In this paper we propose a way to deal with natural language inference by implementing Modern Type Theoretical Semantics in the proof assistant Coq. The paper is a first attempt to deal with NLI and natural language reasoning in general by using the proof assistant technology. Valid NLIs are treated as theorems and as such the adequacy of our account is tested by trying to prove them. We use Luo’s Modern Type Theory with coercive subtyping (...) as the formal language into which we translate natural language semantics, and we further implement these semantics in the Coq proof assistant. It is shown that the use of a MTT with an adequate subtyping mechanism can give us a number of promising results as regards NLI. Specifically, it is shown that a number of inference cases, i.e. quantifiers, adjectives, conjoined noun phrases and temporal reference among other things can be successfully dealt with. It is then shown, that even though Coq is an interactive and not an automated theorem prover, automation of all of the test examples is possible by introducing user-defined automated tactics. Lastly, the paper offers a number of innovative approaches to NL phenomena like adjectives, collective predication, comparatives and factive verbs among other things, contributing in this respect to the theoretical study of formal semantics using MTTs. (shrink)

Transformer-based Pre-Trained Language Models currently dominate the field of Natural Language Inference (NLI). We first survey existing NLI datasets, and we systematize them according to the different kinds of logical inferences that are being distinguished. This shows two gaps in the current dataset landscape, which we propose to address with one dataset that has been developed in argumentative writing research as well as a new one building on syllogistic logic. Throughout, we also explore the promises of (...) ChatGPT. Our results show that our new datasets do pose a challenge to existing methods and models, including ChatGPT, and that tackling this challenge via fine-tuning yields only partly satisfactory results. (shrink)

Epistemic closure refers to the assumption that humans are able to recognize what entails or contradicts what they believe and know, or more accurately, that humans’ epistemic states are closed under logical inferences. Epistemic closure is part of a larger theory of mind ability, which is arguably crucial for downstream NLU tasks, such as inference, QA and conversation. In this project, we introduce a new automatically constructed natural language inference dataset that tests inferences related to epistemic (...) closure. We test and further fine tune the model RoBERTa-large-mnli on the new dataset, with limited positive results. (shrink)

We propose an approach to natural language inference based on a model of natural logic, which identifies valid inferences by their lexical and syntactic features, without full semantic interpretation. We greatly extend past work in natural logic, which has focused solely on semantic containment and monotonicity, to incorporate both semantic exclusion and implicativity. Our system decomposes an inference problem into a sequence of atomic edits linking premise to hypothesis; predicts a lexical entailment relation for (...) each edit using a statistical classifier; propagates these relations upward through a syntax tree according to semantic properties of intermediate nodes; and composes the resulting entailment relations across the edit sequence. We evaluate our system on the FraCaS test suite, and achieve a 27% reduction in error from previous work. We also show that hybridizing an existing RTE system with our natural logic system yields significant gains on the RTE3 test suite. (shrink)

The alignment problem—establishing links between corresponding phrases in two related sentences—is as important in natural language inference (NLI) as it is in machine translation (MT). But the tools and techniques of MT alignment do not readily transfer to NLI, where one cannot assume semantic equivalence, and for which large volumes of bitext are lacking. We present a new NLI aligner, the MANLI system, designed to address these challenges. It uses a phrase-based alignment representation, exploits external lexical resources, (...) and capitalizes on a new set of supervised training data. We compare the performance of MANLI to existing NLI and MT aligners on an NLI alignment task over the well-known Recognizing Textual Entailment data. We show that MANLI significantly outperforms existing aligners, achieving gains of 6.2% in F1 over a representative NLI aligner and 10.5% over GIZA++. (shrink)

How can computers distinguish the coherent from the unintelligible, recognize new information in a sentence, or draw inferences from a natural language passage? Computational semantics is an exciting new field that seeks answers to these questions, and this volume is the first textbook wholly devoted to this growing subdiscipline. The book explains the underlying theoretical issues and fundamental techniques for computing semantic representations for fragments of natural language. This volume will be an essential text for computer (...) scientists, linguists, and anyone interested in the development of computational semantics. (shrink)

