Understanding pragmatics-the use of language in context-is crucial for developing NLP systems capable of interpreting nuanced language use. Despite recent advances in language technologies, including large language models, evaluating their ability to handle pragmatic phenomena such as implicatures and references remains challenging. To advance pragmatic abilities in models, it is essential to understand current evaluation trends and identify existing limitations. In this survey, we provide a comprehensive review of resources designed for evaluating pragmatic capabilities in NLP, categorizing datasets by the pragmatic phenomena they address. We analyze task designs, data collection methods, evaluation approaches, and their relevance to real-world applications. By examining these resources in the context of modern language models, we highlight emerging trends, challenges, and gaps in existing benchmarks. Our survey aims to clarify the landscape of pragmatic evaluation and guide the development of more comprehensive and targeted benchmarks, ultimately contributing to more nuanced and context-aware NLP models.
This study investigates whether vision-language models (VLMs) can perform pragmatic inference, focusing on ignorance implicatures, utterances that imply the speaker's lack of precise knowledge. To test this, we systematically manipulated contextual cues: the visually depicted situation (visual cue) and QUD-based linguistic prompts (linguistic cue). When only visual cues were provided, three state-of-the-art VLMs (GPT-4o, Gemini 1.5 Pro, and Claude 3.5 sonnet) produced interpretations largely based on the lexical meaning of the modified numerals. When linguistic cues were added to enhance contextual informativeness, Claude exhibited more human-like inference by integrating both types of contextual cues. In contrast, GPT and Gemini favored precise, literal interpretations. Although the influence of contextual cues increased, they treated each contextual cue independently and aligned them with semantic features rather than engaging in context-driven reasoning. These findings suggest that although the models differ in how they handle contextual cues, Claude's ability to combine multiple cues may signal emerging pragmatic competence in multimodal models.
The development of artificial intelligence systems capable of understanding and reasoning about complex real-world scenarios is a significant challenge. In this work we present a novel approach to enhance and exploit LLM reactive capability to address complex problems and interpret deeply contextual real-world meaning. We introduce a method and a tool for creating a multimodal, knowledge-augmented formal representation of meaning that combines the strengths of large language models with structured semantic representations. Our method begins with an image input, utilizing state-of-the-art large language models to generate a natural language description. This description is then transformed into an Abstract Meaning Representation (AMR) graph, which is formalized and enriched with logical design patterns, and layered semantics derived from linguistic and factual knowledge bases. The resulting graph is then fed back into the LLM to be extended with implicit knowledge activated by complex heuristic learning, including semantic implicatures, moral values, embodied cognition, and metaphorical representations. By bridging the gap between unstructured language models and formal semantic structures, our method opens new avenues for tackling intricate problems in natural language understanding and reasoning.
Markedness in natural language is often associated with non-literal meanings in discourse. Differential Object Marking (DOM) in Korean is one instance of this phenomenon, where post-positional markers are selected based on both the semantic features of the noun phrases and the discourse features that are orthogonal to the semantic features. Previous work has shown that distributional models of language recover certain semantic features of words -- do these models capture implied discourse-level meanings as well? We evaluate whether a set of large language models are capable of associating discourse meanings with different object markings in Korean. Results suggest that discourse meanings of a grammatical marker can be more challenging to encode than that of a discourse marker.
Human communication is based on a variety of inferences that we draw from sentences, often going beyond what is literally said. While there is wide agreement on the basic distinction between entailment, implicature, and presupposition, the status of many inferences remains controversial. In this paper, we focus on three inferences of plain and embedded disjunctions, and compare them with regular scalar implicatures. We investigate this comparison from the novel perspective of the predictions of state-of-the-art large language models, using the same experimental paradigms as recent studies investigating the same inferences with humans. The results of our best performing models mostly align with those of humans, both in the large differences we find between those inferences and implicatures, as well as in fine-grained distinctions among different aspects of those inferences.
Understanding the non-literal meaning of an utterance is critical for large language models (LLMs) to become human-like social communicators. In this work, we introduce SwordsmanImp, the first Chinese multi-turn-dialogue-based dataset aimed at conversational implicature, sourced from dialogues in the Chinese sitcom $\textit{My Own Swordsman}$. It includes 200 carefully handcrafted questions, all annotated on which Gricean maxims have been violated. We test eight close-source and open-source LLMs under two tasks: a multiple-choice question task and an implicature explanation task. Our results show that GPT-4 attains human-level accuracy (94%) on multiple-choice questions. CausalLM demonstrates a 78.5% accuracy following GPT-4. Other models, including GPT-3.5 and several open-source models, demonstrate a lower accuracy ranging from 20% to 60% on multiple-choice questions. Human raters were asked to rate the explanation of the implicatures generated by LLMs on their reasonability, logic and fluency. While all models generate largely fluent and self-consistent text, their explanations score low on reasonability except for GPT-4, suggesting that most LLMs cannot produce satisfactory explanations of the implicatures in the conversation. Moreover, we find LLMs' performance does not vary significantly by Gricean maxims, suggesting that LLMs do not seem to process implicatures derived from different maxims differently. Our data and code are available at https://github.com/sjtu-compling/llm-pragmatics.
