Ancient texts often lack punctuation marks, making it challenging to determine sentence boundaries and clause boundaries. Texts may contain sequences of hundreds of words without any period or indication of a full stop. Determining such boundaries is a crucial step in various NLP pipelines, especially regarding language models such as BERT that have context window constraints and regarding machine translation models which may become far less accurate when fed too much text at a time. In this paper, we consider several novel approaches to automatic segmentation of unpunctuated ancient texts into grammatically complete or semi-complete units. Our work here focuses on ancient and historical Hebrew and Aramaic texts, but the tools developed can be applied equally to similar languages. We explore several approaches to addressing this task: masked language models (MLM) to predict the next token; fewshot completions via an open-source foundational LLM; and the “Segment-Any-Text” (SaT) tool by Frohmann et al. (Frohmann et al., 2024). These are then compared to instructbased flows using commercial (closed, managed) LLMs, to be used as a benchmark. To evaluate these approaches, we also introduce a new ground truth (GT) dataset of manually segmented texts. We explore the performance of our different approaches on this dataset. We release both our segmentation tools and the dataset to support further research into computational processing and analysis of ancient texts, which can be found here ‘https://github.com/ERC-Midrash/rabbinic_chunker’.

The linguistic nature of Qumran Hebrew (QH) remains a central debate in the study of the Dead Sea Scrolls (DSS). Although some schol-ars view QH as an artificial imitation of Biblical Hebrew (BH), others argue that it represents a spoken dialect of ancient Judea. The present study employs computational lin-guistic techniques, clustering, classification, and machine learning, to analyze the relation-ship of QH with Biblical and Mishnaic He-brew. Preliminary findings confirm existing scholarly conclusions regarding the linguistic affinity of certain texts. This demonstrates that our methodology has a fundamental capacity to identify linguistic relationships. They also contribute new leads, on which we are now working to refine and enhance our analytical methods so as to provide founded insights into the historical development of Hebrew and the process of DSS textual composition.

We investigate the effect of duplicating multihead self-attention layers in large language models (LLMs) across a range of language tasks, with and without fine-tuning. The results demonstrate that duplicating the initial layers once or twice often yields a significant performance boost. Attention analysis uncovered the underlying mechanisms driving the improvement when performing layer duplication. This method enhances LLM capabilities with or without additional training or labeled data.

The digital era has made millions of manuscript images in Hebrew available to all. However, despite major advancements in handwritten text recognition over the past decade, an efficient pipeline for large scale and accurate conversion of these manuscripts into useful machine-readable form is still sorely lacking.We propose a pipeline that significantly improves recognition models for automatic transcription of Hebrew manuscripts. Transfer learning is used to fine-tune pretrained models. For post-recognition correction, it leverages text reuse, a common phenomenon in medieval manuscripts, and state-of-the-art large language models for medieval Hebrew.The framework successfully handles noisy transcriptions and consistently suggests alternate, better readings. Initial results show that word level accuracy increased by 10% for new readings proposed by text-reuse detection. Moreover, the character level accuracy improved by 18% by fine-tuning models on the first few pages of each manuscript.

The challenges of automated transliteration and code-switching–detection in Judeo-Arabic texts are addressed. We introduce two novel machine-learning models, one focused on transliterating Judeo-Arabic into Arabic, and another aimed at identifying non-Arabic words, predominantly Hebrew and Aramaic. Unlike prior work, our models are based on a bilingual Arabic-Hebrew language model, providing a unique advantage in capturing shared linguistic nuances. Evaluation results show that our models outperform prior solutions for the same tasks. As a practical contribution, we present a comprehensive pipeline capable of taking Judeo-Arabic text, identifying non-Arabic words, and then transliterating the Arabic portions into Arabic script. This work not only advances the state of the art but also offers a valuable toolset for making Judeo-Arabic texts more accessible to a broader Arabic-speaking audience.

We present a pipeline for a statistical stylometric exploration of a hypothesized partition of a text. Given a parameterization of the text, our pipeline: (1) detects literary features yielding the optimal overlap between the hypothesized and unsupervised partitions, (2) performs a hypothesis-testing analysis to quantify the statistical significance of the optimal overlap, while conserving implicit correlations between units of text that are more likely to be grouped, and (3) extracts and quantifies the importance of features most responsible for the classification, estimates their statistical stability and cluster-wise abundance. We apply our pipeline to the first two books in the Bible, where one stylistic component stands out in the eyes of biblical scholars, namely, the Priestly component. We identify and explore statistically significant stylistic differences between the Priestly and non-Priestly components.

