The performance of Machine Translation (MT) systems varies significantly with inputs of diverging features such as topics, genres, and surface properties. Though there are many MT evaluation metrics that generally correlate with human judgments, they are not directly useful in identifying specific shortcomings of MT systems. In this demo, we present a benchmarking interface that enables improved evaluation of specific MT systems in isolation or multiple MT systems collectively by quantitatively evaluating their performance on many tasks across multiple domains and evaluation metrics. Further, it facilitates effective debugging and error analysis of MT output via the use of dynamic filters that help users hone in on problem sentences with specific properties, such as genre, topic, sentence length, etc. The interface can be extended to include additional filters such as lexical, morphological, and syntactic features. Aside from helping debug MT output, it can also help in identifying problems in reference translations and evaluation metrics.
Emotion detection can provide us with a window into understanding human behavior. Due to the complex dynamics of human emotions, however, constructing annotated datasets to train automated models can be expensive. Thus, we explore the efficacy of cross-lingual approaches that would use data from a source language to build models for emotion detection in a target language. We compare three approaches, namely: i) using inherently multilingual models; ii) translating training data into the target language; and iii) using an automatically tagged parallel corpus. In our study, we consider English as the source language with Arabic and Spanish as target languages. We study the effectiveness of different classification models such as BERT and SVMs trained with different features. Our BERT-based monolingual models that are trained on target language data surpass state-of-the-art (SOTA) by 4% and 5% absolute Jaccard score for Arabic and Spanish respectively. Next, we show that using cross-lingual approaches with English data alone, we can achieve more than 90% and 80% relative effectiveness of the Arabic and Spanish BERT models respectively. Lastly, we use LIME to analyze the challenges of training cross-lingual models for different language pairs.
Arabic diacritic recovery is important for a variety of downstream tasks such as text-to-speech. In this paper, we introduce a new Gulf Arabic diacritization dataset composed of 19,850 words based on a subset of the Gumar corpus. We provide comprehensive set of guidelines for diacritization to enable the diacritization of more data. We also report on diacritization results based on the new corpus using a Hidden Markov Model and character-based sequence to sequence models.
NatiQ is end-to-end text-to-speech system for Arabic. Our speech synthesizer uses an encoder-decoder architecture with attention. We used both tacotron-based models (tacotron- 1 and tacotron-2) and the faster transformer model for generating mel-spectrograms from characters. We concatenated Tacotron1 with the WaveRNN vocoder, Tacotron2 with the WaveGlow vocoder and ESPnet transformer with the parallel wavegan vocoder to synthesize waveforms from the spectrograms. We used in-house speech data for two voices: 1) neu- tral male “Hamza”- narrating general content and news, and 2) expressive female “Amina”- narrating children story books to train our models. Our best systems achieve an aver- age Mean Opinion Score (MOS) of 4.21 and 4.40 for Amina and Hamza respectively.The objective evaluation of the systems using word and character error rate (WER and CER) as well as the response time measured by real- time factor favored the end-to-end architecture ESPnet. NatiQ demo is available online at https://tts.qcri.org.
Proper dialect identification is important for a variety of Arabic NLP applications. In this paper, we present a method for rapidly constructing a tweet dataset containing a wide range of country-level Arabic dialects —covering 18 different countries in the Middle East and North Africa region. Our method relies on applying multiple filters to identify users who belong to different countries based on their account descriptions and to eliminate tweets that either write mainly in Modern Standard Arabic or mostly use vulgar language. The resultant dataset contains 540k tweets from 2,525 users who are evenly distributed across 18 Arab countries. Using intrinsic evaluation, we show that the labels of a set of randomly selected tweets are 91.5% accurate. For extrinsic evaluation, we are able to build effective country level dialect identification on tweets with a macro-averaged F1-score of 60.6% across 18 classes.