The aim of this project was to explore the possibility of constructing a psychologic inference engine that might enhance introspective self-awareness by delivering inferences about a user based on what he said in interactive dialogues about his closest opposite-sex relation. To implement this aim, we developed a computer program (guru) with the capacity to simulate human conversation in colloquial natural language. The psychologic inferences offered represent the authors' simulations of their commonsense psychology responses to expected user-input expressions. (...) The heuristics of the natural language processor and its relation to output responses are described in enough detail for the operations of the implementation to be understood. Evaluation of this new cognitive agent presents, we hope, puzzles for artificial intelligence and cognitive science. (shrink)

We present a novel generative model for natural language tree structures in which semantic (lexical dependency) and syntactic (PCFG) structures are scored with separate models. This factorization provides conceptual simplicity, straightforward opportunities for separately improving the component models, and a level of performance comparable to similar, non-factored models. Most importantly, unlike other modern parsing models, the factored model admits an extremely effective A* parsing algorithm, which enables efficient, exact inference.

Traditionally, analytic philosophy has been affiliated with a formalist conception of inference which understands reasoning as a process that exploits syntactic properties of natural language according to a set of formal rules that are insensitive to conceptual content. This chapter discusses an alternative approach that takes semantic properties as the underlying forces driving rational inference. Building on Wilfird Sellars’ notion of material inference and analytic tools from cognitive linguistics, I will show how parts of the (...) inferential structure of natural language can be explained in terms of semantic relations between extra-logical concepts. In the end, I will outline a strategy for explicating some of these inference-types using Peter Gärdenfors’ theory of conceptual spaces. (shrink)

Natural language syntax yields an unbounded array of hierarchically structured expressions. We claim that these are used in the service of active inference in accord with the free-energy principle (FEP). While conceptual advances alongside modelling and simulation work have attempted to connect speech segmentation and linguistic communication with the FEP, we extend this program to the underlying computations responsible for generating syntactic objects. We argue that recently proposed principles of economy in language design—such as “minimal search” (...) criteria from theoretical syntax—adhere to the FEP. This affords a greater degree of explanatory power to the FEP—with respect to higher language functions—and offers linguistics a grounding in first principles with respect to computability. While we mostly focus on building new principled conceptual relations between syntax and the FEP, we also show through a sample of preliminary examples how both tree-geometric depth and a Kolmogorov complexity estimate (recruiting a Lempel–Ziv compression algorithm) can be used to accurately predict legal operations on syntactic workspaces, directly in line with formulations of variational free energy minimization. This is used to motivate a general principle of language design that we term Turing–Chomsky Compression (TCC). We use TCC to align concerns of linguists with the normative account of self-organization furnished by the FEP, by marshalling evidence from theoretical linguistics and psycholinguistics to ground core principles of efficient syntactic computation within active inference. (shrink)

This paper develops an inference system for natural language within the ‘Natural Logic’ paradigm as advocated by van Benthem, Sánchez and others. The system that we propose is based on the Lambek calculus and works directly on the Curry-Howard counterparts for syntactic representations of natural language, with no intermediate translation to logical formulae. The Lambek -based system we propose extends the system by Fyodorov et~al., which is based on the Ajdukiewicz/Bar-Hillel calculus Bar Hillel,. This (...) enables the system to deal with new kinds of inferences, involving relative clauses, non-constituent coordination, and meaning postulates that involve complex expressions. Basing the system on the Lambek calculus leads to problems with non-normalized proof terms, which are treated by using normalization axioms. (shrink)

We present a novel generative model for natural language tree structures in which semantic (lexical dependency) and syntactic (PCFG) structures are scored with separate models. This factorization provides conceptual simplicity, straightforward opportunities for separately improving the component models, and a level of performance comparable to similar, non-factored models. Most importantly, unlike other modern parsing models, the factored model admits an extremely effective A* parsing algorithm, which enables efficient, exact inference.

