Apache cTAKES: clinical Text Analysis and Knowledge Extraction System is an open-source Natural Language Processing (NLP) system that extracts clinical information from electronic health record unstructured text. It processes clinical notes, identifying types of clinical named entities — drugs, diseases/disorders, signs/symptoms, anatomical sites and procedures. Each named entity has attributes for the text span, the ontology mapping code, context (family history of, current, unrelated to patient), and negated/not negated. cTAKES was built using the UIMA Unstructured Information Management Architecture framework and OpenNLP natural language processing toolkit. == Components == Components of cTAKES are specifically trained for the clinical domain, and create rich linguistic and semantic annotations that can be utilized by clinical decision support systems and clinical research. These components include: Named Section identifier Sentence boundary detector Rule-based tokenizer Formatted list identifier Normalizer Context dependent tokenizer Part-of-speech tagger Phrasal chunker Dictionary lookup annotator Context annotator Negation detector Uncertainty detector Subject detector Dependency parser patient smoking status identifier Drug mention annotator == History == Development of cTAKES began at the Mayo Clinic in 2006. The development team, led by Dr. Guergana Savova and Dr. Christopher Chute, included physicians, computer scientists and software engineers. After its deployment, cTAKES became an integral part of Mayo's clinical data management infrastructure, processing more than 80 million clinical notes. When Dr. Savova's moved to Boston Children's Hospital in early 2010, the core development team grew to include members there. Further external collaborations include: University of Colorado Brandeis University University of Pittsburgh University of California at San Diego Such collaborations have extended cTAKES' capabilities into other areas such as Temporal Reasoning, Clinical Question Answering, and coreference resolution for the clinical domain. In 2010, cTAKES was adopted by the i2b2 program and is a central component of the SHARP Area 4. In 2013, cTAKES released their first release as an Apache Software Foundation incubator project: cTAKES 3.0. In March 2013, cTAKES became an Apache Software Foundation Top Level Project (TLP).
TargetLink
TargetLink is a software for automatic code generation, based on a subset of Simulink/Stateflow models, produced by dSPACE GmbH. TargetLink requires an existing MATLAB/Simulink model to work on. TargetLink generates both ANSI-C and production code optimized for specific processors. It also supports the generation of AUTOSAR-compliant code for software components for the automotive sector. The management of all relevant information for code generation takes place in a central data container, called the Data Dictionary. Testing of the generated code is implemented in Simulink, which is also used for the specification of the underlying simulation models. TargetLink supports three simulation modes to test the generated code: Model-in-the-loop simulation (MIL): this mode allows the model design to be checked. An MIL simulation is also known as a floating-point simulation, since the variables are typically floating-point variables. Software-in-the-loop (SIL): the simulation is based on the execution of generated code, which runs on a PC system. The variables are typically plain or fixed point numbers. Processor-in-the-loop (PIL): in a PIL simulation, the generated code runs on the target hardware or on an evaluation board. So-called real-time frames are included, making it possible to transfer the simulation results as well as memory consumption and runtime information to the PC. The Motor Industry Software Reliability Association (MISRA) published official MISRA modeling guidelines for TargetLink in late 2007, which are particularly important for functional safety of safety-critical applications. In 2009, TÜV SÜD certified TargetLink for use during the development of safety-critical systems to ISO DIS 26262 and IEC 61508.