Detecting offensive language on Twitter has many applications ranging from detecting/predicting bullying to measuring polarization. In this paper, we focus on building a large Arabic offensive tweet dataset. We introduce a method for building a dataset that is not biased by topic, dialect, or target. We produce the largest Arabic dataset to date with special tags for vulgarity and hate speech. We thoroughly analyze the dataset to determine which topics, dialects, and gender are most associated with offensive tweets and how Arabic speakers useoffensive language. Lastly, we conduct many experiments to produce strong results (F1 =83.2) on the dataset using SOTA techniques.
User stance detection entails ascertaining the position of a user towards a target, such as an entity, topic, or claim. Recent work that employs unsupervised classification has shown that performing stance detection on vocal Twitter users, who have many tweets on a target, can be highly accurate (+98%). However, such methods perform poorly or fail completely for less vocal users, who may have authored only a few tweets about a target. In this paper, we tackle stance detection for such users using two approaches. In the first approach, we improve user-level stance detection by representing tweets using contextualized embeddings, which capture latent meanings of words in context. We show that this approach outperforms two strong baselines and achieves 89.6% accuracy and 91.3% macro F-measure on eight controversial topics. In the second approach, we expand the tweets of a given user using their Twitter timeline tweets, which may not be topically relevant, and then we perform unsupervised classification of the user, which entails clustering a user with other users in the training set. This approach achieves 95.6% accuracy and 93.1% macro F-measure.
This system demonstration paper describes ASAD: Arabic Social media Analysis and unDerstanding, a suite of seven individual modules that allows users to determine dialects, sentiment, news category, offensiveness, hate speech, adult content, and spam in Arabic tweets. The suite is made available through a web API and a web interface where users can enter text or upload files.
With the emergence of the COVID-19 pandemic, the political and the medical aspects of disinformation merged as the problem got elevated to a whole new level to become the first global infodemic. Fighting this infodemic has been declared one of the most important focus areas of the World Health Organization, with dangers ranging from promoting fake cures, rumors, and conspiracy theories to spreading xenophobia and panic. Addressing the issue requires solving a number of challenging problems such as identifying messages containing claims, determining their check-worthiness and factuality, and their potential to do harm as well as the nature of that harm, to mention just a few. To address this gap, we release a large dataset of 16K manually annotated tweets for fine-grained disinformation analysis that (i) focuses on COVID-19, (ii) combines the perspectives and the interests of journalists, fact-checkers, social media platforms, policy makers, and society, and (iii) covers Arabic, Bulgarian, Dutch, and English. Finally, we show strong evaluation results using pretrained Transformers, thus confirming the practical utility of the dataset in monolingual vs. multilingual, and single task vs. multitask settings.
This paper provides an overview of the offensive language detection shared task at the 4th workshop on Open-Source Arabic Corpora and Processing Tools (OSACT4). There were two subtasks, namely: Subtask A, involving the detection of offensive language, which contains unacceptable or vulgar content in addition to any kind of explicit or implicit insults or attacks against individuals or groups; and Subtask B, involving the detection of hate speech, which contains insults or threats targeting a group based on their nationality, ethnicity, race, gender, political or sport affiliation, religious belief, or other common characteristics. In total, 40 teams signed up to participate in Subtask A, and 14 of them submitted test runs. For Subtask B, 33 teams signed up to participate and 13 of them submitted runs. We present and analyze all submissions in this paper.
During the last two decades, we have progressively turned to the Internet and social media to find news, entertain conversations and share opinion. Recently, OpenAI has developed a machine learning system called GPT-2 for Generative Pre-trained Transformer-2, which can produce deepfake texts. It can generate blocks of text based on brief writing prompts that look like they were written by humans, facilitating the spread false or auto-generated text. In line with this progress, and in order to counteract potential dangers, several methods have been proposed for detecting text written by these language models. In this paper, we propose a transfer learning based model that will be able to detect if an Arabic sentence is written by humans or automatically generated by bots. Our dataset is based on tweets from a previous work, which we have crawled and extended using the Twitter API. We used GPT2-Small-Arabic to generate fake Arabic Sentences. For evaluation, we compared different recurrent neural network (RNN) word embeddings based baseline models, namely: LSTM, BI-LSTM, GRU and BI-GRU, with a transformer-based model. Our new transfer-learning model has obtained an accuracy up to 98%. To the best of our knowledge, this work is the first study where ARABERT and GPT2 were combined to detect and classify the Arabic auto-generated texts.
