Michael Elhadad

Large Language Models (LLMs) have spurred progress in text-to-SQL, the task of generating SQL queries from natural language questions based on a given database schema. Despite the declarative nature of SQL, it continues to be a complex programming language. In this paper, we investigate the potential of an alternative query language with simpler syntax and modular specification of complex queries. The purpose is to create a query language that can be learned more easily by modern neural semantic parsing architectures while also enabling non-programmers to better assess the validity of the query plans produced by an interactive query plan assistant. The proposed alternative query language is called Query Plan Language (QPL). It is designed to be modular and can be translated into a restricted form of SQL Common Table Expressions (CTEs). The aim of QPL is to make complex data retrieval accessible to non-programmers by allowing users to express their questions in natural language while also providing an easier-to-verify target language. The paper demonstrates how neural LLMs can benefit from QPL's modularity to generate complex query plans in a compositional manner. This involves a question decomposition strategy and a planning stage. We conduct experiments on a version of the Spider text-to-SQL dataset that has been converted to QPL. The hierarchical structure of QPL programs enables us to measure query complexity naturally. Based on this assessment, we identify the low accuracy of existing text-to-SQL systems on complex compositional queries. We present ways to address the challenge of complex queries in an iterative, user-controlled manner, using fine-tuned LLMs and a variety of prompting strategies in a compositional manner.

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.

Yuval Pinter, Michael Elhadad. Findings of the Association for Computational Linguistics: EMNLP 2023. 2023.

Objective: In Hebrew online health communities, participants commonly write medical terms that appear as transliterated forms of a source term in English. Such transliterations introduce high variability in text and challenge text-analytics methods. To reduce their variability, medical terms must be normalized, such as linking them to Unified Medical Language System (UMLS) concepts. We present a method to identify both transliterated and translated Hebrew medical terms and link them with UMLS entities. Materials and Methods: We investigate the effect of linking terms in Camoni, a popular Israeli online health community in Hebrew. Our method, MDTEL (Medical Deep Transliteration Entity Linking), includes (1) an attention-based recurrent neural network encoder-decoder to transliterate words and mapping UMLS from English to Hebrew, (2) an unsupervised method for creating a transliteration dataset in any language without manually labeled data, and (3) an efficient way to identify and link medical entities in the Hebrew corpus to UMLS concepts, by producing a high-recall list of candidate medical terms in the corpus, and then filtering the candidates to relevant medical terms. Results: We carry out experiments on 3 disease-specific communities: diabetes, multiple sclerosis, and depression. MDTEL tagging and normalizing on Camoni posts achieved 99% accuracy, 92% recall, and 87% precision. When tagging and normalizing terms in queries from the Camoni search logs, UMLS-normalized queries improved search results in 46% of the cases. Conclusions: Cross-lingual UMLS entity linking from Hebrew is possible and improves search performance across communities. Annotated datasets, annotation guidelines, and code are made available online (https://github.com/yonatanbitton/mdtel).

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 PropBank 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.

In a global setting, texts contain transliterated names from many cultural origins. Correct transliteration depends not only on target and source languages but also, on the source language of the name. We introduce a novel methodology for transliteration of names originating in different languages using only monolingual resources. Our method is based on a step of noisy transliteration and then ranking of the results based on origin specific letter models. The transliteration table used for noisy generation is learned in an unsupervised manner for each possible origin language. We present a solution for gathering monolingual training data used by our method by mining of social media sites such as Facebook and Wikipedia. We present results in the context of transliterating from English to Hebrew and provide an online web service for transliteration from English to Hebrew

Query Focused Summarization (QFS) has been addressed mostly using extractive methods. Such methods, however, produce text which suffers from low coherence. We investigate how abstractive methods can be applied to QFS, to overcome such limitations. Recent developments in neural-attention based sequence-to-sequence models have led to state-of-the-art results on the task of abstractive generic single document summarization. Such models are trained in an end to end method on large amounts of training data. We address three aspects to make abstractive summarization applicable to QFS: (a)since there is no training data, we incorporate query relevance into a pre-trained abstractive model; (b) since existing abstractive models are trained in a single-document setting, we design an iterated method to embed abstractive models within the multi-document requirement of QFS; (c) the abstractive models we adapt are trained to generate text of specific length (about 100 words), while we aim at generating output of a different size (about 250 words); we design a way to adapt the target size of the generated summaries to a given size ratio. We compare our method (Relevance Sensitive Attention for QFS) to extractive baselines and with various ways to combine abstractive models on the DUC QFS datasets and demonstrate solid improvements on ROUGE performance.