Adobe Prelude was an ingest and logging software application for tagging media with metadata for searching, post-production workflows, and footage lifecycle management. Adobe Prelude is also made to work closely with Adobe Premiere Pro. It is part of the Adobe Creative Cloud and is geared towards professional video editing alone or with a group. The software also offers features like rough cut creation. A speech transcription feature was removed in December 2014. == History == Adobe announced that on April 23, 2012 Adobe OnLocation would be shut down and Adobe Prelude would launch on May 7, 2012. Adobe stated OnLocation's production was stopping because of the growing trend in the industry toward tapeless, native workflows, Adobe stresses that Adobe Prelude is not a direct replacement for OnLocation. Adobe OnLocation was available in CS5 but not in CS6 and Adobe Prelude is only available in CS6. Adobe still offers technical support for OnLocation. In 2021, Adobe announced they would be discontinuing Adobe Prelude, starting by removing it from their website on September 8, 2021. Support for existing users will continue through September 8, 2024. == Features == Prelude is used to tag media, log data, create and export metadata and generate rough cuts that can be sent to Adobe Premiere Pro. A user can add a tag to a piece of media that will show up on Premiere Pro or if another user opens that media with Prelude. Ingest Footage Prelude can ingest all kinds of file types. Once ingested, Prelude can duplicate, transcode and verify the files. Log Footage Prelude can log data only using the keyboard. Create Rough Cuts Prelude is able to generate Rough Cuts. Rough Cuts are a combination of sub clips that will hold any metadata a user feeds into it. Rough cuts can hold metadata such as markers and comments, and this metadata will stay on this footage. Workflow Accessibility Prelude is an XMP - based open platform that allows for custom integration into many video editing platforms. == Features from OnLocation == Many features from Adobe OnLocation went to Adobe Prelude or Adobe Premiere Pro. Adobe OnLocation thrived on tape - based cameras and setting up a shot before shooting it, with the change in the industry, this problem is irrelevant in post production. Adobe OnLocation also allowed the user to add tags and scripting metadata that would carry over to Premiere Pro. OnLocation also had a Media Browser pane, which is the standard for any Adobe program today, Prelude has this Media Browser as well. == Prelude Live Logger == Prelude Live Logger is an application integrated with Prelude CC. Prelude Live Logger is designed to capture notes to use during video logging and editing while you shoot footage on an iPad's camera. Editors can import and combine this metadata with footage from Prelude throughout editing to facilitate various tasks.
Audio inpainting
Audio inpainting (also known as audio interpolation) is an audio restoration task which deals with the reconstruction of missing or corrupted portions of a digital audio signal. Inpainting techniques are employed when parts of the audio have been lost due to various factors such as transmission errors, data corruption or errors during recording. The goal of audio inpainting is to fill in the gaps (i.e., the missing portions) in the audio signal seamlessly, making the reconstructed portions indistinguishable from the original content and avoiding the introduction of audible distortions or alterations. Many techniques have been proposed to solve the audio inpainting problem and this is usually achieved by analyzing the temporal and spectral information surrounding each missing portion of the considered audio signal. Classic methods employ statistical models or digital signal processing algorithms to predict and synthesize the missing or damaged sections. Recent solutions, instead, take advantage of deep learning models, thanks to the growing trend of exploiting data-driven methods in the context of audio restoration. Depending on the extent of the lost information, the inpainting task can be divided in three categories. Short inpainting refers to the reconstruction of few milliseconds (approximately less than 10) of missing signal, that occurs in the case of short distortions such as clicks or clipping. In this case, the goal of the reconstruction is to recover the lost information exactly. In long inpainting instead, with gaps in the order of hundreds of milliseconds or even seconds, this goal becomes unrealistic, since restoration techniques cannot rely on local information. Therefore, besides providing a coherent reconstruction, the algorithms need to generate new information that has to be semantically compatible with the surrounding context (i.e., the audio signal surrounding the gaps). The case of medium duration gaps lays between short and long inpainting. It refers to the reconstruction of tens of millisecond of missing data, a scale where the non-stationary characteristic of audio already becomes important. == Definition == Consider a digital audio signal x {\displaystyle \mathbf {x} } . A corrupted version of x {\displaystyle \mathbf {x} } , which is the audio signal presenting missing gaps to be reconstructed, can be defined as x ~ = m ∘ x {\displaystyle \mathbf {\tilde {x}} =\mathbf {m} \circ \mathbf {x} } , where m {\displaystyle \mathbf {m} } is a binary mask encoding the reliable or missing samples of x {\displaystyle \mathbf {x} } , and ∘ {\displaystyle \circ } represents the element-wise product. Audio inpainting aims at finding x ^ {\displaystyle \mathbf {\hat {x}} } (i.e., the reconstruction), which is an estimation of x {\displaystyle \mathbf {x} } . This is an ill-posed inverse problem, which is characterized by a non-unique set of solutions. For this reason, similarly to the formulation used for the inpainting problem in other domains, the reconstructed audio signal can be found through an optimization problem that is formally expressed as x ^ ∗ = argmin X ^ L ( m ∘ x ^ , x ~ ) + R ( x ^ ) {\displaystyle \mathbf {\hat {x}} ^{}={\underset {\hat {\mathbf {X} }}{\text{argmin}}}~L(\mathbf {m} \circ \mathbf {\hat {x}} ,\mathbf {\tilde {x}} )+R(\mathbf {\hat {x}} )} . In particular, x ^ ∗ {\displaystyle \mathbf {\hat {x}} ^{}} is the optimal reconstructed audio signal and L {\displaystyle L} is a distance measure term that computes the reconstruction accuracy between the corrupted audio signal and the estimated one. For example, this term can be expressed with a mean squared error or similar metrics. Since L {\displaystyle L} is computed only on the reliable frames, there are many solutions that can minimize L ( m ∘ x ^ , x ~ ) {\displaystyle L(\mathbf {m} \circ \mathbf {\hat {x}} ,\mathbf {\tilde {x}} )} . It is thus necessary to add a constraint to the minimization, in order to restrict the results only to the valid solutions. This is expressed through the regularization term R {\displaystyle R} that is computed on the reconstructed audio signal x ^ {\displaystyle \mathbf {\hat {x}} } . This term encodes some kind of a-priori information on the audio data. For example, R {\displaystyle R} can express assumptions on the stationarity of the signal, on the sparsity of its representation or can be learned from data. == Techniques == There exist various techniques to perform audio inpainting. These can vary significantly, influenced by factors such as the specific application requirements, the length of the gaps and the available data. In the literature, these techniques are broadly divided in model-based techniques (sometimes also referred as signal processing techniques) and data-driven techniques. === Model-based techniques === Model-based techniques involve the exploitation of mathematical models or assumptions about the underlying structure of the audio signal. These models can be based on prior knowledge of the audio content or statistical properties observed in the data. By leveraging these models, missing or corrupted portions of the audio signal can be inferred or estimated. An example of a model-based techniques are autoregressive models. These methods interpolate or extrapolate the missing samples based on the neighboring values, by using mathematical functions to approximate the missing data. In particular, in autoregressive models the missing samples are completed through linear prediction. The autoregressive coefficients necessary for this prediction are learned from the surrounding audio data, specifically from the data adjacent to each gap. Some more recent techniques approach audio inpainting by representing audio signals as sparse linear combinations of a limited number of basis functions (as for example in the Short Time Fourier Transform). In this context, the aim is to find the sparse representation of the missing section of the signal that most accurately matches the surrounding, unaffected signal. The aforementioned methods exhibit optimal performance when applied to filling in relatively short gaps, lasting only a few tens of milliseconds, and thus they can be included in the context of short inpainting. However, these signal-processing techniques tend to struggle when dealing with longer gaps. The reason behind this limitation lies in the violation of the stationarity condition, as the signal often undergoes significant changes after the gap, making it substantially different from the signal preceding the gap. As a way to overcome these limitations, some approaches add strong assumptions also about the fundamental structure of the gap itself, exploiting sinusoidal modeling or similarity graphs to perform inpainting of longer missing portions of audio signals. === Data-driven techniques === Data-driven techniques rely on the analysis and exploitation of the available audio data. These techniques often employ deep learning algorithms that learn patterns and relationships directly from the provided data. They involve training models on large datasets of audio examples, allowing them to capture the statistical regularities present in the audio signals. Once trained, these models can be used to generate missing portions of the audio signal based on the learned representations, without being restricted by stationarity assumptions. Data-driven techniques also offer the advantage of adaptability and flexibility, as they can learn from diverse audio datasets and potentially handle complex inpainting scenarios. As of today, such techniques constitute the state-of-the-art of audio inpainting, being able to reconstruct gaps of hundreds of milliseconds or even seconds. These performances are made possible by the use of generative models that have the capability to generate novel content to fill in the missing portions. For example, generative adversarial networks, which are the state-of-the-art of generative models in many areas, rely on two competing neural networks trained simultaneously in a two-player minmax game: the generator produces new data from samples of a random variable, the discriminator attempts to distinguish between generated and real data. During the training, the generator's objective is to fool the discriminator, while the discriminator attempts to learn to better classify real and fake data. In GAN-based inpainting methods the generator acts as a context encoder and produces a plausible completion for the gap only given the available information surrounding it. The discriminator is used to train the generator and tests the consistency of the produced inpainted audio. Recently, also diffusion models have established themselves as the state-of-the-art of generative models in many fields, often beating even GAN-based solutions. For this reason they have also been used to solve the audio inpainting problem, obtaining valid results. These models generate new data instances by inverting the