Automatic categorization of short texts, such as news headlines and social media posts, has many applications ranging from content analysis to recommendation systems. In this paper, we use such text categorization i.e., labeling the social media posts to categories like ‘sports’, ‘politics’, ‘human-rights’ among others, to showcase the efficacy of models across different sources and varieties of Arabic. In doing so, we show that diversifying the training data, whether by using diverse training data for the specific task (an increase of 21% macro F1) or using diverse data to pre-train a BERT model (26% macro F1), leads to overall improvements in classification effectiveness. In our work, we also introduce two new Arabic text categorization datasets, where the first is composed of social media posts from a popular Arabic news channel that cover Twitter, Facebook, and YouTube, and the second is composed of tweets from popular Arabic accounts. The posts in the former are nearly exclusively authored in modern standard Arabic (MSA), while the tweets in the latter contain both MSA and dialectal Arabic.
Discovering the stances of media outlets and influential people on current, debatable topics is important for social statisticians and policy makers. Many supervised solutions exist for determining viewpoints, but manually annotating training data is costly. In this paper, we propose a cascaded method that uses unsupervised learning to ascertain the stance of Twitter users with respect to a polarizing topic by leveraging their retweet behavior; then, it uses supervised learning based on user labels to characterize both the general political leaning of online media and of popular Twitter users, as well as their stance with respect to the target polarizing topic. We evaluate the model by comparing its predictions to gold labels from the Media Bias/Fact Check website, achieving 82.6% accuracy.
Developing a platform that analyzes the content of curricula can help identify their shortcomings and whether they are tailored to specific desired outcomes. In this paper, we present a system to analyze Arabic curricula and provide insights into their content. It allows users to explore word presence, surface-forms used, as well as contrasting statistics between different countries from which the curricula were selected. Also, it provides a facility to grade text in reference to given grade-level and gives users feedback about the complexity or difficulty of words used in a text.
Arabic text is typically written without short vowels (or diacritics). However, their presence is required for properly verbalizing Arabic and is hence essential for applications such as text to speech. There are two types of diacritics, namely core-word diacritics and case-endings. Most previous works on automatic Arabic diacritic recovery rely on a large number of manually engineered features, particularly for case-endings. In this work, we present a unified character level sequence-to-sequence deep learning model that recovers both types of diacritics without the use of explicit feature engineering. Specifically, we employ a standard neural machine translation setup on overlapping windows of words (broken down into characters), and then we use voting to select the most likely diacritized form of a word. The proposed model outperforms all previous state-of-the-art systems. Our best settings achieve a word error rate (WER) of 4.49% compared to the state-of-the-art of 12.25% on a standard dataset.
When speakers code-switch between their native language and a second language or language variant, they follow a syntactic pattern where words and phrases from the embedded language are inserted into the matrix language. This paper explores the possibility of utilizing this pattern in improving code-switching identification between Modern Standard Arabic (MSA) and Egyptian Arabic (EA). We try to answer the question of how strong is the POS signal in word-level code-switching identification. We build a deep learning model enriched with linguistic features (including POS tags) that outperforms the state-of-the-art results by 1.9% on the development set and 1.0% on the test set. We also show that in intra-sentential code-switching, the selection of lexical items is constrained by POS categories, where function words tend to come more often from the dialectal language while the majority of content words come from the standard language.
This paper describes the QC-GO team submission to the MADAR Shared Task Subtask 1 (travel domain dialect identification) and Subtask 2 (Twitter user location identification). In our participation in both subtasks, we explored a number of approaches and system combinations to obtain the best performance for both tasks. These include deep neural nets and heuristics. Since individual approaches suffer from various shortcomings, the combination of different approaches was able to fill some of these gaps. Our system achieves F1-Scores of 66.1% and 67.0% on the development sets for Subtasks 1 and 2 respectively.
