Thailand

Large language models (LLMs) play a crucial role in a wide range of real world applications. However, concerns about their safety and ethical implications are growing. While research on LLM safety is expanding, there is a noticeable gap in evaluating safety across multiple languages, especially in Arabic and Russian. We address this gap by exploring biases in LLMs across different languages and contexts, focusing on GPT-3.5 and Gemini. Through carefully designed argument-based prompts and scenarios in Arabic, English, and Russian, we examine biases in cultural, political, racial, religious, and gender domains. Our findings reveal biases in these domains. In particular, our investigation uncovers subtle biases where each model tends to present winners as those speaking the primary language the model is prompted with. Our study contributes to ongoing efforts to ensure justice and equality in LLM development and emphasizes the importance of further research towards responsible progress in this field.

ACL 2024

John vs. Ahmed: Debate-Induced Bias in Multilingual LLMs

debate

llms

religion

culture

race

politics

gender

safety

bias

workshop paper

### Welcome!
The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. Our Virtual Poster Sessions will take place online Thursday, August 22, 2024.

You are required to register for this event. **Please register [here](https://2024.aclweb.org/registration). **

If you have already registered, please check your inbox for an email from Underline granting you access to ACL 2024 content.

Please register!

The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. More information will be announced soon.

Arabic Optical Character Recognition (OCR) and Handwriting Recognition (HWR) pose unique challenges due to the cursive and context-sensitive nature of the Arabic script. This study introduces ***Qalam***, a novel foundation model designed for Arabic OCR and HWR, built on a SwinV2 encoder and RoBERTa decoder architecture. Our model significantly outperforms existing methods, achieving a Word Error Rate (WER) of just 0.80% in HWR tasks and 1.18% in OCR tasks. We train ***Qalam*** on a diverse dataset, including over 4.5 million images from Arabic manuscripts and a synthetic dataset comprising 60k image-text pairs. Notably, ***Qalam*** demonstrates exceptional handling of Arabic diacritics, a critical feature in Arabic scripts. Furthermore, it shows a remarkable ability to process high-resolution inputs, addressing a common limitation in current OCR systems. These advancements underscore ***Qalam***'s potential as a leading solution for Arabic script recognition, offering a significant leap in accuracy and efficiency.

Qalam: A Multimodal LLM for Arabic Optical Character and Handwriting Recognition

The rapid advancements in Large Language Models (LLMs) have led to significant improvements in various natural language processing tasks. However, the evaluation of LLMs’ legal knowledge, particularly in non English languages such as Arabic, remains under-explored. To address this gap, we introduce ArabLegalEval, a multitask benchmark dataset for assessing the Arabic legal knowledge of LLMs. Inspired by the MMLU and LegalBench datasets, ArabLegalEval consists of multiple tasks sourced from Saudi legal documents and synthesized questions. In this work, we aim to analyze the capabilities required to solve legal problems in Arabic and benchmark the performance of state-of-the-art LLMs. We explore the impact of in-context learning on performance and investigate various evaluation methods. Additionally, we explore workflows for automatically generating questions with automatic validation to enhance the dataset’s quality. By releasing ArabLegalEval and our code, we hope to accelerate AI research in the Arabic Legal domain

ArabLegalEval: A Multitask Benchmark for Assessing Arabic Legal Knowledge in Large Language Models

Tashkeel, or Arabic Text Diacritization (ATD), greatly enhances the comprehension of Arabic text by removing ambiguity and minimizing the risk of misinterpretations caused by its absence. It plays a crucial role in improving Arabic text processing, particularly in applications such as text-to-speech and machine translation. This paper introduces a new approach to training ATD models. First, we finetuned two transformers, encoder-only and encoder-decoder, that were initialized from a pretrained character-based BERT. Then, we applied the Noisy-Student approach to boost the performance of the best model. We evaluated our models alongside 11 commercial and open-source models using two manually labeled benchmark datasets: WikiNews and our CATT dataset. Our findings show that our top model surpasses all evaluated models by relative Diacritic Error Rates (DERs) of 30.83% and 35.21% on WikiNews and CATT, respectively, achieving state-of-the-art in ATD. In addition, we show that our model outperforms GPT-4-turbo on CATT dataset by a relative DER of 9.36%. We open-source our CATT models and benchmark dataset for the research community footnote{ https://github.com/abjadai/catt }.

CATT: Character-based Arabic Tashkeel Transformer

Learning morphophonological mappings between the spoken form of a language and its underlying morphological structures is crucial for enriching resources for morphologically rich languages like Arabic. In this work, we focus on Egyptian Arabic as our case study and explore the integration of linguistic knowledge with a neural transformer model. Our approach involves learning to correct the residual errors from hand-crafted rules to predict the spoken form from a given underlying morphological representation. We demonstrate that using a minimal set of rules, we can effectively recover errors even in very low-resource settings.

Picking Up Where the Linguist Left Off: Mapping Morphology to Phonology through Learning the Residuals

Development of pre-trained language models has predominantly relied on large amounts of datasets. However, this dependence on abundant data has limited the applicability of these models in low-resource settings. In this work, we investigate the utility of exploiting synthetic datasets acquired from different sources to pre-train language models for Arabic. Namely, we leverage data derived based on four different methods: optical character recognition (OCR), automatic speech recognition (ASR), machine translation (MT), and generative language models. We use these datasets to pre-train models in three different architectures: encoder-only (BERTtextsubscript{Base}), encoder-decoder (T5), and decoder-only (GPT-2). We test the capabilities of resulting models on Arabic natural language understanding (NLU) tasks using the ORCA benchmark. Our results show that utilizing synthetic data can achieve performance comparable to, or even surpassing, those trained on gold data. For example, our model based on a GPT-2 architecture trained on a combined synthetic dataset surpasses the baseline model ARBERTtextsubscript{v2}. Overall, our models pre-trained on synthetic data demonstrate robust performance across various tasks. This highlights the potential of synthetic datasets in augmenting language model training in low-resource settings.