Best AI Code Generators in 2026
Comparing the best AI code generator? An AI code generator is software that uses machine learning to help you get more done — it lowers the barrier so anyone can produce professional output. Privacy matters too: check whether your data trains the model and whether a no-log or enterprise tier is available. Whether you are a beginner or a pro, the right AI code generator slots into your workflow and pays for itself fast. We tested the leading options and ranked them by quality, value, and ease of use.
Best AI Logo Makers in 2026
Looking for the best AI logo maker? An AI logo maker is software that uses machine learning to help you get more done — it can save you hours every week by automating repetitive work. Most options offer a generous free tier, with paid plans unlocking higher limits, faster processing, and team features. Whether you are a beginner or a pro, the right AI logo maker slots into your workflow and pays for itself fast. Read on for hands-on impressions, pricing tiers, and the standout features that matter.
Apertium
Apertium is a free/open-source rule-based machine translation platform. It is free software and released under the terms of the GNU General Public License. == Overview == Apertium is a transfer-based machine translation system, which uses finite state transducers for all of its lexical transformations, and Constraint Grammar taggers as well as hidden Markov models or Perceptrons for part-of-speech tagging / word category disambiguation. A structural transfer component is responsible for word movement and agreement; most Apertium language pairs up until now have used "chunking" or shallow transfer rules, though newer pairs use (possibly recursive) rules defined in a Context-free grammar. Many existing machine translation systems available at present are commercial or use proprietary technologies, which makes them very hard to adapt to new usages. Apertium code and data is free software and uses a language-independent specification, to allow for the ease of contributing to Apertium, more efficient development, and enhancing the project's overall growth. At present (December 2020), Apertium has released 51 stable language pairs, delivering fast translation with reasonably intelligible results (errors are easily corrected). Being an open-source project, Apertium provides tools for potential developers to build their own language pair and contribute to the project. == History == Apertium originated as one of the machine translation engines in the project OpenTrad, which was funded by the Spanish government, and developed by the Transducens research group at the Universitat d'Alacant. It was originally designed to translate between closely related languages, although it has recently been expanded to treat more divergent language pairs. To create a new machine translation system, one just has to develop linguistic data (dictionaries, rules) in well-specified XML formats. Language data developed for it (in collaboration with the Universidade de Vigo, the Universitat Politècnica de Catalunya and the Universitat Pompeu Fabra) currently support (in stable version) the Arabic, Aragonese, Asturian, Basque, Belarusian, Breton, Bulgarian, Catalan, Crimean Tatar, Danish, English, Esperanto, French, Galician, Hindi, Icelandic, Indonesian, Italian, Kazakh, Macedonian, Malaysian, Maltese, Northern Sami, Norwegian (Bokmål and Nynorsk), Occitan, Polish, Portuguese, Romanian, Russian, Sardinian, Serbo-Croatian, Silesian, Slovene, Spanish, Swedish, Tatar, Ukrainian, Urdu, and Welsh languages. A full list is available below. Several companies are also involved in the development of Apertium, including Prompsit Language Engineering, Imaxin Software and Eleka Ingeniaritza Linguistikoa. The project has taken part in the 2009, 2010, 2011, 2012, 2013 and 2014 editions of Google Summer of Code and the 2010, 2011, 2012, 2013, 2014, 2015, 2016 and 2017 editions of Google Code-In. == Translation methodology == This is an overall, step-by-step view how Apertium works. The diagram displays the steps that Apertium takes to translate a source-language text (the text we want to translate) into a target-language text (the translated text). Source language text is passed into Apertium for translation. The deformatter removes formatting markup (HTML, RTF, etc.) that should be kept in place but not translated. The morphological analyser segments the text (expanding elisions, marking set phrases, etc.), and looks up segments in the language dictionaries, returning dictionary forms and tags for all matches. In pairs that involve agglutinative morphology, including a number of Turkic languages, a Helsinki Finite State Transducer (HFST) is used. Otherwise, an Apertium-specific finite state transducer system called lttoolbox, is used. The morphological disambiguator (the morphological analyser and the morphological disambiguator together form the