Short vowels, aka diacritics, are more often omitted when writing different varieties of Arabic including Modern Standard Arabic (MSA), Classical Arabic (CA), and Dialectal Arabic (DA). However, diacritics are required to properly pronounce words, which makes diacritic restoration (a.k.a. diacritization) essential for language learning and text-to-speech applications. In this paper, we present a system for diacritizing MSA, CA, and two varieties of DA, namely Moroccan and Tunisian. The system uses a character level sequence-to-sequence deep learning model that requires no feature engineering and beats all previous SOTA systems for all the Arabic varieties that we test on.
We introduce Tanbih, a news aggregator with intelligent analysis tools to help readers understanding what’s behind a news story. Our system displays news grouped into events and generates media profiles that show the general factuality of reporting, the degree of propagandistic content, hyper-partisanship, leading political ideology, general frame of reporting, and stance with respect to various claims and topics of a news outlet. In addition, we automatically analyse each article to detect whether it is propagandistic and to determine its stance with respect to a number of controversial topics.
In this paper, we present a new and fast state-of-the-art Arabic diacritizer that guesses the diacritics of words and then their case endings. We employ a Viterbi decoder at word-level with back-off to stem, morphological patterns, and transliteration and sequence labeling based diacritization of named entities. For case endings, we use Support Vector Machine (SVM) based ranking coupled with morphological patterns and linguistic rules to properly guess case endings. We achieve a low word level diacritization error of 3.29% and 12.77% without and with case endings respectively on a new multi-genre free of copyright test set. We are making the diacritizer available for free for research purposes.
The automated processing of Arabic Dialects is challenging due to the lack of spelling standards and to the scarcity of annotated data and resources in general. Segmentation of words into its constituent parts is an important processing building block. In this paper, we show how a segmenter can be trained using only 350 annotated tweets using neural networks without any normalization or use of lexical features or lexical resources. We deal with segmentation as a sequence labeling problem at the character level. We show experimentally that our model can rival state-of-the-art methods that rely on additional resources.
This paper focuses on comparing between using Support Vector Machine based ranking (SVM-Rank) and Bidirectional Long-Short-Term-Memory (bi-LSTM) neural-network based sequence labeling in building a state-of-the-art Arabic part-of-speech tagging system. Using SVM-Rank leads to state-of-the-art results, but with a fair amount of feature engineering. Using bi-LSTM, particularly when combined with word embeddings, may lead to competitive POS-tagging results by automatically deducing latent linguistic features. However, we show that augmenting bi-LSTM sequence labeling with some of the features that we used for the SVM-Rank based tagger yields to further improvements. We also show that gains that realized by using embeddings may not be additive with the gains achieved by the features. We are open-sourcing both the SVM-Rank and the bi-LSTM based systems for free.
In this paper, we present our work on detecting abusive language on Arabic social media. We extract a list of obscene words and hashtags using common patterns used in offensive and rude communications. We also classify Twitter users according to whether they use any of these words or not in their tweets. We expand the list of obscene words using this classification, and we report results on a newly created dataset of classified Arabic tweets (obscene, offensive, and clean). We make this dataset freely available for research, in addition to the list of obscene words and hashtags. We are also publicly releasing a large corpus of classified user comments that were deleted from a popular Arabic news site due to violations the site’s rules and guidelines.
Arabic dialects do not just share a common koiné, but there are shared pan-dialectal linguistic phenomena that allow computational models for dialects to learn from each other. In this paper we build a unified segmentation model where the training data for different dialects are combined and a single model is trained. The model yields higher accuracies than dialect-specific models, eliminating the need for dialect identification before segmentation. We also measure the degree of relatedness between four major Arabic dialects by testing how a segmentation model trained on one dialect performs on the other dialects. We found that linguistic relatedness is contingent with geographical proximity. In our experiments we use SVM-based ranking and bi-LSTM-CRF sequence labeling.
