditions on indefinites and definites, as well as presupposition as definedness conditions on file operations. Subsequent research demonstrates how dynamic approaches explain presupposition projection, donkey anaphora, and hierarchical context updates, often simplifying or reorganizing phenomena that appear ad hoc in static models. [glossa-journal](https://www.glossa-journal.org/article/id/4848/download/pdf/)
-
-**Implementations & benchmarks.** Dynamic semantics has been instantiated in computational discourse representation frameworks that build structured representations (boxes or files) incrementally as text is processed, supporting tasks such as pronoun resolution and discourse coherence checking. Benchmarks typically involve coverage of classic test cases (donkey sentences, presupposition projection in embeddings) and performance on corpus‑based anaphora resolution or discourse parsing tasks, though direct quantitative evaluations relative to alternative theoretical frameworks remain limited. Formal comparisons indicate that certain presupposition projection patterns are captured more systematically by dynamic systems than by purely static, truth‑conditional approaches. [patrickdelliott](https://www.patrickdelliott.com/anaphora-seminar/handout1.pdf)
-
-**Strengths & limitations.** Dynamic semantics captures the intuition that understanding a specification involves not just evaluating isolated sentences, but updating a growing representation of system behavior, environment assumptions, and obligations, which aligns closely with engineering intuitions about requirements documents. It offers explicit mechanisms for modeling cross‑sentential dependencies, such as references to previously introduced entities or conditions, which are sources of both necessary cohesion and potential ambiguity. However, the added complexity of dynamic formalisms can make them harder to learn, implement, and scale, and there remains debate about how much of the observed behavior truly requires dynamic machinery rather than enriched static models plus pragmatics. [stokhof](https://stokhof.org/wp-content/uploads/2020/10/stokhof_adm.pdf)
-
-### 4.7 Situation Semantics
-
-**Theory & mechanism.** Situation semantics, developed by Barwise and Perry, evaluates meanings relative to **situations** (partial, concrete parts of the world) and **infons** (units of information), rather than full possible worlds, thereby encoding partiality and information flow at the foundational level. The meaning of an utterance is analyzed as a relation among a discourse situation, a resource or connective situation encoding background information, and a described situation that the utterance concerns, with constraints linking these components. The causal mechanism posits that linguistic expressions classify or constrain situations, and that understanding involves tracking how utterances license inferences about described situations given available resource situations. [john.jperry](http://john.jperry.net/cv/1997g.pdf)
-
-**Literature evidence.** The situation semantics entry outlines how the framework arose as an alternative to extensional model theory, particularly for analyzing perception and attitude reports where reference to partial situations seems more natural than to complete worlds. Perry’s historical survey explains the evolution of situation theory, the role of infons and constraints, and applications to perception, belief, and discourse, emphasizing how situations support or fail to support informational items. Later developments connect situation‑based approaches to more standard possible‑worlds semantics by treating situations as parts of worlds and incorporating situation variables into otherwise orthodox frameworks. [en.wikipedia](https://en.wikipedia.org/wiki/Situation_semantics)
-
-**Implementations & benchmarks.** Situation semantics has been used in certain computational representations of context, particularly in artificial intelligence systems that model local situations and information flow rather than full state spaces, although large‑scale empirical benchmarks are relatively sparse. In linguistic semantics, situation variables appear in analyses of tense, aspect, and focus, often implemented within standard model‑theoretic systems that enrich the domain with situations. Evaluation remains largely theoretical, based on the approach’s ability to handle problematic constructions such as naked infinitive perception reports and de re belief ascriptions more naturally than competing frameworks. [john.jperry](http://john.jperry.net/cv/1997g.pdf)