Scalar adjectives pertain to various domain scales and vary in intensity within each scale (e.g. certain is more intense than likely on the likelihood scale). Scalar implicatures arise from the consideration of alternative statements which could have been made. They can be triggered by scalar adjectives and require listeners to reason pragmatically about them. Some scalar adjectives are more likely to trigger scalar implicatures than others. This phenomenon is referred to as scalar diversity. In this study, we probe different families of Large Language Models such as GPT-4 for their knowledge of the lexical semantics of scalar adjectives and one specific aspect of their pragmatics, namely scalar diversity. We find that they encode rich lexical-semantic information about scalar adjectives. However, the rich lexical-semantic knowledge does not entail a good understanding of scalar diversity. We also compare current models of different sizes and complexities and find that larger models are not always better. Finally, we explain our probing results by leveraging linguistic intuitions and model training objectives.
We present a model of pragmatic language understanding, where utterances are produced and understood by searching for regularized equilibria of signaling games. In this model (which we call ReCo, for Regularized Conventions), speakers and listeners search for contextually appropriate utterance--meaning mappings that are both close to game-theoretically optimal conventions and close to a shared, ''default'' semantics. By characterizing pragmatic communication as equilibrium search, we obtain principled sampling algorithms and formal guarantees about the trade-off between communicative success and naturalness. Across several datasets capturing real and idealized human judgments about pragmatic implicatures, ReCo matches or improves upon predictions made by best response and rational speech act models of language understanding.
Conversational implicatures are pragmatic inferences that require listeners to deduce the intended meaning conveyed by a speaker from their explicit utterances. Although such inferential reasoning is fundamental to human communication, recent research indicates that large language models struggle to comprehend these implicatures as effectively as the average human. This paper demonstrates that by incorporating Grice's Four Maxims into the model through chain-of-thought prompting, we can significantly enhance its performance, surpassing even the average human performance on this task.
Interactive Task Learning (ITL) concerns learning about unforeseen domain concepts via natural interactions with human users. The learner faces a number of significant constraints: learning should be online, incremental and few-shot, as it is expected to perform tangible belief updates right after novel words denoting unforeseen concepts are introduced. In this work, we explore a challenging symbol grounding task--discriminating among object classes that look very similar--within the constraints imposed by ITL. We demonstrate empirically that more data-efficient grounding results from exploiting the truth-conditions of the teacher's generic statements (e.g., "Xs have attribute Z.") and their implicatures in context (e.g., as an answer to "How are Xs and Ys different?", one infers Y lacks attribute Z).
We investigate the optimal model size and number of tokens for training a transformer language model under a given compute budget. We find that current large language models are significantly undertrained, a consequence of the recent focus on scaling language models whilst keeping the amount of training data constant. By training over \nummodels language models ranging from 70 million to over 16 billion parameters on 5 to 500 billion tokens, we find that for compute-optimal training, the model size and the number of training tokens should be scaled equally: for every doubling of model size the number of training tokens should also be doubled. We test this hypothesis by training a predicted compute-optimal model, \chinchilla, that uses the same compute budget as \gopher but with 70B parameters and 4$\times$ more more data. \chinchilla uniformly and significantly outperforms \Gopher (280B), GPT-3 (175B), Jurassic-1 (178B), and Megatron-Turing NLG (530B) on a large range of downstream evaluation tasks. This also means that \chinchilla uses substantially less compute for fine-tuning and inference, greatly facilitating downstream usage. As a highlight, \chinchilla reaches a state-of-the-art average accuracy of 67.5\% on the MMLU benchmark, greater than a 7\% improvement over \gopher.
Language modelling provides a step towards intelligent communication systems by harnessing large repositories of written human knowledge to better predict and understand the world. In this paper, we present an analysis of Transformer-based language model performance across a wide range of model scales -- from models with tens of millions of parameters up to a 280 billion parameter model called Gopher. These models are evaluated on 152 diverse tasks, achieving state-of-the-art performance across the majority. Gains from scale are largest in areas such as reading comprehension, fact-checking, and the identification of toxic language, but logical and mathematical reasoning see less benefit. We provide a holistic analysis of the training dataset and model's behaviour, covering the intersection of model scale with bias and toxicity. Finally we discuss the application of language models to AI safety and the mitigation of downstream harms.
As pre-trained language models (LMs) continue to dominate NLP, it is increasingly important that we understand the depth of language capabilities in these models. In this paper, we target pre-trained LMs' competence in pragmatics, with a focus on pragmatics relating to discourse connectives. We formulate cloze-style tests using a combination of naturally-occurring data and controlled inputs drawn from psycholinguistics. We focus on testing models' ability to use pragmatic cues to predict discourse connectives, models' ability to understand implicatures relating to connectives, and the extent to which models show humanlike preferences regarding temporal dynamics of connectives. We find that although models predict connectives reasonably well in the context of naturally-occurring data, when we control contexts to isolate high-level pragmatic cues, model sensitivity is much lower. Models also do not show substantial humanlike temporal preferences. Overall, the findings suggest that at present, dominant pre-training paradigms do not result in substantial pragmatic competence in our models.