Progress in philosophy is difficult to achieve because our methods are evidentially and rhetorically weak. In the last two decades, experimental philosophers have begun to employ the methods of the social sciences to address philosophical questions. However, the adequacy of these methods has been called into question by repeated failures of replication. Experimental philosophers need to incorporate more robust methods to achieve a multi-modal perspective. In this chapter, we describe and showcase cutting-edge methods for data-mining and visualization. Big data is (...) a useful investigatory tool for moral psychology, and it fits well with the Ramsification method the first author advances in a series of recent papers. The guiding insight of these papers is that we can infer the meaning and structure of concepts from patterns of assertions and inferential associations in natural language. (shrink)

We describe a knowledge representation and inference formalism, based on an intensional propositional semantic network, in which variables are structures terms consisting of quantifier, type, and other information. This has three important consequences for natural language processing. First, this leads to an extended, more natural formalism whose use and representations are consistent with the use of variables in natural language in two ways: the structure of representations mirrors the structure of the language and (...) allows re-use phenomena such as pronouns and ellipsis. Second, the formalism allows the specification of description subsumption as a partial ordering on related concepts (variable nodes in a semantic network) that relates more general concepts to more specific instances of that concept, as is done in language. Finally, this structured variable representation simplifies the resolution of some representational difficulties with certain classes of natural language sentences, namely, donkey sentences and sentences involving branching quantifiers. The implementation of this formalism is called ANALOG (A NAtural LOGIC) and its utility for natural language processing tasks is illustrated. (shrink)

The topic of temporal meaning in texts has received considerable attention in recent years from scholars in linguistics, logical semantics, cognitive science, and artificial intelligence. Representing Time in Natural Language offers a systematic and detailed account of how we use temporal information contained in a text or in discourse to reason about the flow of time, inferring the order in which events happened when this is not explicitly stated. A new representational system is designed to formalize an appropriately (...) context-dependent notion of situated inference. The Dynamic Aspect Tree, representing temporal dependencies, constitutes a novel and important dynamic temporal logic, one that makes it easy to see "what follows when" from the information given in an ordinary English text. (shrink)

In this article, we first present the platform and the Dislog language, designed for discourse analysis with a logic and linguistic perspective. The platform has now reached a certain level of maturity which allows the recognition of a large diversity of discourse structures including general-purpose rhetorical structures as well as domain-specific discourse structures. The Dislog language is based on linguistic considerations and includes knowledge access and inference capabilities. Functionalities of the language are presented together with a (...) method for writing discourse analysis rules. Efficiency and portability of the system over domains and languages are investigated to conclude this first part. In a second part, we analyse the different types of arguments found in several document genres, most notably: procedures, didactic texts and requirements. Arguments form a large class of discourse relations. A generic and frequently encountered form emerges from our analysis: ?reasons for conclusion? which constitutes a homogeneous family of arguments from a language, functional and conceptual point of view. This family can be viewed as a kind of proto-argument. We then elaborate its linguistic structure and show how it is implemented in . We then investigate the cooperation between explanation and arguments, in particular in didactic texts where they are particularly rich and elaborated. This article ends with a prospective section that develops current and potential uses of this work and how it can be extended to the recognition of other forms of arguments. (shrink)

‘all A are B’ ; A ⊆ B ‘no A are B’ ; A ⊆ B ‘some A are not B’ ; A ⊆ B ‘some A are B’ ; A ⊆ B (equivalently: A ∩ B = ∅). A knowledge base is a list of triples (Class1, Class2, Boolean) where (A, B, ) expresses that A ⊆ B, and (A, B, ⊥) expresses that A ⊆ B.

I revisit my earlier arguments for the (trivial) inconsistency of natural languages, and take up the objection that no such argument can be established on the basis of surface usage. I respond with the evidential centrality of surface usage: the ways it can and can't be undercut by linguistic science. Then some important ramifications of having an inconsistent natural language are explored: (1) the temptation to engage in illegitimate reductio reasoning, (2) the breakdown of the knowledge idiom (...) (because its facticity isn't comfortable with every sentence being true and false). Restoring the utility of the knowledge idiom motivates regimentation - the consistentist reinterpretation of natural language inferences (as they occur in real time). It's then sketched how to give the semantics of a natural language despite its inconsistency: Necessary and sufficient truth conditions are dropped in favor of necessary truth conditions and sufficient truth conditions. (shrink)

I revisit my earlier arguments for the (trivial) inconsistency of natural languages, and take up the objection that no such argument can be established on the basis of surface usage. I respond with the evidential centrality of surface usage: the ways it can and can't be undercut by linguistic science. Then some important ramifications of having an inconsistent natural language are explored: (1) the temptation to engage in illegitimate reductio reasoning, (2) the breakdown of the knowledge idiom (...) (because its facticity isn't comfortable with every sentence being true and false). Restoring the utility of the knowledge idiom motivates regimentation—the consistentist reinterpretation of natural language inferences (as they occur in real time). It's then sketched how to give the semantics of a natural language despite its inconsistency: Necessary and sufficient truth conditions are dropped in favor of necessary truth conditions and sufficient truth conditions. (shrink)