part of speech tagger) resolves ambiguous segments (i.e., when there is more than one match) by choosing one match. Apertium uses Constraint Grammar rules (with the vislcg3 parser) for most of its language pairs. Retokenisation uses a finite state transducer to match sequences of lexical units and may reorder or translate tags (often used for translating idiomatic expressions into something that more approaches the target language grammar) Lexical transfer looks up disambiguated source-language basewords to find their target-language equivalents (i.e., mapping source language to target language). For lexical transfer, Apertium uses an XML-based dictionary format called bidix. Lexical selection chooses between alternative translations when the source text word has alternative meanings. Apertium uses a specific XML-based technology, apertium-lex-tools, to perform lexical selection. Structural transfer (i.e., it is an XML format that allows writing complex structural transfer rules) can consist of one-step chunking transfer, three-step chunking transfer or a CFG-based transfer module. The chunking modules flag grammatical differences between the source language and target language (e.g. gender or number agreement) by creating a sequence of chunks containing markers for this. They then reorder or modify chunks in order to produce a grammatical translation in the target-language. The newer CFG-based module matches input sequences into possible parse trees, selecting the best-ranking one and applying transformation rules on the tree. The morphological generator uses the tags to deliver the correct target language surface form. The morphological generator is a morphological transducer, just like the morphological analyser. A morphological transducer both analyses and generates forms. The post-generator makes any necessary orthographic changes due to the contact of words (e.g. elisions). The reformatter replaces formatting markup (HTML, RTF, etc.) that was removed by the deformatter in the first step. Apertium delivers the target-language translation. == Supported languages == As of June 2026, the following 108 pairs and 51 languages and languages varieties are supported by Apertium.
Ayoba
Ayoba is an African communication platform developed in South Africa. It is owned by Progressive Tech Holdings in Mauritius and managed by SIMFY Africa. Launched on May 4, 2019, as of April 2024, it has over 35 million active users. == History == Ayoba was first published on Google Play in February 2019. Its first marketing campaign and brand launch took place in Cameroon on May 4, 2019. In June 2019, the platform introduced its first eight channels. In November 2019, the platform reached one million active users, which increased to two million by June 2020. Subsequently, ayoba expanded its services, including the launch of games for Android in February 2020, Momo (Mobile Money) in Cameroon in May 2020, and MicroApps in May 2020. It also launched music and voice and video calling features in 12 territories in August 2020. The first version of ayoba for iOS was released in September 2020. In December of the same year, games and Messaging 2.0 were launched on the platform. In November 2020, it won Best Mobile Application at the African Digital Awards. In 2021, it won OTT Brand of the Year at the Marketing World Awards in Ghana. In December 2022, it received Top Innovative Technology and Telecom Product of the Year at the National Communications Awards in December 2022. In June 2023 ayoba partnered with BoomPlay and as of April 2024, it had 35 million monthly active users. Ayoba has partnered with Jumia Ghana to offer exclusive deals to users. Ayoba users can get a 10% discount on selected Jumia purchases through the app, with no data charges for MTN users. This partnership aims to make online shopping more affordable and accessible by integrating Jumia's offers into the ayoba app. Ayoba supports over 35 million users across Africa and provides services in 22 languages. To access the deals, users can download the ayoba app from the Google Play Store, iOS Store, or the official website. == Platform features == Chat, Call and Share: ayoba enables instant messaging, voice notes, picture sharing, and file sharing with contacts, even if they do not have the app installed. The app supports voice and video calls on both Android and iOS, as well as group chats, help channel and SMS continuity (non ayoba users receive messages as SMS, their responses appear in the ayoba app). Music: ayoba offers a free music player with daily updates on international and African music. Users can find playlists for different genres. Games: ayoba provides a selection of interactive games, including action, adventure, and children's games available on both Android and iOS. Mobile Money Transfers: In certain territories, ayoba supports mobile money transfers using MTN Mobile Money (MoMo) for transactions within the app. MicroApps: ayoba features individual MicroApps within the platform that offer content and services, including streaming channels, podcasts, and specialized apps. The availability of these apps may vary by country. == Operations == ayoba primarily focuses on the following territories: Nigeria, Cameroon, South Africa, Ghana, Côte d'Ivoire, Uganda, Republic of Congo, Benin, Zambia, Tanzania, Kenya, Senegal, Togo, Guinea Bissau, Guinea Conakry, Sudan, South Sudan, and Liberia. The company operates from its offices in Cape Town and Johannesburg, South Africa. David Gillaranz served as the CEO from 2019 to 2021, and Burak Akinci has been the CEO since 2021.