-
-**Strengths & limitations.** Situation semantics foregrounds partiality and information flow, which resonates with the structure of specifications that often describe local system behaviors, test scenarios, or failure modes rather than complete world states. By distinguishing discourse, resource, and described situations, it clarifies how background knowledge and contextual constraints shape the interpretation of particular descriptive clauses in complex documents. However, the original theory encountered foundational challenges, and later, more conservative variants often resemble enriched possible‑worlds semantics, raising questions about whether the additional complexity yields sufficient empirical or practical payoff. [cs.uwaterloo](https://cs.uwaterloo.ca/~dberry/natural.language.html)
-
-### 4.8 Vagueness and Semantic Indeterminacy
-
-**Theory & mechanism.** Vagueness concerns expressions with borderline cases and Sorites‑susceptible predicates, such as “heap,” “bald,” or “reliable,” where small changes do not seem to affect applicability but large accumulations do. Competing theories attribute vagueness either to **truth‑value gaps** (supervaluationism and related views) or to epistemic ignorance about sharp boundaries (epistemicism), with Williamson’s epistemicism arguing that vague predicates have precise extensions that are unknowable in principle. The causal mechanism differs accordingly: on gap theories, semantic rules intentionally leave some cases undefined, whereas on epistemic views, social and usage patterns fix sharp boundaries through complex micro‑facts about use, which remain beyond agents’ discriminative capacities. [robert-williams](https://robert-williams.org/2009/10/26/vagueness-survey-paper-iii-epistemicism/)
-
-**Literature evidence.** Kamp and Sassoon’s survey provides an overview of Sorites arguments, manifestations of vagueness across adjectives, nouns, verbs, and quantifiers, and various theoretical responses, including supervaluationism, degree semantics, and contextualism. Williamson’s *Vagueness* defends epistemicism, combining safety‑based epistemology with claims about sharp but unstable boundaries that depend sensitively on patterns of use. These works document that vagueness penetrates even apparently precise scientific and measurement vocabulary, suggesting that natural language rarely achieves absolute sharpness outside mathematical domains. [semanticsarchive](https://semanticsarchive.net/Archive/jhjNTc2Z/Vagueness.pdf)
-
-**Implementations & benchmarks.** Fuzzy logics, degree semantics, and probabilistic models have operationalized some treatments of vagueness, representing predicate application as graded or context‑dependent rather than binary, and these formalisms appear in control systems and approximate reasoning. Empirical benchmarks involve experimental studies of categorization, tolerance to small changes, and acceptability of Sorites premises and conclusions, although direct head‑to‑head testing of philosophical theories remains challenging. In requirements engineering, analyses often document how terms like “high availability,” “user‑friendly,” or “secure” function as vague predicates whose thresholds are rarely specified numerically, making vagueness a practical source of interpretive divergence. [taylorfrancis](https://www.taylorfrancis.com/books/mono/10.4324/9780203014264/vagueness-timothy-williamson)
-
-**Strengths & limitations.** Theories of vagueness demonstrate that a significant portion of interpretive indeterminacy arises not from ambiguity between discrete senses, but from systematic tolerance and context variability within single predicates, which complicates any attempt to regiment specifications purely by disambiguation. Gap theories align with the intuition that some borderline statements lack determinate truth values, while epistemicism preserves classical logic at the cost of postulating unknowable sharp boundaries, each choice introducing trade‑offs for formal reasoning about specifications that involve vague terms. Many engineering practices effectively adopt context‑sensitive or degree‑based interpretations without fully articulating underlying theories, leaving a gap between informal handling of vagueness and formally justified treatments. [robert-williams](https://robert-williams.org/2009/10/26/vagueness-survey-paper-iii-epistemicism/)