Using contemporary science, the paper builds on Wittgenstein’s views of human language. Rather than ascribing reality to inscription-like entities, it links embodiment with distributed cognition. The verbal or (quasi) technological aspect of language is traced to not action, but human specific interactivity. This species-specific form of sense-making sustains, among other things, using texts, making/construing phonetic gestures and thinking. Human action is thus grounded in appraisals or sense-saturated coordination. To illustrate interactivity at work, the paper focuses on a case (...) study. Over 11 s, a crime scene investigator infers that she is probably dealing with an inside job: she uses not words, but intelligent gaze. This connects professional expertise to circumstances and the feeling of thinking. It is suggested that, as for other species, human appraisal is based in synergies. However, since the verbal aspect of language constrains action and thinking, we also develop customary ways of behaving. Humans extend embodiment by linking real-time activity to actions through which the collectivity imposes a variable degree of control over how individuals realise values. (shrink)

Inner speech is a pervasive feature of our conscious mental lives. Yet its function and character remain an issue of philosophical debate. The present paper focuses on the relation between inner speech and natural language and on the cognitive functions that various contributors have ascribed to inner speech. In particular, it is argued that inner speech does not consist of bare, context-free internal presentations of sentential (or subsentential) content, but rather has an ineliminably perspectival element. The proposed model (...) of inner speech, which characterizes inner speech as akin to the testimony of an inner interlocutor, accounts for this perspectival element and, it is argued, is explanatorily superior, insofar as it better explains, amongst other phenomena, the often condensed character of inner speech. (shrink)

This text offers an innovative approach to the teaching of logic, which is rigorous but entirely non-symbolic. By introducing students to deductive inferences in natural language, the book breaks new ground pedagogically. Cannon focuses on such topics as using a tableaux technique to assess inconsistency; using generative grammar; employing logical analyses of sentences; and dealing with quantifier expressions and syllogisms. An appendix covers truth-functional logic.

We present a formal system for reasoning about inclusion and exclusion in natural language, following work by MacCartney and Manning. In particular, we show that an extension of the Monotonicity Calculus, augmented by six new type markings, is sufficient to derive novel inferences beyond monotonicity reasoning, and moreover gives rise to an interesting logic of its own. We prove soundness of the resulting calculus and discuss further logical and linguistic issues, including a new connection to the classes of (...) weak, strong, and superstrong negative polarity items. (shrink)

As modern deep networks become more complex, and get closer to human‐like capabilities in certain domains, the question arises as to how the representations and decision rules they learn compare to the ones in humans. In this work, we study representations of sentences in one such artificial system for natural language processing. We first present a diagnostic test dataset to examine the degree of abstract composable structure represented. Analyzing performance on these diagnostic tests indicates a lack of systematicity (...) in representations and decision rules, and reveals a set of heuristic strategies. We then investigate the effect of training distribution on learning these heuristic strategies, and we study changes in these representations with various augmentations to the training set. Our results reveal parallels to the analogous representations in people. We find that these systems can learn abstract rules and generalize them to new contexts under certain circumstances—similar to human zero‐shot reasoning. However, we also note some shortcomings in this generalization behavior—similar to human judgment errors like belief bias. Studying these parallels suggests new ways to understand psychological phenomena in humans as well as informs best strategies for building artificial intelligence with human‐like language understanding. (shrink)

Syllogism is a type of everyday reasoning. For instance, given that ‘Avicenna wrote the famous book the Canon of Medicine’ and ‘The Canon of Medicine has influenced modern medicine,’ it can be conc...