Mona Diab
Mona Talat Diab (Arabic: منى طلعت دياب) is a computer science professor and director of Carnegie Mellon University's Language Technologies Institute. Previously, she was a professor at George Washington University and a research scientist with Facebook AI. Her research focuses on natural language processing, computational linguistics, cross lingual/multilingual processing, computational socio-pragmatics, Arabic language processing, and applied machine learning. == Education == Diab completed her M.Sc. in computer science with a major in machine learning and artificial intelligence at The George Washington University (1997) and her Ph.D. in computational linguistics at the University of Maryland, Linguistics Department and University of Maryland Institute for Advanced Computer Studies (UMIACS) in 2003, under the supervision of Philip Resnik. She was also a postdoctoral research scientist at Stanford University (2003–2005) under the mentorship of Dan Jurafsky, where she was a part of the Stanford NLP Group. == Career == After her postdoc at Stanford, Diab took a position as research scientist (principal investigator) at the Center for Computational Learning Systems (CCLS) in Columbia University, where she was also adjunct professor in the computer science department. In 2013 she joined the George Washington University as an associate professor, where she was promoted to full professor in 2017. Diab is the founder and director of the GW NLP lab CARE4Lang. Diab served as an elected faculty senator at Columbia University for 6 years (2007–2012) and an elected faculty senator at GW (2013–2014). She served the computational linguistics community as elected member, secretary and president of ACL SIGLEX (2005–2016) and elected president of ACL SIGSemitic. She currently serves as the elected VP-elect for ACL SIGDAT. In 2017 Diab joined Amazon AWS AI Deep Learning Group for Human Language Technologies, where she led the AWS Lex project for task oriented dialogue systems for enterprises. A couple of years later, she moved to Facebook AI as a research scientist. In the fall of 2023, she became the director of CMU's Language Technologies Institute -- the first full time director since the passing of its founder Jaime Carbonell. == Research == Diab's research interests include several areas in computational linguistics/natural language processing, like conversational AI, computational lexical semantics, multilingual and cross lingual processing, social media processing with an emphasis on computational socio- pragmatics, information extraction & text analytics, machine translation. Besides this, she also has special interests in Arabic NLP and low resource scenarios. Diab co-established two research trends in the computational linguistics field, computational approaches to linguistic code switching in 2007 and semantic textual similarity in 2010. Diab together with Nizar Habash and Owen Rambow, co-founded CADIM in 2005, a global reference point in Arabic dialect processing. In 2012, Diab together with Eneko Agirre and Johan Bos, brought together two ACL communities SIGLEX and SIGSEM and established the 1st tier conference SEM. == Awards and recognition == Selected as one of top 150 leaders and visionaries in AI nationwide to participate in White House AI Summit in Government, Washington, D.C., US, September 2019 March 2017: 3 Muslim Women in STEM You Should Know About, Teen Vogue, March 2017 May 2017: Behind Every Strong Woman Is...Another Strong Woman: Ten women give thanks to the women who supported them on the way up. Elle, May 2017. Google Faculty Research Award – Tharwa++: Building a multidialectal Arabic Lexical Repository, (PI), 09.2015 –12.2016. Google Faculty Research Award – Nuanced Sentiment and