-
-### 4.9 Semantic Underspecification and Ambiguity Management
-
-**Theory & mechanism.** Semantic underspecification frameworks represent the meanings of ambiguous expressions using compact structures that encode multiple possible readings simultaneously, subject to constraints that restrict how subexpressions can be combined. The causal channel runs from syntactic or lexical ambiguity (such as quantifier scope or attachment) through construction of an underspecified semantic representation (USR), which captures shared structure across readings while deferring full resolution until additional information becomes available. Rather than forcing early disambiguation, these frameworks treat ambiguity as a structured space of possibilities, allowing subsequent processing stages to eliminate redundant or implausible interpretations. [let.uvt](https://let.uvt.nl/general/people/bunt/docs/book-b.pdf)
-
-**Literature evidence.** Bunt’s survey on semantic underspecification details various techniques, including Underspecified Discourse Representation Theory (UDRT) and constraint‑based formalisms using labels and dominance constraints, to represent multiple readings compactly. Koller and colleagues show how regular tree grammars can serve as a formalism for underspecified processing of scope ambiguities, including algorithms for eliminating equivalent readings and computing preferred readings efficiently. This literature reveals convergence on a “standard model” of scope underspecification, in which readings are treated as trees or formulas generated from a USR subject to structural constraints. [aclanthology](https://aclanthology.org/P08-1026.pdf)
-
-**Implementations & benchmarks.** Underspecification techniques have been implemented in computational semantics systems, including those accompanying broad‑coverage grammars that generate large numbers of potential scope readings, which are then compactly represented and pruned. Benchmarks include counts of possible readings for sentences with multiple quantifiers, coverage of known ambiguity patterns, and efficiency in enumerating or ranking interpretations, sometimes evaluated on semantically annotated treebanks. These systems can substantially reduce combinatorial explosion by avoiding full enumeration of readings until necessary, which is crucial for practical processing of complex specifications and legal documents. [let.uvt](https://let.uvt.nl/general/people/bunt/docs/book-b.pdf)
-
-**Strengths & limitations.** Underspecification directly addresses the problem that natural‑language specifications often remain legitimately ambiguous at early stages, yet still need to support partial reasoning and consistency checks, by providing formalisms that represent ambiguity without collapsing distinctions or forcing premature commitments. However, these frameworks typically focus on structural ambiguities of scope and attachment, and offer less direct treatment of vagueness, context dependence, or pragmatic enrichment, which are equally significant sources of interpretive variability. Additionally, selecting preferred readings still requires algorithms or heuristics grounded in broader semantic or pragmatic theories, so underspecification functions as an intermediate representation rather than a complete theory of meaning. [semanticsarchive](https://semanticsarchive.net/Archive/jhjNTc2Z/Vagueness.pdf)
-
-## 5. Comparative Synthesis
-
-The following table compares the approaches from §4 along several dimensions relevant to precision and ambiguity in natural language.
-
-| Approach | Conceptual orientation | Primary target phenomena | Formalization strength | Ambiguity handling style | Evidence base maturity |
-|----------|------------------------|--------------------------|------------------------|---------------------------|------------------------|
-| Use‑based meaning (§4.1) | Philosophical, practice‑oriented | Rule‑following, language games, meaning shifts | Low to moderate (informal, some systematization) | Accepts and describes multiplicity of uses; relies on social context | Mature philosophical tradition, limited formal testing [en.wikipedia](https://en.wikipedia.org/wiki/Philosophical_Investigations) |
-| Truth‑conditional semantics (§4.2) | Model‑theoretic, compositional | Truth conditions, quantification, intensionality | High (typed logics, models, textbooks) | Resolves ambiguity via logical form enrichment and type distinctions | Highly developed with extensive linguistic applications [plato.stanford](https://plato.stanford.edu/entries/montague-semantics/) |
