We propose a new model for metaphor detection in which an expectation component estimates representations of expected word meanings in a given context, whereas a realization component computes representations of target word meanings in context. We also introduce a systematic evaluation methodology that estimates generalization performance in three settings: within distribution, a new strong out of distribution setting, and a novel out-of-pretraining setting. Across all settings, the expectation-realization model obtains results that are competitive with or better than previous metaphor detection models.

Socratic questioning is a teaching strategy where the student is guided towards solving a problem on their own, instead of being given the solution directly. In this paper, we introduce a dataset of Socratic conversations where an instructor helps a novice programmer fix buggy solutions to simple computational problems. The dataset is then used for benchmarking the Socratic debugging abilities of GPT-based language models. While GPT-4 is observed to perform much better than GPT-3.5, its precision, and recall still fall short of human expert abilities, motivating further work in this area.

The ever-growing complexity of mathematical proofs makes their manual verification by mathematicians very cognitively demanding. Autoformalization seeks to address this by translating proofs written in natural language into a formal representation that is computer-verifiable via interactive theorem provers. In this paper, we introduce a semantic parsing approach, based on the Universal Transformer architecture, that translates elementary mathematical proofs into an equivalent formalization in the language of the Coq interactive theorem prover. The same architecture is also trained to translate simple imperative code decorated with Hoare triples into formally verifiable proofs of correctness in Coq. Experiments on a limited domain of artificial and human-written proofs show that the models generalize well to intermediate lengths not seen during training and variations in natural language.

Novelty or surprise is a fundamental attribute of creative output. As such, we postulate that a writer’s creative use of language leads to word choices and, more importantly, corresponding semantic structures that are unexpected for the reader. In this paper we investigate measures of surprise that rely solely on word distributions computed by language models and show empirically that creative language such as humor and metaphor is strongly correlated with surprise. Surprisingly at first, information content is observed to be at least as good a predictor of creative language as any of the surprise measures investigated. However, the best prediction performance is obtained when information and surprise measures are combined, showing that surprise measures capture an aspect of creative language that goes beyond information content.

We describe a new semantic parsing setting that allows users to query the system using both natural language questions and actions within a graphical user interface. Multiple time series belonging to an entity of interest are stored in a database and the user interacts with the system to obtain a better understanding of the entity’s state and behavior, entailing sequences of actions and questions whose answers may depend on previous factual or navigational interactions. We design an LSTM-based encoder-decoder architecture that models context dependency through copying mechanisms and multiple levels of attention over inputs and previous outputs. When trained to predict tokens using supervised learning, the proposed architecture substantially outperforms standard sequence generation baselines. Training the architecture using policy gradient leads to further improvements in performance, reaching a sequence-level accuracy of 88.7% on artificial data and 74.8% on real data.

The automation of tasks in community question answering (cQA) is dominated by machine learning approaches, whose performance is often limited by the number of training examples. Starting from a neural sequence learning approach with attention, we explore the impact of two data augmentation techniques on question ranking performance: a method that swaps reference questions with their paraphrases, and training on examples automatically selected from external datasets. Both methods are shown to lead to substantial gains in accuracy over a strong baseline. Further improvements are obtained by changing the model architecture to mirror the structure seen in the data.