We suggest a model for partial diacritization of deep orthographies. We focus on Arabic, where the optional indication of selected vowels by means of diacritics can resolve ambiguity and improve readability. Our partial diacritizer restores short vowels only when they contribute to the ease of understandability during reading a given running text. The idea is to identify those uncertainties of absent vowels that require the reader to look ahead to disambiguate. To achieve this, two independent neural networks are used for predicting diacritics, one that takes the entire sentence as input and another that considers only the text that has been read thus far. Partial diacritization is then determined by retaining precisely those vowels on which the two networks disagree, preferring the reading based on consideration of the whole sentence over the more naïve reading-order diacritization. For evaluation, we prepared a new dataset of Arabic texts with both full and partial vowelization. In addition to facilitating readability, we find that our partial diacritizer improves translation quality compared either to their total absence or to random selection. Lastly, we study the benefit of knowing the text that follows the word in focus toward the restoration of short vowels during reading, and we measure the degree to which lookahead contributes to resolving ambiguities encountered while reading. L’Herbelot had asserted, that the most ancient Korans, written in the Cufic character, had no vowel points; and that these were first invented by Jahia–ben Jamer, who died in the 127th year of the Hegira. “Toderini’s History of Turkish Literature,” Analytical Review (1789)

Natural language processing tools have been shown to be effective for detecting symptoms of schizophrenia in transcribed speech. We analyze and assess the contribution of the various syntactic and morphological categories towards successful machine classification of texts produced by subjects with schizophrenia and by others. Specifically, we fine-tune a language model for the classification task, and mask all words that are attributed with each category of interest. The speech samples were generated in a controlled way by interviewing inpatients who were officially diagnosed with schizophrenia, and a corresponding group of healthy controls. All participants are native Hebrew speakers. Our results show that nouns are the most significant category for classification performance.

Metaphor detection has been a challenging task in the NLP domain both before and after the emergence of transformer-based language models. The difficulty lies in subtle semantic nuances that are required to detect metaphor and in the scarcity of labeled data. We explore few-shot setups for metaphor detection, and also introduce new question answering data that can enhance classifiers that are trained on a small amount of data. We formulate the classification task as a question-answering one, and train a question-answering model. We perform extensive experiments for few shot on several architectures and report the results of several strong baselines. Thus, the answer to the question posed in the title is a definite “Yes!”

Midrash collections are complex rabbinic works that consist of text in multiple languages, that evolved through long processes of instable oral and written transmission. Determining the origin of a given passage in such a compilation is not always straightforward and is often a matter disputed by scholars, yet it is essential for scholars’ understanding of the passage and its relationship to other texts in the rabbinic corpus. To help solve this problem, we propose a system for classification of rabbinic literature based on its style, leveraging recently released pretrained Transformer models for Hebrew. Additionally, we demonstrate how our method can be applied to uncover lost material from the Midrash Tanhuma.

We trained a model to automatically transliterate Judeo-Arabic texts into Arabic script, enabling Arabic readers to access those writings. We employ a recurrent neural network (RNN), combined with the connectionist temporal classification (CTC) loss to deal with unequal input/output lengths. This obligates adjustments in the training data to avoid input sequences that are shorter than their corresponding outputs. We also utilize a pretraining stage with a different loss function to improve network converge. Since only a single source of parallel text was available for training, we take advantage of the possibility of generating data synthetically. We train a model that has the capability to memorize words in the output language, and that also utilizes context for distinguishing ambiguities in the transliteration. We obtain an improvement over the baseline 9.5% character error, achieving 2% error with our best configuration. To measure the contribution of context to learning, we also tested word-shuffled data, for which the error rises to 2.5%.

Natural language processing tools are used to automatically detect disturbances in transcribed speech of schizophrenia inpatients who speak Hebrew. We measure topic mutation over time and show that controls maintain more cohesive speech than inpatients. We also examine differences in how inpatients and controls use adjectives and adverbs to describe content words and show that the ones used by controls are more common than the those of inpatients. We provide experimental results and show their potential for automatically detecting schizophrenia in patients by means only of their speech patterns.

An implementation of a non-structural Example-Based Machine Translation system that translates sentences from Arabic to English, using a parallel corpus aligned at the sentence level, is described. Source-language synonyms were derived automatically and used to help locate potential translation examples for fragments of a given input sentence. The smaller the parallel corpus, the greater the contribution provided by synonyms. Considering the degree of relevance of the subject matter of a potential match contributes to the quality of the final results.