Models of category learning have been extensively studied in cognitive science and primarily tested on perceptual abstractions or artificial stimuli. In this paper, we focus on categories acquired from natural language stimuli, that is, words. We present a Bayesian model that, unlike previous work, learns both categories and their features in a single process. We model category induction as two interrelated subproblems: the acquisition of features that discriminate among categories, and the grouping of concepts into categories based on (...) those features. Our model learns categories incrementally using particle filters, a sequential Monte Carlo method commonly used for approximate probabilistic inference that sequentially integrates newly observed data and can be viewed as a plausible mechanism for human learning. Experimental results show that our incremental learner obtains meaningful categories which yield a closer fit to behavioral data compared to related models while at the same time acquiring features which characterize the learned categories. (shrink)

A formal, computational, semantically clean representation of natural language is presented. This representation captures the fact that logical inferences in natural language crucially depend on the semantic relation of entailment between sentential constituents such as determiner, noun, adjective, adverb, preposition, and verb phrases.The representation parallels natural language in that it accounts for human intuition about entailment of sentences, it preserves its structure, it reflects the semantics of different syntactic categories, it simulates conjunction, disjunction, and (...) negation in natural language by computable operations with provable mathematical properties, and it allows one to represent coordination on different syntactic levels. (shrink)

We describe a prototype ontology-driven information system (ODIS) that exploits what we call Portion of Reality (POR) representations. The system takes both sensor data and natural language text as inputs and composes on this basis logically structured POR assertions. The goal of our prototype is to represent both natural language and sensor data within a single framework that is able to support both axiomatic reasoning and computation. In addition, the framework should be capable of discovering and (...) representing new kinds of situations and thematic roles, (e.g., roles such as agent, patient and instrument), based on new compositions of existing representations. We applied our prototype in an intelligence analysis use case to test the hypothesis that a framework of this sort can produce usefully structured information from combined natural language and sensor data inputs. We further tested our hypothesis by adding an enhanced US Air Force ontology framework to our ODIS in order to (1) process a collection of sensor data, intel reports, and mission plans; (2) build composite POR representations from these data; and (3) machine analyze the fused results to infer mission threats. (shrink)

In this paper a logical interpretation of semantic nets and graph grammars is proposed for modelling natural language understanding and creating language understanding computer systems. An example of parsing a Finnish question by graph grammars and inferring the answer to it by a semantic net representation is provided.

In a recent paper Johan van Benthem reviews earlier work done by himself and colleagues on ‘natural logic’. His paper makes a number of challenging comments on the relationships between traditional logic, modern logic and natural logic. I respond to his challenge, by drawing what I think are the most significant lines dividing traditional logic from modern. The leading difference is in the way logic is expected to be used for checking arguments. For traditionals the checking is local, (...) i.e. separately for each inference step. Between inference steps, several kinds of paraphrasing are allowed. Today we formalise globally: we choose a symbolisation that works for the entire argument, and thus we eliminate intuitive steps and changes of viewpoint during the argument. Frege and Peano recast the logical rules so as to make this possible. I comment also on the traditional assumption that logical processing takes place at the top syntactic level, and I question Johan’s view that natural logic is ‘natural’. (shrink)

In this paper we present an approach to conditional reasoning tasks based on two main ideas. The first idea is that, in contrast with what is usually assumed, an ‘if… then…’ sentence is not an adequate translation in natural language of a logical formula containing a material implication as its principal operator. The second idea is that when subjects are required to check the validity of a sentence in a task, their inferences are not driven uniquely by the (...) content of the sentence, but also by other information embedded in the task scenario and, eventually, by their knowledge about the topic (i.e. information stored in memory). Data from 30 subjects tested on six different tasks are reported as evidence for our approach. The results show that conditional tasks are significantly more complex only when they are presented with ‘if… then…’ sentences and when the subject cannot rely on any extra information, such as contextual information or knowledge about the topic. (shrink)

This paper presents the first use of a computational model of natural logic—a system of logical inference which operates over natural language—for textual inference. Most current approaches to the PAS- CAL RTE textual inference task achieve robustness by sacrificing semantic precision; while broadly effective, they are easily confounded by ubiquitous inferences involving monotonicity. At the other extreme, systems which rely on first-order logic and theorem proving are precise, but excessively brittle. This work aims at (...) a middle way. Our system finds a low-cost edit sequence which transforms the premise into the hypothesis; learns to classify entailment relations across atomic edits; and composes atomic entailments into a top-level entailment judgment. We provide the first reported results for any system on the FraCaS test suite. We also evaluate on RTE3 data, and show that hybridizing an existing RTE system with our natural logic system yields significant performance gains. (shrink)

