Text-to-SQL semantic parsing faces challenges in generalizing to cross-domain and complex queries. Recent research has employed a question decomposition strategy to enhance the parsing of complex SQL queries.However, this strategy encounters two major obstacles: (1) existing datasets lack question decomposition; (2) due to the syntactic complexity of SQL, most complex queries cannot be disentangled into sub-queries that can be readily recomposed. To address these challenges, we propose a new modular Query Plan Language (QPL) that systematically decomposes SQL queries into simple and regular sub-queries. We develop a translator from SQL to QPL by leveraging analysis of SQL server query optimization plans, and we augment the Spider dataset with QPL programs. Experimental results demonstrate that the modular nature of QPL benefits existing semantic-parsing architectures, and training text-to-QPL parsers is more effective than text-to-SQL parsing for semantically equivalent queries. The QPL approach offers two additional advantages: (1) QPL programs can be paraphrased as simple questions, which allows us to create a dataset of (complex question, decomposed questions). Training on this dataset, we obtain a Question Decomposer for data retrieval that is sensitive to database schemas. (2) QPL is more accessible to non-experts for complex queries, leading to more interpretable output from the semantic parser.

We call into question the recently popularized method of direct model editing as a means of correcting factual errors in LLM generations. We contrast model editing with three similar but distinct approaches that pursue better defined objectives: (1) retrieval-based architectures, which decouple factual memory from inference and linguistic capabilities embodied in LLMs; (2) concept erasure methods, which aim at preventing systemic bias in generated text; and (3) attribution methods, which aim at grounding generations into identified textual sources. We argue that direct model editing cannot be trusted as a systematic remedy for the disadvantages inherent to LLMs, and while it has proven potential in improving model explainability, it opens risks by reinforcing the notion that models can be trusted for factuality. We call for cautious promotion and application of model editing as part of the LLM deployment process, and for responsibly limiting the use cases of LLMs to those not relying on editing as a critical component.

We study cross-lingual UMLS named entity linking, where mentions in a given source language are mapped to UMLS concepts, most of which are labeled in English. Our cross-lingual framework includes an offline unsupervised construction of a translated UMLS dictionary and a per-document pipeline which identifies UMLS candidate mentions and uses a fine-tuned pretrained transformer language model to filter candidates according to context. Our method exploits a small dataset of manually annotated UMLS mentions in the source language and uses this supervised data in two ways: to extend the unsupervised UMLS dictionary and to fine-tune the contextual filtering of candidate mentions in full documents. We demonstrate results of our approach on both Hebrew and English. We achieve new state-of-the-art (SOTA) results on the Hebrew Camoni corpus, +8.9 F1 on average across three communities in the dataset. We also achieve new SOTA on the English dataset MedMentions with +7.3 F1.

Masked language modeling (MLM) is one of the key sub-tasks in vision-language pretraining. In the cross-modal setting, tokens in the sentence are masked at random, and the model predicts the masked tokens given the image and the text. In this paper, we observe several key disadvantages of MLM in this setting. First, as captions tend to be short, in a third of the sentences no token is sampled. Second, the majority of masked tokens are stop-words and punctuation, leading to under-utilization of the image. We investigate a range of alternative masking strategies specific to the cross-modal setting that address these shortcomings, aiming for better fusion of text and image in the learned representation. When pre-training the LXMERT model, our alternative masking strategies consistently improve over the original masking strategy on three downstream tasks, especially in low resource settings. Further, our pre-training approach substantially outperforms the baseline model on a prompt-based probing task designed to elicit image objects. These results and our analysis indicate that our method allows for better utilization of the training data.

Recent works have shown that supervised models often exploit data artifacts to achieve good test scores while their performance severely degrades on samples outside their training distribution. Contrast sets (Gardneret al., 2020) quantify this phenomenon by perturbing test samples in a minimal way such that the output label is modified. While most contrast sets were created manually, requiring intensive annotation effort, we present a novel method which leverages rich semantic input representation to automatically generate contrast sets for the visual question answering task. Our method computes the answer of perturbed questions, thus vastly reducing annotation cost and enabling thorough evaluation of models’ performance on various semantic aspects (e.g., spatial or relational reasoning). We demonstrate the effectiveness of our approach on the GQA dataset and its semantic scene graph image representation. We find that, despite GQA’s compositionality and carefully balanced label distribution, two high-performing models drop 13-17% in accuracy compared to the original test set. Finally, we show that our automatic perturbation can be applied to the training set to mitigate the degradation in performance, opening the door to more robust models.

Data-to-text generation systems are trained on large datasets, such as WebNLG, Ro-toWire, E2E or DART. Beyond traditional token-overlap evaluation metrics (BLEU or METEOR), a key concern faced by recent generators is to control the factuality of the generated text with respect to the input data specification. We report on our experience when developing an automatic factuality evaluation system for data-to-text generation that we are testing on WebNLG and E2E data. We aim to prepare gold data annotated manually to identify cases where the text communicates more information than is warranted based on the in-put data (extra) or fails to communicate data that is part of the input (missing). While analyzing reference (data, text) samples, we encountered a range of systematic uncertainties that are related to cases on implicit phenomena in text, and the nature of non-linguistic knowledge we expect to be involved when assessing factuality. We derive from our experience a set of evaluation guidelines to reach high inter-annotator agreement on such cases.

We present a semantic role labeling resource for Hebrew built semi-automatically through annotation projection from English. This corpus is derived from the multilingual OpenSubtitles dataset and includes short informal sentences, for which reliable linguistic annotations have been computed. We provide a fully annotated version of the data including morphological analysis, dependency syntax and semantic role labeling in both FrameNet and ProbBank styles. Sentences are aligned between English and Hebrew, both sides include full annotations and the explicit mapping from the English arguments to the Hebrew ones. We train a neural SRL model on this Hebrew resource exploiting the pre-trained multilingual BERT transformer model, and provide the first available baseline model for Hebrew SRL as a reference point. The code we provide is generic and can be adapted to other languages to bootstrap SRL resources.

Recent work in the field of automatic summarization and headline generation focuses on maximizing ROUGE scores for various news datasets. We present an alternative, extrinsic, evaluation metric for this task, Answering Performance for Evaluation of Summaries. APES utilizes recent progress in the field of reading-comprehension to quantify the ability of a summary to answer a set of manually created questions regarding central entities in the source article. We first analyze the strength of this metric by comparing it to known manual evaluation metrics. We then present an end-to-end neural abstractive model that maximizes APES, while increasing ROUGE scores to competitive results.

We present the Hebrew FrameNet project, describe the development and annotation processes and enumerate the challenges we faced along the way. We have developed semi-automatic tools to help speed the annotation and data collection process. The resource currently covers 167 frames, 3,000 lexical units and about 500 fully annotated sentences. We have started training and testing automatic SRL tools on the seed data.

We report on an effort to build a corpus of Modern Hebrew tagged with part-of-speech and morphology. We designed a tagset specific to Hebrew while focusing on four aspects: the tagset should be consistent with common linguistic knowledge; there should be maximal agreement among taggers as to the tags assigned to maintain consistency; the tagset should be useful for machine taggers and learning algorithms; and the tagset should be effective for applications relying on the tags as input features. In this paper, we illustrate these issues by explaining our decision to introduce a tag for beinoni forms in Hebrew. We explain how this tag is defined, and how it helped us improve manual tagging accuracy to a high-level, while improving automatic tagging and helping in the task of syntactic chunking.