Developers of text generation models rely on automated evaluation metrics as a stand-in for slow and expensive manual evaluations. However, image captioning metrics have struggled to give accurate learned estimates of the semantic and pragmatic success of output text. We address this weakness by introducing the first discourse-aware learned generation metric for evaluating image descriptions. Our approach is inspired by computational theories of discourse for capturing information goals using coherence. We present a dataset of image–description pairs annotated with coherence relations. We then train a coherence-aware metric on a subset of the Conceptual Captions dataset and measure its effectiveness—its ability to predict human ratings of output captions—on a test set composed of out-of-domain images. We demonstrate a higher Kendall Correlation Coefficient for our proposed metric with the human judgments for the results of a number of state-of-the-art coherence-aware caption generation models when compared to several other metrics including recently proposed learned metrics such as BLEURT and BERTScore.

We use coherence relations inspired by computational models of discourse to study the information needs and goals of image captioning. Using an annotation protocol specifically devised for capturing image–caption coherence relations, we annotate 10,000 instances from publicly-available image–caption pairs. We introduce a new task for learning inferences in imagery and text, coherence relation prediction, and show that these coherence annotations can be exploited to learn relation classifiers as an intermediary step, and also train coherence-aware, controllable image captioning models. The results show a dramatic improvement in the consistency and quality of the generated captions with respect to information needs specified via coherence relations.

All communication aims at achieving common ground (grounding): interlocutors can work together effectively only with mutual beliefs about what the state of the world is, about what their goals are, and about how they plan to make their goals a reality. Computational dialogue research offers some classic results on grouding, which unfortunately offer scant guidance to the design of grounding modules and behaviors in cutting-edge systems. In this tutorial, we focus on three main topic areas: 1) grounding in human-human communication; 2) grounding in dialogue systems; and 3) grounding in multi-modal interactive systems, including image-oriented conversations and human-robot interactions. We highlight a number of achievements of recent computational research in coordinating complex content, show how these results lead to rich and challenging opportunities for doing grounding in more flexible and powerful ways, and canvass relevant insights from the literature on human–human conversation. We expect that the tutorial will be of interest to researchers in dialogue systems, computational semantics and cognitive modeling, and hope that it will catalyze research and system building that more directly explores the creative, strategic ways conversational agents might be able to seek and offer evidence about their understanding of their interlocutors.

In many domains, dialogue systems need to work collaboratively with users to successfully reconstruct the meaning the user had in mind. In this paper, we show how cognitive models of users’ communicative strategies can be leveraged in a reinforcement learning approach to dialogue planning to enable interactive systems to give targeted, effective feedback about the system’s understanding. We describe a prototype system that collaborates on reference tasks that distinguish arbitrarily varying color patches from similar distractors, and use experiments with crowd workers and analyses of our learned policies to document that our approach leads to context-sensitive clarification strategies that focus on key missing information, elicit correct answers that the system understands, and contribute to increasing dialogue success.

The interpretation of the lexical aspect of verbs in English plays a crucial role in tasks such as recognizing textual entailment and learning discourse-level inferences. We show that two elementary dimensions of aspectual class, states vs. events, and telic vs. atelic events, can be modelled effectively with distributional semantics. We find that a verb’s local context is most indicative of its aspectual class, and we demonstrate that closed class words tend to be stronger discriminating contexts than content words. Our approach outperforms previous work on three datasets. Further, we present a new dataset of human-human conversations annotated with lexical aspects and present experiments that show the correlation of telicity with genre and discourse goals.

We analyze a corpus of referential communication through the lens of quantitative models of speaker reasoning. Different models place different emphases on linguistic reasoning and collaborative reasoning. This leads models to make different assessments of the risks and rewards of using specific utterances in specific contexts. By fitting a latent variable model to the corpus, we can exhibit utterances that give systematic evidence of the diverse kinds of reasoning speakers employ, and build integrated models that recognize not only speaker reference but also speaker reasoning.

We study verbs in image–text corpora, contrasting caption corpora, where texts are explicitly written to characterize image content, with depiction corpora, where texts and images may stand in more general relations. Captions show a distinctively limited distribution of verbs, with strong preferences for specific tense, aspect, lexical aspect, and semantic field. These limitations, which appear in data elicited by a range of methods, restrict the utility of caption corpora to inform image retrieval, multimodal document generation, and perceptually-grounded semantic models. We suggest that these limitations reflect the discourse constraints in play when subjects write texts to accompany imagery, so we argue that future development of image–text corpora should work to increase the diversity of event descriptions, while looking explicitly at the different ways text and imagery can be coherently related.

This paper presents a novel crowd-sourced resource for multimodal discourse: our resource characterizes inferences in image-text contexts in the domain of cooking recipes in the form of coherence relations. Like previous corpora annotating discourse structure between text arguments, such as the Penn Discourse Treebank, our new corpus aids in establishing a better understanding of natural communication and common-sense reasoning, while our findings have implications for a wide range of applications, such as understanding and generation of multimodal documents.

Arrows are a key ingredient of schematic pictorial communication. This paper investigates the interpretation of arrows through linguistic, crowdsourcing and machine-learning methodology. Our work establishes a novel analogy between arrows and verbs: we advocate representing arrows in terms of qualitatively different structural and semantic frames, and resolving frames to specific interpretations using shallow world knowledge.

A key component in surface realization in natural language generation is to choose concrete syntactic relationships to express a target meaning. We develop a new method for syntactic choice based on learning a stochastic tree grammar in a neural architecture. This framework can exploit state-of-the-art methods for modeling word sequences and generalizing across vocabulary. We also induce embeddings to generalize over elementary tree structures and exploit a tree recurrence over the input structure to model long-distance influences between NLG choices. We evaluate the models on the task of linearizing unannotated dependency trees, documenting the contribution of our modeling techniques to improvements in both accuracy and run time.

Natural language meanings allow speakers to encode important real-world distinctions, but corpora of grounded language use also reveal that speakers categorize the world in different ways and describe situations with different terminology. To learn meanings from data, we therefore need to link underlying representations of meaning to models of speaker judgment and speaker choice. This paper describes a new approach to this problem: we model variability through uncertainty in categorization boundaries and distributions over preferred vocabulary. We apply the approach to a large data set of color descriptions, where statistical evaluation documents its accuracy. The results are available as a Lexicon of Uncertain Color Standards (LUX), which supports future efforts in grounded language understanding and generation by probabilistically mapping 829 English color descriptions to potentially context-sensitive regions in HSV color space.