-| Speech act theory (§4.3) | Normative‑institutional | Illocutionary force, performatives | Moderate (taxonomies, some logics) | Differentiates act types; content ambiguity often externalized | Established but formal extensions uneven [en.wikipedia](https://en.wikipedia.org/wiki/How_to_Do_Things_with_Words) |
-| Gricean / formal pragmatics (§4.4) | Inferential, game‑theoretic | Implicature, scalar terms, indirectness | Moderate to high (probabilistic and game models) | Reduces ambiguity via cooperative reasoning and implicature | Active cross‑disciplinary research with experimental support [faculty.washington](https://faculty.washington.edu/smcohen/453/GriceLogicDisplay.pdf) |
-| Context / indexicals / presupposition (§4.5) | Parameterized semantic | Indexicals, presuppositions, common ground | High (context indices, presupposition logics) | Attributes ambiguity to varying contexts, models accommodation | Well‑developed with ongoing debates [plato.stanford](https://plato.stanford.edu/entries/indexicals/) |
-| Dynamic semantics (§4.6) | Procedural, update‑based | Anaphora, presupposition projection, discourse | High (update logics, DRT, FCS) | Represents ambiguity as alternative update paths or unresolved files | Mature theoretical field, selective empirical tests [semprag](http://semprag.org/article/download/sp.4.3/article-pdf) |
-| Situation semantics (§4.7) | Information‑theoretic, partial | Perception reports, attitudes, partial information | Moderate (situation theory, infons) | Treats ambiguity as multiple supporting situations or infons | Historically significant, less mainstream today [en.wikipedia](https://en.wikipedia.org/wiki/Situation_semantics) |
-| Vagueness theories (§4.8) | Logical–epistemic | Borderline cases, Sorites, graded predicates | High (supervaluation, epistemicism, fuzzy logics) | Attributes indeterminacy to gaps, degrees, or ignorance | Extensive philosophical and some empirical work [semanticsarchive](https://semanticsarchive.net/Archive/jhjNTc2Z/Vagueness.pdf) |
-| Underspecification (§4.9) | Algorithmic, representational | Scope, attachment, structural ambiguity | High (constraint formalisms, RTGs) | Encodes ambiguity compactly, defers resolution | Technically mature in limited domains [aclanthology](https://aclanthology.org/P08-1026.pdf) |
-
-This comparison reveals several non‑obvious trade‑offs. First, high formalization strength does not entail comprehensive coverage of all ambiguity sources, since truth‑conditional and underspecification frameworks excel at structural ambiguities while leaving vagueness and pragmatic enrichment largely external. Second, approaches that foreground social practice or institutional context, such as use‑based theories and speech act theory, illuminate aspects of meaning crucial for specifications (like commitments and authority) yet remain less amenable to direct formal encoding, which complicates their integration into automated reasoning about requirements. Third, dynamic and context‑sensitive frameworks occupy an intermediate space, offering sophisticated formal tools that better match discourse and update phenomena, but at the cost of increased complexity and ongoing debates about how to partition semantic and pragmatic responsibilities. [philosopheasy](https://www.philosopheasy.com/p/deep-dive-ludwig-wittgensteins-language)
-
-## 6. Open Problems & Gaps
-
-1. **Integration of semantic and pragmatic theories into a unified specification‑oriented framework.** Existing theories specialize in different aspects of meaning, such as truth conditions, speech acts, or dynamic updates, yet there is no widely accepted architecture that combines model‑theoretic precision, speech‑act level commitments, and Gricean reasoning in a way that directly supports the drafting and automated analysis of real‑world specifications. Resolving this gap would require a principled account of the semantics–pragmatics interface calibrated to institutional and engineering needs. [plato.stanford](https://plato.stanford.edu/entries/dynamic-semantics/)
-