On the Utility of Pretraining Language Models on Synthetic Data

Open-sourced large language models (LLMs) have exhibited remarkable performance in a variety of NLP tasks, often catching up with the closed-sourced LLMs like ChatGPT. Among these open LLMs, LLaMA-3-70B has emerged as the most recent and the most prominent one. However, how LLaMA-3-70B would situate itself in multilingual settings, especially in a rich morphological language like Arabic, has yet to be explored. In this work, we focus to bridge this gap by evaluating LLaMA-3-70B on a diverse set of Arabic natural language generation (NLG) benchmarks. To the best of our knowledge, this is the first study that comprehensively evaluates LLaMA-3-70B on tasks related to Arabic natural language generation. Our study reveals that LLaMA-3-70B lags behind the closed LLMs like ChatGPT, both in modern standard Arabic (MSA) and dialectal Arabic (DA). We further compare the performance of LLaMA-3-70B with our smaller and dedicated finetuned Arabic models. We find that both LLaMA-3-70B and ChatGPT are outperformed by comparatively smaller dedicated Arabic models, indicating the scope for potential improvement with Arabic-focused LLMs.

Benchmarking LLaMA-3 on Arabic Language Generation Tasks

Event-argument extraction is a challenging task, particularly in Arabic due to sparse linguistic resources. To fill this gap, we introduce the hadath corpus ($550$k tokens) as an extension of Wojood, enriched with event-argument annotations. We used three types of event arguments: $agent$, $location$, and $date$, which we annotated as relation types. Our inter-annotator agreement evaluation resulted in $82.23%$ $Kappa$ score and $87.2%$ $F_1$-score. Additionally, we propose a novel method for event relation extraction using BERT, in which we treat the task as text entailment. This method achieves an $F_1$-score of $94.01%$. To further evaluate the generalization of our proposed method, we collected and annotated another out-of-domain corpus (about $80$k tokens) called testNLI and used it as a second test set, on which our approach achieved promising results ($83.59%$ $F_1$-score). Last but not least, we propose an end-to-end system for event-arguments extraction. This system is implemented as part of SinaTools, and both corpora are publicly available at {small url{https://sina.birzeit.edu/wojood}}

Event-Arguments Extraction Corpus and Modeling using BERT for Arabic

Recent advancements have significantly enhanced the capabilities of Multimodal Large Language Models (MLLMs) in generating and understanding image-to-text content. Despite these successes, progress is predominantly limited to English due to the scarcity of high-quality multimodal resources in other languages. This limitation impedes the development of competitive models in languages such as Arabic. To alleviate this situation, we introduce an efficient Arabic multimodal assistant, dubbed ***Dallah***, that utilizes an advanced language model based on LLaMA-2 to facilitate multimodal interactions. ***Dallah*** demonstrates state-of-the-art performance in Arabic MLLMs. Through fine-tuning six Arabic dialects, ***Dallah*** showcases its capability to handle complex dialectal interactions incorporating both textual and visual elements. The model excels in two benchmark tests: one evaluating its performance on Modern Standard Arabic (MSA) and another specifically designed to assess dialectal responses. Beyond its robust performance in multimodal interaction tasks, ***Dallah*** has the potential to pave the way for further development of dialect-aware Arabic MLLMs.

Dallah: A Dialect-Aware Multimodal Large Language Model for Arabic

Automated Essay Scoring (AES) has emerged as a significant research problem within natural language processing, providing valuable support for educators in assessing student writing skills. In this paper, we introduce QAES, the first publicly available trait-specific annotations for Arabic AES, built on the Qatari Corpus of Argumentative Writing (QCAW). QAES includes a diverse collection of essays in Arabic, each of them annotated with holistic and trait-specific scores, including relevance, organization, vocabulary, style, development, mechanics, and grammar. In total, it comprises 195 Arabic essays (with lengths ranging from 239 to 806 words) across two distinct argumentative writing tasks. We benchmark our dataset against the state-of-the-art English baselines and a feature-based approach. In addition, we discuss the adopted guidelines and the challenges encountered during the annotation process. Finally, we provide insights into potential areas for improvement and future directions in Arabic AES research.

QAES: First Publicly-Available Trait-Specific Annotations for Automated Scoring of Arabic Essays

Text classification is of paramount importance in a wide range of applications, including information retrieval, extraction and sentiment analysis. The challenge of classifying and labelling text genres, especially in web-based corpora, has received considerable attention. The frequent absence of unambiguous genre information complicates the identification of text types. To address these issues, the Functional Text Dimensions (FTD) method has been introduced to provide a universal set of categories for text classification. This study presents the Arabic Functional Text Dimensions Corpus (AFTD Corpus), a carefully curated collection of documents for evaluating text classification in Arabic. The AFTD Corpus which we are making available to the community, consists of 3400 documents spanning 17 different class categories. Through a comprehensive evaluation using traditional machine learning and neural models, we assess the effectiveness of the FTD approach in the Arabic context. CAMeLBERT, a state-of-the-art model, achieved an impressive F1 score of 0.81 on our corpus. This research highlights the potential of the FTD method for improving text classification, especially for Arabic content, and underlines the importance of robust classification models in web applications.

Premium content

John vs. Ahmed: Debate-Induced Bias in Multilingual LLMs

Next from ACL 2024

Qalam: A Multimodal LLM for Arabic Optical Character and Handwriting Recognition

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES