Estimating item parameters (e.g., the difficulty of a question) is an important part of modern high-stakes tests. Conventional methods require lengthy pilots to collect response data from a representative population of test-takers. The need for these pilots limit item bank size and how often those item banks can be refreshed, impacting test security, while increasing costs needed to support the test and taking up the test-taker’s valuable time. Our paper presents a novel explanatory item response theory (IRT) model, BERT-IRT, that has been used on the Duolingo English Test (DET), a high-stakes test of English, to reduce the length of pilots by a factor of 10. Our evaluation shows how the model uses BERT embeddings and engineered NLP features to accelerate item piloting without sacrificing criterion validity or reliability.

The popularization of large language models (LLMs) such as OpenAI’s GPT-3 and GPT-4 have led to numerous innovations in the field of AI in education. With respect to automated writing evaluation (AWE), LLMs have reduced challenges associated with assessing writing quality characteristics that are difficult to identify automatically, such as discourse coherence. In addition, LLMs can provide rationales for their evaluations (ratings) which increases score interpretability and transparency. This paper investigates one approach to producing ratings by training GPT-4 to assess discourse coherence in a manner consistent with expert human raters. The findings of the study suggest that GPT-4 has strong potential to produce discourse coherence ratings that are comparable to human ratings, accompanied by clear rationales. Furthermore, the GPT-4 ratings outperform traditional NLP coherence metrics with respect to agreement with human ratings. These results have implications for advancing AWE technology for learning and assessment.

Standard test sets for supervised learning evaluate in-distribution generalization. Unfortunately, when a dataset has systematic gaps (e.g., annotation artifacts), these evaluations are misleading: a model can learn simple decision rules that perform well on the test set but do not capture the abilities a dataset is intended to test. We propose a more rigorous annotation paradigm for NLP that helps to close systematic gaps in the test data. In particular, after a dataset is constructed, we recommend that the dataset authors manually perturb the test instances in small but meaningful ways that (typically) change the gold label, creating contrast sets. Contrast sets provide a local view of a model’s decision boundary, which can be used to more accurately evaluate a model’s true linguistic capabilities. We demonstrate the efficacy of contrast sets by creating them for 10 diverse NLP datasets (e.g., DROP reading comprehension, UD parsing, and IMDb sentiment analysis). Although our contrast sets are not explicitly adversarial, model performance is significantly lower on them than on the original test sets—up to 25% in some cases. We release our contrast sets as new evaluation benchmarks and encourage future dataset construction efforts to follow similar annotation processes.

Language models have emerged as a central component across NLP, and a great deal of progress depends on the ability to cheaply adapt them (e.g., through finetuning) to new domains and tasks. A language model’s vocabulary—typically selected before training and permanently fixed later—affects its size and is part of what makes it resistant to such adaptation. Prior work has used compositional input embeddings based on surface forms to ameliorate this issue. In this work, we go one step beyond and propose a fully compositional output embedding layer for language models, which is further grounded in information from a structured lexicon (WordNet), namely semantically related words and free-text definitions. To our knowledge, the result is the first word-level language model with a size that does not depend on the training vocabulary. We evaluate the model on conventional language modeling as well as challenging cross-domain settings with an open vocabulary, finding that it matches or outperforms previous state-of-the-art output embedding methods and adaptation approaches. Our analysis attributes the improvements to sample efficiency: our model is more accurate for low-frequency words.

Despite advances in dependency parsing, languages with small treebanks still present challenges. We assess recent approaches to multilingual contextual word representations (CWRs), and compare them for crosslingual transfer from a language with a large treebank to a language with a small or nonexistent treebank, by sharing parameters between languages in the parser itself. We experiment with a diverse selection of languages in both simulated and truly low-resource scenarios, and show that multilingual CWRs greatly facilitate low-resource dependency parsing even without crosslingual supervision such as dictionaries or parallel text. Furthermore, we examine the non-contextual part of the learned language models (which we call a “decontextual probe”) to demonstrate that polyglot language models better encode crosslingual lexical correspondence compared to aligned monolingual language models. This analysis provides further evidence that polyglot training is an effective approach to crosslingual transfer.

We introduce Rosita, a method to produce multilingual contextual word representations by training a single language model on text from multiple languages. Our method combines the advantages of contextual word representations with those of multilingual representation learning. We produce language models from dissimilar language pairs (English/Arabic and English/Chinese) and use them in dependency parsing, semantic role labeling, and named entity recognition, with comparisons to monolingual and non-contextual variants. Our results provide further evidence for the benefits of polyglot learning, in which representations are shared across multiple languages.

Previous approaches to multilingual semantic dependency parsing treat languages independently, without exploiting the similarities between semantic structures across languages. We experiment with a new approach where we combine resources from different languages in the CoNLL 2009 shared task to build a single polyglot semantic dependency parser. Notwithstanding the absence of parallel data, and the dissimilarity in annotations between languages, our approach results in improvement in parsing performance on several languages over a monolingual baseline. Analysis of the polyglot models’ performance provides a new understanding of the similarities and differences between languages in the shared task.