Recognizing a basic difference between the semiotics of humans and machines presents a possibility to overcome the shortcomings of current speech assistive devices. For the machine, the meaning of a (human) utterance is defined by its own scope of actions. Machines, thus, do not need to understand the conventional meaning of an utterance. Rather, they draw conversational implicatures in the sense of (neo-)Gricean pragmatics. For speech assistive devices, the learning of machine-specific meanings of human utterances, i.e. the fossilization of conversational implicatures into conventionalized ones by trial and error through lexicalization appears to be sufficient. Using the quite trivial example of a cognitive heating device, we show that - based on dynamic semantics - this process can be formalized as the reinforcement learning of utterance-meaning pairs (UMP).
Natural language inference (NLI) is an increasingly important task for natural language understanding, which requires one to infer whether a sentence entails another. However, the ability of NLI models to make pragmatic inferences remains understudied. We create an IMPlicature and PRESupposition diagnostic dataset (IMPPRES), consisting of >25k semiautomatically generated sentence pairs illustrating well-studied pragmatic inference types. We use IMPPRES to evaluate whether BERT, InferSent, and BOW NLI models trained on MultiNLI (Williams et al., 2018) learn to make pragmatic inferences. Although MultiNLI appears to contain very few pairs illustrating these inference types, we find that BERT learns to draw pragmatic inferences. It reliably treats scalar implicatures triggered by "some" as entailments. For some presupposition triggers like "only", BERT reliably recognizes the presupposition as an entailment, even when the trigger is embedded under an entailment canceling operator like negation. BOW and InferSent show weaker evidence of pragmatic reasoning. We conclude that NLI training encourages models to learn some, but not all, pragmatic inferences.
Human dialogue often contains utterances having meanings entirely different from the sentences used and are clearly understood by the interlocutors. But in human-computer interactions, the machine fails to understand the implicated meaning unless it is trained with a dataset containing the implicated meaning of an utterance along with the utterance and the context in which it is uttered. In linguistic terms, conversational implicatures are the meanings of the speaker's utterance that are not part of what is explicitly said. In this paper, we introduce a dataset of dialogue snippets with three constituents, which are the context, the utterance, and the implicated meanings. These implicated meanings are the conversational implicatures. The utterances are collected by transcribing from listening comprehension sections of English tests like TOEFL (Test of English as a Foreign Language) as well as scraping dialogues from movie scripts available on IMSDb (Internet Movie Script Database). The utterances are manually annotated with implicatures.
Controlled natural languages (CNLs) are effective languages for knowledge representation and reasoning. They are designed based on certain natural languages with restricted lexicon and grammar. CNLs are unambiguous and simple as opposed to their base languages. They preserve the expressiveness and coherence of natural languages. In this report, we focus on a class of CNLs, called machine-oriented CNLs, which have well-defined semantics that can be deterministically translated into formal languages, such as Prolog, to do logical reasoning. Over the past 20 years, a number of machine-oriented CNLs emerged and have been used in many application domains for problem solving and question answering. However, few of them support non-monotonic inference. In our work, we propose non-monotonic extensions of CNL to support defeasible reasoning. In the first part of this report, we survey CNLs and compare three influential systems: Attempto Controlled English (ACE), Processable English (PENG), and Computer-processable English (CPL). We compare their language design, semantic interpretations, and reasoning services. In the second part of this report, we first identify typical non-monotonicity in natural languages, such as defaults, exceptions and conversational implicatures. Then, we propose their representation in CNL and the corresponding formalizations in a form of defeasible reasoning known as Logic Programming with Defaults and Argumentation Theory (LPDA).
While previous sentiment analysis research has concentrated on the interpretation of explicitly stated opinions and attitudes, this work initiates the computational study of a type of opinion implicature (i.e., opinion-oriented inference) in text. This paper described a rule-based framework for representing and analyzing opinion implicatures which we hope will contribute to deeper automatic interpretation of subjective language. In the course of understanding implicatures, the system recognizes implicit sentiments (and beliefs) toward various events and entities in the sentence, often attributed to different sources (holders) and of mixed polarities; thus, it produces a richer interpretation than is typical in opinion analysis.
Semantic textual similarity (STS) systems are designed to encode and evaluate the semantic similarity between words, phrases, sentences, and documents. One method for assessing the quality or authenticity of semantic information encoded in these systems is by comparison with human judgments. A data set for evaluating semantic models was developed consisting of 775 English word-sentence pairs, each annotated for semantic relatedness by human raters engaged in a Maximum Difference Scaling (MDS) task, as well as a faster alternative task. As a sample application of this relatedness data, behavior-based relatedness was compared to the relatedness computed via four off-the-shelf STS models: n-gram, Latent Semantic Analysis (LSA), Word2Vec, and UMBC Ebiquity. Some STS models captured much of the variance in the human judgments collected, but they were not sensitive to the implicatures and entailments that were processed and considered by the participants. All text stimuli and judgment data have been made freely available.