This paper trials new experimental methods for the analysis of natural language reasoning and the development of critical ordinary language philosophy in the wake of J.L. Austin. Philosophical arguments and thought experiments are strongly shaped by default pragmatic inferences, including stereotypical inferences. Austin suggested that contextually inappropriate stereotypical inferences are at the root of some philosophical paradoxes and problems, and that these can be resolved by exposing those verbal fallacies. This paper builds on recent efforts to empirically (...) document inappropriate stereotypical inferences that may drive philosophical arguments. We demonstrate that previously employed questionnaire-based output measures do not suffice to exclude relevant confounds. We then report an experiment that combines reading time measurements with plausibility ratings. The study seeks to provide evidence of inappropriate stereotypical inferences from appearance verbs that have been suggested to lie at the root of the influential ‘argument from illusion’. Our findings support a diagnostic reconstruction of this argument. They provide the missing component for proof of concept for an experimental implementation of critical ordinary language philosophy that is in line with the ambitions of current ‘evidential’ experimental philosophy. (shrink)

Semantics plays a role in grammar in at least three guises. (A) Linguists seek to account for speakers‘ knowledge of what linguistic expressions mean. This goal is typically achieved by assigning a model theoretic interpretation in a compositional fashion. For example, *No whale flies* is true if and only if the intersection of the sets of whales and fliers is empty in the model. (B) Linguists seek to account for the ability of speakers to make various inferences based on semantic (...) knowledge. For example, *No whale flies* entails *No blue whale flies* and *No whale flies high*. (C) The wellformedness of a variety of syntactic constructions depends on morpho-syntactic features with a semantic flavor. For example, *Under no circumstances would a whale fly* is grammatical, whereas *Under some circumstances would a whale fly* is not, corresponding to the downward vs. upward monotonic features of the preposed phrases. It is usually assumed that once a compositional model theoretic interpretation is assigned to all expressions, its fruits can be freely enjoyed by inferencing and syntax. What place might proof theory have in this picture? (shrink)

A first-order language with a defined identity predicate is proposed whose apparatus for atomic predication is sensitive to grammatical categories of natural language. Subatomic natural deduction systems are defined for this naturalistic first-order language. These systems contain subatomic systems which govern the inferential relations which obtain between naturalistic atomic sentences and between their possibly composite components. As a main result it is shown that normal derivations in the defined systems enjoy the subexpression property which subsumes (...) the subformula property with respect to atomic and identity formulae as a special case. The systems admit a proof-theoretic semantics which does not only apply to logically compound but also to atomic and identity formulae—as well as to their components. The potential of the defined systems for a meticulous first-order analysis of natural inferences whose validity crucially depends on expressions of some of the aforementioned categories is demonstrated. (shrink)

Natural languages make prolific use of conventional constituent‐ordering patterns to indicate “who did what to whom,” yet the mechanisms through which these regularities arise are not well understood. A series of recent experiments demonstrates that, when prompted to express meanings through silent gesture, people bypass native language conventions, revealing apparent biases underpinning word order usage, based on the semantic properties of the information to be conveyed. We extend the scope of these studies by focusing, experimentally and computationally, on (...) the interpretation of silent gesture. We show cross‐linguistic experimental evidence that people use variability in constituent order as a cue to obtain different interpretations. To illuminate the computational principles that govern interpretation of non‐conventional communication, we derive a Bayesian model of interpretation via biased inductive inference and estimate these biases from the experimental data. Our analyses suggest people's interpretations balance the ambiguity that is characteristic of emerging language systems, with ordering preferences that are skewed and asymmetric, but defeasible. (shrink)

Natural languages make prolific use of conventional constituent‐ordering patterns to indicate “who did what to whom,” yet the mechanisms through which these regularities arise are not well understood. A series of recent experiments demonstrates that, when prompted to express meanings through silent gesture, people bypass native language conventions, revealing apparent biases underpinning word order usage, based on the semantic properties of the information to be conveyed. We extend the scope of these studies by focusing, experimentally and computationally, on (...) the interpretation of silent gesture. We show cross‐linguistic experimental evidence that people use variability in constituent order as a cue to obtain different interpretations. To illuminate the computational principles that govern interpretation of non‐conventional communication, we derive a Bayesian model of interpretation via biased inductive inference and estimate these biases from the experimental data. Our analyses suggest people's interpretations balance the ambiguity that is characteristic of emerging language systems, with ordering preferences that are skewed and asymmetric, but defeasible. (shrink)