-2. **Systematic treatment of vagueness in safety‑critical and legal specifications.** Philosophical theories of vagueness provide multiple incompatible accounts of borderline cases, but there is limited work on how these theories translate into operational norms for documents where ambiguity tolerance is low, such as medical device standards or financial regulations. Addressing this problem would involve linking logics of vagueness with domain‑specific risk assessments and interpretive conventions. [taylorfrancis](https://www.taylorfrancis.com/books/mono/10.4324/9780203014264/vagueness-timothy-williamson)
-
-3. **Empirical evaluation of dynamic and situation semantics in applied settings.** Dynamic and situation‑based frameworks claim advantages for modeling discourse and partial information, yet their benefits relative to enriched static models and practical annotation schemes have not been systematically evaluated on large corpora of specifications, contracts, or technical standards. Progress here would require constructing annotated datasets and comparative tools that implement competing theoretical architectures. [coli.uni-saarland](https://www.coli.uni-saarland.de/courses/semantics-17/lectures/ST08-Dynamic.pdf)
-
-4. **Formal modeling of institutional and organizational language games.** Use‑based theories highlight that meaning depends on practices and forms of life, particularly in specialized domains like law, finance, or engineering, but such practices are rarely modeled formally beyond rough typologies. Clarifying how domain‑specific rules of use interact with general semantic and pragmatic mechanisms remains an open area with implications for cross‑organizational interoperability and standardization. [pmc.ncbi.nlm.nih](https://pmc.ncbi.nlm.nih.gov/articles/PMC8450915/)
-
-5. **Scalable representations of underspecification across semantic dimensions.** Current underspecification formalisms mainly target structural ambiguities in quantifier scope and attachment, while leaving contextual, pragmatic, and vague aspects outside the representational core. Developing frameworks that can encode and manage underspecification across these additional dimensions, without incurring intractable complexity, constitutes a significant open challenge. [aclanthology](https://aclanthology.org/P08-1026.pdf)
-
-6. **Bridging theoretical semantics with distributional and neural models.** Recent work on contextualized embeddings demonstrates that large language models encode context‑sensitive meaning in high‑dimensional spaces, but the relationship between these representations and classical notions of truth conditions, context updates, or infons remains poorly understood. Clarifying this relationship is important for assessing whether such models can support specification‑level precision or whether their representational indeterminacy fundamentally limits their role. [arxiv](https://arxiv.org/pdf/2209.00154.pdf)
-
-## 7. Conclusion
-
-The landscape surveyed here documents a rich array of approaches to how natural language encodes and fails to encode precise meaning, ranging from use‑based and speech act theories that foreground social practice, to model‑theoretic, dynamic, and situation‑based frameworks that formalize different aspects of semantic content and context. Each family of approaches isolates specific mechanisms by which ambiguity, vagueness, and context dependence arise, whether through multiplicity of language games, underdetermined logical forms, graded predicates, or evolving information states, and provides tools for analyzing these mechanisms at varying levels of abstraction. [plato.stanford](https://plato.stanford.edu/entries/wittgenstein/)
-
-A recurrent structural trade‑off concerns the balance between precision and coverage. Highly regimented truth‑conditional and underspecification frameworks achieve formal clarity for carefully delimited phenomena but often presuppose idealized contexts and interpretations, whereas use‑based and pragmatic theories capture lived interpretive practices at the cost of reduced formal tractability. Dynamic and context‑sensitive approaches attempt intermediate solutions that align better with discourse and specification practices, yet they introduce additional complexity and unresolved questions about the division of labor between semantics and pragmatics. Across these perspectives, ambiguity in specifications appears less as a mere formatting problem than as a manifestation of deep properties of natural language, whose systematic study in philosophy of language and formal semantics provides both explanatory resources and ongoing challenges for any attempt to engineer linguistic precision. [semprag](http://semprag.org/article/download/sp.4.3/article-pdf)
