Image-Grounded Conversations: Multimodal Context for Natural Question and Response Generation

The popularity of image sharing on social media and the engagement it creates between users reflects the important role that visual context plays in everyday conversations. We present a novel task, Image-Grounded Conversations (IGC), in which natural-sounding conversations are generated about a shared image. To benchmark progress, we introduce a new multiple-reference dataset of crowd-sourced, event-centric conversations on images. IGC falls on the continuum between chit-chat and goal-directed conversation models, where visual grounding constrains the topic of conversation to event-driven utterances. Experiments with models trained on social media data show that the combination of visual and textual context enhances the quality of generated conversational turns. In human evaluation, the gap between human performance and that of both neural and retrieval architectures suggests that multi-modal IGC presents an interesting challenge for dialogue research.

Generating Natural Questions About an Image

There has been an explosion of work in the vision & language community during the past few years from image captioning to video transcription, and answering questions about images. These tasks have focused on literal descriptions of the image. To move beyond the literal, we choose to explore how questions about an image are often directed at commonsense inference and the abstract events evoked by objects in the image. In this paper, we introduce the novel task of Visual Question Generation (VQG), where the system is tasked with asking a natural and engaging question when shown an image. We provide three datasets which cover a variety of images from object-centric to event-centric, with considerably more abstract training data than provided to state-of-the-art captioning systems thus far. We train and test several generative and retrieval models to tackle the task of VQG. Evaluation results show that while such models ask reasonable questions for a variety of images, there is still a wide gap with human performance which motivates further work on connecting images with commonsense knowledge and pragmatics. Our proposed task offers a new challenge to the community which we hope furthers interest in exploring deeper connections between vision & language.

Visual Storytelling

We introduce the first dataset for sequential vision-to-language, and explore how this data may be used for the task of visual storytelling. The first release of this dataset, SIND v.1, includes 81,743 unique photos in 20,211 sequences, aligned to both descriptive (caption) and story language. We establish several strong baselines for the storytelling task, and motivate an automatic metric to benchmark progress. Modelling concrete description as well as figurative and social language, as provided in this dataset and the storytelling task, has the potential to move artificial intelligence from basic understandings of typical visual scenes towards more and more human-like understanding of grounded event structure and subjective expression.

A Corpus and Evaluation Framework for Deeper Understanding of Commonsense Stories

Representation and learning of commonsense knowledge is one of the foundational problems in the quest to enable deep language understanding. This issue is particularly challenging for understanding casual and correlational relationships between events. While this topic has received a lot of interest in the NLP community, research has been hindered by the lack of a proper evaluation framework. This paper attempts to address this problem with a new framework for evaluating story understanding and script learning: the 'Story Cloze Test'. This test requires a system to choose the correct ending to a four-sentence story. We created a new corpus of ~50k five-sentence commonsense stories, ROCStories, to enable this evaluation. This corpus is unique in two ways: (1) it captures a rich set of causal and temporal commonsense relations between daily events, and (2) it is a high quality collection of everyday life stories that can also be used for story generation. Experimental evaluation shows that a host of baselines and state-of-the-art models based on shallow language understanding struggle to achieve a high score on the Story Cloze Test. We discuss these implications for script and story learning, and offer suggestions for deeper language understanding.

A Survey of Current Datasets for Vision and Language Research

Integrating vision and language has long been a dream in work on artificial intelligence (AI). In the past two years, we have witnessed an explosion of work that brings together vision and language from images to videos and beyond. The available corpora have played a crucial role in advancing this area of research. In this paper, we propose a set of quality metrics for evaluating and analyzing the vision & language datasets and categorize them accordingly. Our analyses show that the most recent datasets have been using more complex language and more abstract concepts, however, there are different strengths and weaknesses in each.

Probabilistic Frame Induction

In natural-language discourse, related events tend to appear near each other to describe a larger scenario. Such structures can be formalized by the notion of a frame (a.k.a. template), which comprises a set of related events and prototypical participants and event transitions. Identifying frames is a prerequisite for information extraction and natural language generation, and is usually done manually. Methods for inducing frames have been proposed recently, but they typically use ad hoc procedures and are difficult to diagnose or extend. In this paper, we propose the first probabilistic approach to frame induction, which incorporates frames, events, participants as latent topics and learns those frame and event transitions that best explain the text. The number of frames is inferred by a novel application of a split-merge method from syntactic parsing. In end-to-end evaluations from text to induced frames and extracted facts, our method produced state-of-the-art results while substantially reducing engineering effort.

Ambiguity in the Acquisition of Lexical Information

This paper describes an approach to the automatic identification of lexical information in on-line dictionaries. This approach uses bootstrapping techniques, specifically so that ambiguity in the dictionary text can be treated properly. This approach consists of processing an on-line dictionary multiple times, each time refining the lexical information previously acquired and identifying new lexical information. The strength of this approach is that lexical information can be acquired from definitions which are syntactically ambiguous, given that information acquired during the first pass can be used to improve the syntactic analysis of definitions in subsequent passes. In the context of a lexical knowledge base, the types of lexical information that need to be represented cannot be viewed as a fixed set, but rather as a set that will change given the resources of the lexical knowledge base and the requirements of analysis systems which access it.

Automatically Identifying Morphological Relations in = Machine-Readable Dictionaries

We describe an automated method for identifying classes of morphologically related words in an on-line dictionary, and for linking individual senses in the derived form to one or more senses in the base form by means of morphological relation attributes. We also present an algorithm for computing a score reflecting the system=92s certainty in these derivational links; this computation relies on the content of semantic relations associated with each sense, which are extracted automatically by parsing each sense definition and subjecting the parse structure to automated semantic analysis. By processing the entire set of headwords in the dictionary in this fashion we create a large set of directed derivational graphs, which can then be accessed by other components in our broad-coverage NLP system. Spurious or unlikely derivations are not discarded, but are rather added to the dictionary and assigned a negative score; this allows the system to handle non-standard uses of these forms.