Perspective Analysis for Arabic Social Media Text, (PI), 12.2014 –12.2015 QNRF Best Poster Award – Ossama Obeid, Houda Bouamor, Wajdi Zaghouani, Mahmoud Ghoneim, Abdelati Hawwari, Mona Diab, Kemal Oflazer. (2016) MANDIAC: A Web-based Annotation System For Manual Arabic Diacritization. Proceedings of the 2nd Workshop on Arabic Corpora and Processing Tools, LREC 2016. Best Paper Award – Aminian, Maryam, Mahmoud Ghoneim, Mona Diab. (2015) Unsupervised False Friend Disambiguation Using Contextual Word Clusters and Parallel Word Alignments. In Proceedings of Workshop 9th Semantics Syntax Statistical Translation, NAACL 2015, Denver CO, US. == Publications == Diab has over 250 publications, and she is an acting editor for several scientific journals. === Selected publications === Semeval-2012 task 6: A pilot on semantic textual similarity. E. Agirre, D. Cer, M. Diab, A. Gonzalez-Agirre. SEM 2012: The First Joint Conference on Lexical and Computational Semantics–Volume 1: Proceedings of the main conference and the shared task, and Volume 2: Proceedings of the Sixth International Workshop on Semantic Evaluation (SemEval 2012) Predictive linguistic features of schizophrenia. ES Kayi, M Diab, L Pauselli, M Compton, G Coppersmith. arXiv preprint arXiv:1810.09377 Ideological perspective detection using semantic features. H Elfardy, M Diab, C Callison-Burch – Proceedings of SEM 2015 DeSePtion: Dual sequence prediction and adversarial examples for improved fact-checking. Christopher Hidey, Tuhin Chakrabarty, Tariq Alhindi, Siddharth Varia, Kriste Krstovski, Mona Diab, Smaranda Muresan, 2020 Does Causal Coherence Predict Online Spread of Social Media? Pedram Hosseini, Mona Diab, David A Broniatowski. Proceedings of International Conference on Social Computing, Behavioral-Cultural Modeling and Prediction and Behavior Representation in Modeling and Simulation, 2019. Diversity, Density, and Homogeneity: Quantitative Characteristic Metrics for Text Collections. YA Lai, X Zhu, Y Zhang, M Diab, arXiv preprint arXiv:2003.08529, 2020 Readability of written medicine information materials in Arabic language: expert and consumer evaluation. S Al Aqeel, N Abanmy, A Aldayel, H Al-Khalifa, M Al-Yahya, M Diab. BMC health services research 18 (1), 1–7, 2019 Unsupervised word mapping using structural similarities in monolingual embeddings. H Aldarmaki, M Mohan, M Diab – Transactions of the Association for Computational Linguistics, 2018 An unsupervised method for word sense tagging using parallel corpora M Diab, P Resnik. Proceedings of ACL 2002 Overview for the first shared task on language identification in code-switched data. Thamar Solorio, Elizabeth Blair, Suraj Maharjan, Steven Bethard, Mona Diab, Mahmoud Ghoneim, Abdelati Hawwari, Fahad AlGhamdi, Julia Hirschberg, Alison Chang, Pascale Fung. Proceedings of the First Workshop on Computational Approaches to Code Switching, 2014 Modeling sentences in the latent space. W Guo, M Diab – ACL 20 12 Task-based evaluation of multiword expressions: a pilot study in statistical machine translation. M Carpuat, M Diab – NAACL-HLT 2010 Rumor detection and classification for twitter data. S Hamidian, MT Diab – arXiv preprint arXiv:1912.08926, 2019 Subgroup detection in ideological discussions. A Abu-Jbara, P Dasigi, M Diab, D Radev – ACL 2012 Madamira: A fast, comprehensive tool for morphological analysis and disambiguation of arabic. A. Pasha, M. Al-Badrashiny, M. Diab, A. El Kholy, R. Eskander, N. Habash, M. Pooleery, O. Rambow, R. Roth. LREC 14, 1094–1101. 2014 Context-Aware Self-Attentive Natural Language Understanding for Task-Oriented Chatbots. A. Gupta, P. Zhang, G. Lalwani, M. Diab. EMNLP 2019 A multitask learning approach for diacritic restoration. S. Alqahtani, A. Mishra, M. Diab. ACL 2020