The logical negation property (LNP), which implies generating different predictions for semantically opposite inputs (p is true iff ¬p is false), is an important property that a trustworthy language model must satisfy. However, much recent evidence shows that large-size pre-trained language models (PLMs) do not satisfy this property. In this paper, we perform experiments using probing tasks to assess PLMs’ LNP understanding. Unlike previous studies that only examined negation expressions, we expand the boundary of the investigation to lexical semantics. Through experiments, we observe that PLMs violate the LNP frequently. To alleviate the issue, we propose a novel intermediate training task, named meaning-matching, designed to directly learn a meaning text correspondence, instead of relying on the distributional hypothesis. Through multiple experiments, we find that the task enables PLMs to learn lexical semantic information. Also, through fine-tuning experiments on 7 GLUE tasks, we confirm that it is a safe intermediate task that guarantees a similar or better performance of downstream tasks. Finally, we observe that our proposed approach outperforms our previous counterparts despite its time and resource efficiency.

Behavioural consistency is a critical condition for a language model (LM) to become trustworthy like humans. Despite its importance, however, there is little consensus on the definition of LM consistency, resulting in different definitions across many studies. In this paper, we first propose the idea of LM consistency based on behavioural consistency and establish a taxonomy that classifies previously studied consistencies into several sub-categories. Next, we create a new benchmark that allows us to evaluate a model on 19 test cases, distinguished by multiple types of consistency and diverse downstream tasks. Through extensive experiments on the new benchmark, we ascertain that none of the modern pre-trained language models (PLMs) performs well in every test case, while exhibiting high inconsistency in many cases. Our experimental results suggest that a unified benchmark that covers broad aspects (i.e., multiple consistency types and tasks) is essential for a more precise evaluation.

The aim of knowledge base completion is to predict unseen facts from existing facts in knowledge bases. In this work, we introduce the first approach for transfer of knowledge from one collection of facts to another without the need for entity or relation matching. The method works for both canonicalized knowledge bases and uncanonicalized or open knowledge bases, i.e., knowledge bases where more than one copy of a real-world entity or relation may exist. Such knowledge bases are a natural output of automated information extraction tools that extract structured data from unstructured text. Our main contribution is a method that can make use of a large-scale pretraining on facts, collected from unstructured text, to improve predictions on structured data from a specific domain. The introduced method is the most impactful on small datasets such as ReVerb20K, where we obtained a 6% absolute increase of mean reciprocal rank and 65% relative decrease of mean rank over the previously best method, despite not relying on large pre-trained models like BERT.

To increase trust in artificial intelligence systems, a promising research direction consists of designing neural models capable of generating natural language explanations for their predictions. In this work, we show that such models are nonetheless prone to generating mutually inconsistent explanations, such as ”Because there is a dog in the image.” and ”Because there is no dog in the [same] image.”, exposing flaws in either the decision-making process of the model or in the generation of the explanations. We introduce a simple yet effective adversarial framework for sanity checking models against the generation of inconsistent natural language explanations. Moreover, as part of the framework, we address the problem of adversarial attacks with full target sequences, a scenario that was not previously addressed in sequence-to-sequence attacks. Finally, we apply our framework on a state-of-the-art neural natural language inference model that provides natural language explanations for its predictions. Our framework shows that this model is capable of generating a significant number of inconsistent explanations.

Hard cases of pronoun resolution have been used as a long-standing benchmark for commonsense reasoning. In the recent literature, pre-trained language models have been used to obtain state-of-the-art results on pronoun resolution. Overall, four categories of training and evaluation objectives have been introduced. The variety of training datasets and pre-trained language models used in these works makes it unclear whether the choice of training objective is critical. In this work, we make a fair comparison of the performance and seed-wise stability of four models that represent the four categories of objectives. Our experiments show that the objective of sequence ranking performs the best in-domain, while the objective of semantic similarity between candidates and pronoun performs the best out-of-domain. We also observe a seed-wise instability of the model using sequence ranking, which is not the case when the other objectives are used.

The goal of open information extraction (OIE) is to extract facts from natural language text, and to represent them as structured triples of the form <subject,predicate, object>. For example, given the sentence “Beethoven composed the Ode to Joy.”, we are expected to extract the triple <Beethoven, composed, Ode to Joy>. In this work, we systematically compare different neural network architectures and training approaches, and improve the performance of the currently best models on the OIE16 benchmark (Stanovsky and Dagan, 2016) by 0.421 F1 score and 0.420 AUC-PR, respectively, in our experiments (i.e., by more than 200% in both cases). Furthermore, we show that appropriate problem and loss formulations often affect the performance more than the network architecture.

Pronoun resolution is a major area of natural language understanding. However, large-scale training sets are still scarce, since manually labelling data is costly. In this work, we introduce WikiCREM (Wikipedia CoREferences Masked) a large-scale, yet accurate dataset of pronoun disambiguation instances. We use a language-model-based approach for pronoun resolution in combination with our WikiCREM dataset. We compare a series of models on a collection of diverse and challenging coreference resolution problems, where we match or outperform previous state-of-the-art approaches on 6 out of 7 datasets, such as GAP, DPR, WNLI, PDP, WinoBias, and WinoGender. We release our model to be used off-the-shelf for solving pronoun disambiguation.

The Winograd Schema Challenge (WSC) dataset WSC273 and its inference counterpart WNLI are popular benchmarks for natural language understanding and commonsense reasoning. In this paper, we show that the performance of three language models on WSC273 consistently and robustly improves when fine-tuned on a similar pronoun disambiguation problem dataset (denoted WSCR). We additionally generate a large unsupervised WSC-like dataset. By fine-tuning the BERT language model both on the introduced and on the WSCR dataset, we achieve overall accuracies of 72.5% and 74.7% on WSC273 and WNLI, improving the previous state-of-the-art solutions by 8.8% and 9.6%, respectively. Furthermore, our fine-tuned models are also consistently more accurate on the “complex” subsets of WSC273, introduced by Trichelair et al. (2018).