The paper addresses the phenomenology of inference. It proposes that the conscious character of conscious inferences is partly constituted by a sense of meaning; specifically, a sense of what Grice called ‘natural meaning’. In consciously drawing the (outright, categorical) conclusion that Q from a presumed fact that P, one senses the presumed fact that P as meaning that Q, where ‘meaning that’ expresses natural meaning. This sense of natural meaning is phenomenologically analogous, I suggest, to our (...) sense of what is said in fluently comprehending everyday utterances in our first language. The proposal that conscious inference involves a sense of natural meaning is compared with views according to which conscious inference involves taking the premises (i) to be good reasons for the conclusion (as defended by Thomson and Grice), (ii) to support it (as argued by Audi and, recently, Boghossian), or (iii) to imply it (as lately contended by Broome). I argue our proposal can explain certain phenomena handled by alternatives (i) and (ii), but that some further phenomena is handled by our account but not these alternatives. In relation to alternative (iii), I argue that, in so far as implicational and natural-meaning relations come apart, the latter are a better fit for what we sense or take to be so in conscious inference. (shrink)

This paper develops a version of Natural Logic – an inference system that works directly on natural language syntactic representations, with no intermediate translation to logical formulae. Following work by Sánchez, we develop a small fragment that computes semantic order relations between derivation trees in Categorial Grammar. The proposed system has the following new characteristics: It uses orderings between derivation trees as purely syntactic units, derivable by a formal calculus. The system is extended for conjunctive phenomena (...) like coordination and relative clauses. This allows a simple account of non-monotonic expressions that are reducible to conjunctions of monotonic ones. A decision procedure for provability is developed for a fragment of Natural Logic. (shrink)

Linguistic Issues in Language Technology (LiLT) is an open-access journal that focuses on the relationships between linguistic insights and language technology. In conjunction with machine learning and statistical techniques, deeper and more sophisticated models of language and speech are needed to make significant progress in both existing and newly emerging areas of computational language analysis. The vast quantity of electronically accessible natural language data (text and speech, annotated and unannotated, formal and informal) provides unprecedented (...) opportunities for data-intensive analysis of linguistic phenomena, which can in turn enrich computational methods. Taking an eclectic view on methodology, LiLT provides a forum for this work. In this volume, contributors offer new perspectives on semantic representations for textual inference. (shrink)

This article considers natural signs and their role in the origin of language. Natural signs, sometimes called primary signs, are connected with their signified by causal relationships, concomitance, or likeliness. And their acquisition is directed by both objective reality and past experience (memory). The discovery and use of natural signs is a required prerequisite of existence for any living systems because they are indispensable to movement, the search for food, regulation, communication, and many other information-related activities. (...) It is argued that the birth of conventional signs, sometimes called secondary signs, was determined by a connotative use of natural signs and that, regulated and maintained by them, human language developed. At the same time, the origin and development of human language presupposes a ‘rational turn’ from the given and external reality of natural signs to the rationally constructed reality of artificial signs and rules that are internally maintained by the subjects’ deliberate activities, and actual and inherited social tradition (social memory). In view of this, language is defined as a dynamic system that must both be natural and artificial, empirical and a priori, inductive and deductive. This bilateral origin and regulation of language is the dual-inference of language. (shrink)

In this paper we address the modelling of reasoning processes in natural agents. We focus on a very basic kind of non-monotonic inference for which we identify a simple and plausible underlying process, and we develop a family of logical models that allow to match this process. Partial worlds models, as we call them, are a variant of Kraus, Lehmann and Magidor’s cumulative models. We show that the inference relations they induce form a strict subclass of cumulative (...) relations and tackle the issue of providing sound and complete sets of rules to characterise them. Taking inspiration from the work of Gabbay and Schlechta, we analyse the question in terms of definable sets of partial worlds and conclude that completeness is probably unreachable using a standard propositional language. This brings us to enrich our language with an additional connective, which makes it possible to distinguish between two kinds of disjunction in the partial worlds context. Within this renewed framework, we provide two representation theorems: one for inference relations induced by precisification-free smooth models, and the other for inference relations induced by precisification-free ranked models. Finally we give an aperçu of how partial worlds models lend themselves to the modelling of learning. (shrink)