-
-## References
-
-Austin, J. L., 1962, *How to Do Things with Words*, Clarendon Press / Harvard University Press, overview at Wikipedia, https://en.wikipedia.org/wiki/How_to_Do_Things_with_Words. [en.wikipedia](https://en.wikipedia.org/wiki/How_to_Do_Things_with_Words)
-
-Barwise, J., and J. Perry, 1983, *Situations and Attitudes*, MIT Press; summarized in “Situation semantics,” *Wikipedia*, https://en.wikipedia.org/wiki/Situation_semantics. [en.wikipedia](https://en.wikipedia.org/wiki/Situation_semantics)
-
-Biletzki, A., and A. Matar, 2002, “Ludwig Wittgenstein,” *Stanford Encyclopedia of Philosophy*, https://plato.stanford.edu/entries/wittgenstein/. [plato.stanford](https://plato.stanford.edu/entries/wittgenstein/)
-
-Bunt, H., 1999, “Semantic Underspecification: Which Technique for What Purpose?,” Tilburg University preprint, https://let.uvt.nl/general/people/bunt/docs/book-b.pdf. [let.uvt](https://let.uvt.nl/general/people/bunt/docs/book-b.pdf)
-
-Grice, H. P., 1975, “Logic and Conversation,” in *Syntax and Semantics 3: Speech Acts*, Academic Press; PDF via lawandlogic.org, https://lawandlogic.org/wp-content/uploads/2018/07/grice1975logic-and-conversation.pdf. [lawandlogic](https://lawandlogic.org/wp-content/uploads/2018/07/grice1975logic-and-conversation.pdf)
-
-Heim, I., and A. Kratzer, 1998, *Semantics in Generative Grammar*, Blackwell, sample pages via Scribd, https://fr.scribd.com/doc/159832648/Irene-Heim-and-Angelika-Kratzer-Semantics-in-Generative-Grammar. [fr.scribd](https://fr.scribd.com/doc/159832648/Irene-Heim-and-Angelika-Kratzer-Semantics-in-Generative-Grammar)
-
-Heim, I., 1982, “The Semantics of Definite and Indefinite Noun Phrases,” Ph.D. thesis, University of Massachusetts Amherst; discussed in Patrick D. Elliott, “File change semantics: context and composition,” handout, https://www.patrickdelliott.com/anaphora-seminar/handout1.pdf. [patrickdelliott](https://www.patrickdelliott.com/anaphora-seminar/handout1.pdf)
-
-Heim, I., 1983, “On the Projection Problem for Presuppositions,” in *Proceedings of WCCFL 2*; dynamic extensions described in Philippe Schlenker, 2011, “Explaining presupposition projection with dynamic semantics,” *Semantics & Pragmatics* 4(3), http://semprag.org/article/download/sp.4.3/article-pdf. [semprag](http://semprag.org/article/download/sp.4.3/article-pdf)
-
-Kamp, H., and G. W. Sassoon, 2016, “Vagueness,” preprint in Semantics Archive, https://semanticsarchive.net/Archive/jhjNTc2Z/Vagueness.pdf. [semanticsarchive](https://semanticsarchive.net/Archive/jhjNTc2Z/Vagueness.pdf)
-
-Kaplan, D., 1989, “Demonstratives,” in J. Almog et al. (eds.), *Themes from Kaplan*, Oxford University Press; overview in David Braun, 2001, “Indexicals,” *Stanford Encyclopedia of Philosophy*, https://plato.stanford.edu/entries/indexicals/. [plato.stanford](https://plato.stanford.edu/entries/indexicals/)
-
-Koller, A., S. Thater, and J. Niehren, 2008, “Regular Tree Grammars as a Formalism for Scope Underspecification,” *Proceedings of ACL‑08: HLT*, https://aclanthology.org/P08-1026.pdf. [aclanthology](https://aclanthology.org/P08-1026.pdf)
-
-Montague, R., 1970, “Universal Grammar,” reprinted in R. Thomason (ed.), 1974, *Formal Philosophy*, Yale University Press; surveyed in“Montague semantics,” *Stanford Encyclopedia of Philosophy*, https://plato.stanford.edu/entries/montague-semantics/. [plato.stanford](https://plato.stanford.edu/entries/montague-semantics/)
-
-Perry, J., 1997, “History of Situation Semantics,” preprint, http://john.jperry.net/cv/1997g.pdf. [john.jperry](http://john.jperry.net/cv/1997g.pdf)
-
-Roberts, C., 2016, “The Semantics–Pragmatics Interface,” in M. Aloni and P. Dekker (eds.), *The Cambridge Handbook of Formal Semantics*, Cambridge University Press, chapter summary at https://www.cambridge.org/core/books/cambridge-handbook-of-formal-semantics/semanticspragmatics-interface/38D0793106DB47D5467E6C. [cambridge](https://www.cambridge.org/core/books/cambridge-handbook-of-formal-semantics/semanticspragmatics-interface/38D0793106DB47D5467E6C3434A3213F)
-
-Schlenker, P., 2011, “Explaining Presupposition Projection with Dynamic Semantics,” *Semantics & Pragmatics* 4(3), http://semprag.org/article/download/sp.4.3/article-pdf. [semprag](http://semprag.org/article/download/sp.4.3/article-pdf)
-
-Stalnaker, R., 2002, “Common Ground,” *Linguistics and Philosophy* 25(5–6), 701–721, PDF via semantics.uchicago.edu, https://semantics.uchicago.edu/kennedy/classes/f07/pragmatics/stalnaker02.pdf. [semantics.uchicago](https://semantics.uchicago.edu/kennedy/classes/f07/pragmatics/stalnaker02.pdf)
-
-“Dynamic Semantics,” *Stanford Encyclopedia of Philosophy*, 2010, https://plato.stanford.edu/entries/dynamic-semantics/. [plato.stanford](https://plato.stanford.edu/entries/dynamic-semantics/)
-
-“Dynamic semantics,” *Wikipedia*, https://en.wikipedia.org/wiki/Dynamic_semantics. [en.wikipedia](https://en.wikipedia.org/wiki/Dynamic_semantics)
-
-“Philosophical Investigations,” *Wikipedia*, https://en.wikipedia.org/wiki/Philosophical_Investigations. [en.wikipedia](https://en.wikipedia.org/wiki/Philosophical_Investigations)
-
-Williams, J. R. G., 2009, “Vagueness survey paper III (epistemicism),” blog post, https://robert-williams.org/2009/10/26/vagueness-survey-paper-iii-epistemicism/. [robert-williams](https://robert-williams.org/2009/10/26/vagueness-survey-paper-iii-epistemicism/)
-
-Williamson, T., 1994, *Vagueness*, Routledge; eBook information at Taylor & Francis, https://www.taylorfrancis.com/books/mono/10.4324/9780203014264/vagueness-timothy-williamson. [taylorfrancis](https://www.taylorfrancis.com/books/mono/10.4324/9780203014264/vagueness-timothy-williamson)
-
-Wittgenstein, L., 1953, *Philosophical Investigations*, Blackwell; discussed in “Meaning is use: Wittgenstein on the limits of language,” blog post, https://philosophyforchange.wordpress.com/2014/03/11/meaning-is-use-wittgenstein-on-the-limits-of-language/ and in L. K. C. Cheung, 2014, “Meaning, Use and Ostensive Definition in Wittgenstein's *Philosophical Investigations*,” *Philosophical Investigations*, https://onlinelibrary.wiley.com/doi/abs/10.1111/phin.12063. [philosophyforchange.wordpress](https://philosophyforchange.wordpress.com/2014/03/11/meaning-is-use-wittgenstein-on-the-limits-of-language/)
-
-Wittgenstein secondary discussions: “A study on Ludwig Wittgenstein’s concept of language games and the private language argument,” *Sabaragamuwa University Journal* 12(1), 2014, http://suslj.sljol.info/articles/10.4038/suslj.v12i1.7025/galley/5421/download/; and S. B. Biar, 2020, “Wittgenstein’s Concept of Language Games,” *Borneo International Journal of Languages and Education*, http://biarjournal.com/index.php/biolae/article/download/280/314. [biarjournal](http://biarjournal.com/index.php/biolae/article/download/280/314)
-
-Heim, I., 2025, “Irene Heim: File Change Semantics and Dynamic Meaning,” blog essay, https://linguistically.substack.com/p/irene-heim-file-change-semantics. [linguistically.substack](https://linguistically.substack.com/p/irene-heim-file-change-semantics)
-
-Elliott, P. D., 2017, “Semantic Theory week 8 – Dynamic Semantics,” lecture slides, Saarland University, https://www.coli.uni-saarland.de/courses/semantics-17/lectures/ST08-Dynamic.pdf. [coli.uni-saarland](https://www.coli.uni-saarland.de/courses/semantics-17/lectures/ST08-Dynamic.pdf)
-
-Stokhof, M., 1992, “Arguing About Dynamic Meaning,” preprint, https://stokhof.org/wp-content/uploads/2020/10/stokhof_adm.pdf. [stokhof](https://stokhof.org/wp-content/uploads/2020/10/stokhof_adm.pdf)
-
-Horn, L., and K. Turner (eds.), 2018, *Current Research in the Semantics/Pragmatics Interface* series overview, University of Cologne, https://idsl1.phil-fak.uni-koeln.de/current-research-in-the-semantics-pragmatics-interface.html. [idsl1.phil-fak.uni-koeln](https://idsl1.phil-fak.uni-koeln.de/current-research-in-the-semantics-pragmatics-interface.html)
-
-Berry, D. M., “Natural Language in Requirements Engineering,” University of Waterloo, https://cs.uwaterloo.ca/~dberry/natural.language.html. [cs.uwaterloo](https://cs.uwaterloo.ca/~dberry/natural.language.html)
-
-M. Brenndoerfer, 2025, “Montague Semantics – The Formal Foundation of Compositional Natural Language Understanding,” blog article, https://mbrenndoerfer.com/writing/montague-semantics-formal-compositional-natural-language-understanding. [mbrenndoerfer](https://mbrenndoerfer.com/writing/montague-semantics-formal-compositional-natural-language-understanding)
-
-“Semantic Theory: Dynamic Semantics,” lecture slides, University of Saarland, https://www.coli.uni-saarland.de/courses/semantics-17/lectures/ST08-Dynamic.pdf. [coli.uni-saarland](https://www.coli.uni-saarland.de/courses/semantics-17/lectures/ST08-Dynamic.pdf)
-
-(Additional web sources cited in the text provide context and secondary discussion but are not repeated here for brevity.)  
-
-## Practitioner Resources
-
-- **Stanford Encyclopedia of Philosophy (SEP) – Core Entries**: Central articles on Wittgenstein, Montague semantics, indexicals, dynamic semantics, and related topics, providing authoritative overviews and bibliographies, https://plato.stanford.edu. [plato.stanford](https://plato.stanford.edu/entries/indexicals/)
-
-- **Heim & Kratzer, *Semantics in Generative Grammar***: Standard textbook implementing compositional, model‑theoretic semantics with links to syntax, widely used for formal semantic training in linguistics and logic, sample at https://fr.scribd.com/doc/159832648/Irene-Heim-and-Angelika-Kratzer-Semantics-in-Generative-Grammar. [fr.scribd](https://fr.scribd.com/doc/159832648/Irene-Heim-and-Angelika-Kratzer-Semantics-in-Generative-Grammar)
-
-- **Cambridge Handbook of Formal Semantics**: Collection of survey chapters on formal semantics, including the semantics–pragmatics interface and computational aspects, suitable for advanced overviews, summary at https://www.cambridge.org/core/books/cambridge-handbook-of-formal-semantics. [cambridge](https://www.cambridge.org/core/books/cambridge-handbook-of-formal-semantics/semanticspragmatics-interface/38D0793106DB47D5467E6C3434A3213F)
-
-- **Semantics Archive**: Repository of preprints in formal semantics and pragmatics, including work on vagueness and dynamic semantics, accessible at https://semanticsarchive.net. [semanticsarchive](https://semanticsarchive.net/Archive/jhjNTc2Z/Vagueness.pdf)
-
-- **Semantics & Pragmatics (Journal)**: Open‑access journal publishing research on dynamic semantics, presupposition, and related topics, including Schlenker’s paper on presupposition projection, http://semprag.org. [semprag](http://semprag.org/article/download/sp.4.3/article-pdf)
-
-- **Patrick D. Elliott’s Dynamic Semantics Handouts**: Lecture handouts on File Change Semantics and context‑change potentials, useful as concise technical introductions, https://www.patrickdelliott.com/anaphora-seminar/handout1.pdf. [patrickdelliott](https://www.patrickdelliott.com/anaphora-seminar/handout1.pdf)
-
-- **Regular Tree Grammars for Underspecification (Koller et al.)**: ACL 2008 paper and associated tools that demonstrate underspecified semantic representations and algorithms for handling scope ambiguity, https://aclanthology.org/P08-1026.pdf. [aclanthology](https://aclanthology.org/P08-1026.pdf)
-
-- **Bunt’s Survey on Semantic Underspecification**: Overview of underspecification techniques and their suitability for different tasks, helpful for selecting representational frameworks in computational semantics, https://let.uvt.nl/general/people/bunt/docs/book-b.pdf. [let.uvt](https://let.uvt.nl/general/people/bunt/docs/book-b.pdf)
-
-- **Berry, “Natural Language in Requirements Engineering”**: Discussion of natural language ambiguity and imprecision in software requirements, connecting formal language theory with engineering practice, https://cs.uwaterloo.ca/~dberry/natural.language.html. [cs.uwaterloo](https://cs.uwaterloo.ca/~dberry/natural.language.html)
-
-- **Dynamic Semantics Lecture Slides (Saarland University)**: Slide sets introducing context‑change potential and dynamic interpretation, suitable for technically oriented readers seeking formal detail, https://www.coli.uni-saarland.de/courses/semantics-17/lectures/ST08-Dynamic.pdf. [coli.uni-saarland](https://www.coli.uni-saarland.de/courses/semantics-17/lectures/ST08-Dynamic.pdf)