In this paper, we present Farasa (meaning insight in Arabic), which is a fast and accurate Arabic segmenter. Segmentation involves breaking Arabic words into their constituent clitics. Our approach is based on SVMrank using linear kernels. The features that we utilized account for: likelihood of stems, prefixes, suffixes, and their combination; presence in lexicons containing valid stems and named entities; and underlying stem templates. Farasa outperforms or equalizes state-of-the-art Arabic segmenters, namely QATARA and MADAMIRA. Meanwhile, Farasa is nearly one order of magnitude faster than QATARA and two orders of magnitude faster than MADAMIRA. The segmenter should be able to process one billion words in less than 5 hours. Farasa is written entirely in native Java, with no external dependencies, and is open-source.
The paper describes the QCRI submissions to the task of automatic Arabic dialect classification into 5 Arabic variants, namely Egyptian, Gulf, Levantine, North-African, and Modern Standard Arabic (MSA). The training data is relatively small and is automatically generated from an ASR system. To avoid over-fitting on such small data, we carefully selected and designed the features to capture the morphological essence of the different dialects. We submitted four runs to the Arabic sub-task. For all runs, we used a combined feature vector of character bi-grams, tri-grams, 4-grams, and 5-grams. We tried several machine-learning algorithms, namely Logistic Regression, Naive Bayes, Neural Networks, and Support Vector Machines (SVM) with linear and string kernels. However, our submitted runs used SVM with a linear kernel. In the closed submission, we got the best accuracy of 0.5136 and the third best weighted F1 score, with a difference less than 0.002 from the highest score.
Despite many recent papers on Arabic Named Entity Recognition (NER) in the news domain, little work has been done on microblog NER. NER on microblogs presents many complications such as informality of language, shortened named entities, brevity of expressions, and inconsistent capitalization (for cased languages). We introduce simple effective language-independent approaches for improving NER on microblogs, based on using large gazetteers, domain adaptation, and a two-pass semi-supervised method. We use Arabic as an example language to compare the relative effectiveness of the approaches and when best to use them. We also present a new dataset for the task. Results of combining the proposed approaches show an improvement of 35.3 F-measure points over a baseline system trained on news data and an improvement of 19.9 F-measure points over the same system but trained on microblog data.
This paper presents an end-to-end automatic processing system for Arabic. The system performs: correction of common spelling errors pertaining to different forms of alef, ta marbouta and ha, and alef maqsoura and ya; context sensitive word segmentation into underlying clitics, POS tagging, and gender and number tagging of nouns and adjectives. We introduce the use of stem templates as a feature to improve POS tagging by 0.5\% and to help ascertain the gender and number of nouns and adjectives. For gender and number tagging, we report accuracies that are significantly higher on previously unseen words compared to a state-of-the-art system.
In this paper we present an algorithm for automatic extraction of textual elements, namely titles and full text, associated with news stories in news web pages. We propose a supervised machine learning classification technique based on the use of a Support Vector Machine (SVM) classifier to extract the desired textual elements. The technique uses internal structural features of a webpage without relying on the Document Object Model to which many content authors fail to adhere. The classifier uses a set of features which rely on the length of text, the percentage of hypertext, etc. The resulting classifier is nearly perfect on previously unseen news pages from different sites. The proposed technique is successfully employed in Alzoa.com, which is the largest Arabic news aggregator on the web.
This paper describes the development of an Arabic document image collection containing 34,651 documents from 1,378 different books and 25 topics with their relevance judgments. The books from which the collection is obtained are a part of a larger collection 75,000 books being scanned for archival and retrieval at the bibliotheca Alexandrina (BA). The documents in the collection vary widely in topics, fonts, and degradation levels. Initial baseline experiments were performed to examine the effectiveness of different index terms, with and without blind relevance feedback, on Arabic OCR degraded text.