ACM SIGEVO
The ACM SIGEVO is a Special Interest Group of the Association of Computing Machinery for members of that organization who are practitioners, academics, students or others with interests in evolutionary computation and related algorithms. == History == ACM SIGEVO was founded in 2005 when the International Society for Genetic and Evolutionary Computation (ISGEC) became an ACM Special Interest Group under its present title. The ISGEC had been formed in 1999 by the merger of the Genetic Programming conference organization with the International Conference on Genetic Algorithms (ICGA) leading to the first Genetic and Evolutionary Computation Conference (GECCO). == Membership == Members of this SIG pay a small fee in addition to the ACM membership fee. In return they have access to a quarterly online newsletter, but more importantly can obtain reduced registration rates at the two conferences organised by ACM SIGEVO: GECCO and the Foundations of Genetic Algorithms conference (FOGA). They can also access material on evolutionary computation and related topics in the ACM Digital Library. In addition they can subscribe to email mailing lists in order to keep informed about news over time. For students, ACM SIGEVO sponsors Travel Awards for attendance at the GECCO Conference and FOGA (the Foundations of Genetic Algorithms conference). ACM SIGEVO also sponsors a Graduate Student Workshop. ACM also sponsors Awards to be competed for by attendees at the conferences it organises. == Conferences == ACM SIGEVO organises two major conferences in the field of evolutionary computation. The Genetic and Evolutionary Conference (GECCO) is held annually, while the Foundations of Genetic Algorithms conference (FOGA) is held biennially. === GECCO === The first GECCO conference was held prior to the formation of ACM SIGEVO but since 2005 (see History above) it has been organised annually by ACM SIGEVO. The latest (2025) was held in Málaga, Spain. The next (2026) will be held in San José, Costa Rica. === FOGA === Foundations of Genetic Algorithms (FOGA) is a biennial peer-reviewed research conference focusing on the theoretical principles underlying genetic algorithms, other evolutionary algorithms and related heuristics. It is organized by ACM SIGEVO. Its relevance to the computer science research community has been reflected in an A-rating in the CORE computer science conference assessment system. The Foundations of Genetic Algorithms (FOGA) conference originated as a workshop in 1990 in order to create an opportunity for researchers on genetic algorithms and related areas of evolutionary computation to focus on the theoretical principles underlying their field. From the start its multi-day duration made it comparable to conferences in the field, and since 2015 its proceedings have used conference rather than workshop in their titles. In 2005 ACM SIGEVO the Association for Computing Machinery Special Interest Group on Genetic and Evolutionary Computation was formed and every FOGA conference since then has been supported by SIGEVO. The table below shows FOGA conferences by year, location, websites (where available) and publisher of proceedings. A citation follows the reference to the publisher giving the full details of each FOGA proceedings. Papers accepted at recent conferences have been presented as digital or print posters in poster sessions at the conference, before being published in written form in the conference proceedings. FOGA is comparable in its multi-day duration to other conferences on evolutionary computation such as CEC, GECCO and PPSN. The main difference is that FOGA focuses on the theoretical basis of evolutionary computation and related subjects. While the above conferences devote some time to theory they also cover a wide range of other topics including competitions and applications. This focus on theoretical computer science was reflected in the CORE computer science conference assessment exercise, where FOGA was given an A-ranking in the 2023 assessment. GECCO and PPSN also obtained A-rankings, but many other conferences in the field of evolutionary computation obtained lower rankings. This suggests that FOGA is a relevant conference in its field, comparable with others including the much larger CEC or GECCO. Keynote speakers at past conferences have been: == Awards == ACM SIGEVO sponsors a number of awards. === SIGEVO Outstanding Contribution Award === The SIGEVO Outstanding Contribution Award commenced in 2023, and these awards are designed to recognise distinctive contributions to the field of evolutionary computation when evaluated over a period of at least 15 years. As a result many recipients to date are notable academics or industrial practitioners, and include Anne Auger, Kalyanmoy Deb, Stephanie Forrest, Emma Hart and Hans-Paul Schwefel. === SIGEVO Dissertation Award === The SIGEVO Dissertation Award recognises thesis research in the field of evolutionary computation completed at least by the year prior to a GECCO conference. Theses are submitted and reviewed by a panel that selects one winner and a maximum of two honourable mentions. Awards will be made to the winner and any others at the next GECCO conference. === SIGEVO Chair Award === The SIGEVO Chair Award, established in 2016 is a lecture sponsored by ACM SIGEVO, to take place on the last day of the GECCO conference. It recognizes through the lectures that the lecturers are influential researchers in the field of evolutionary computation. The more recent lectures are available online. The 2024 Award winner was Una-May O'Reilly. === SIGEVO Impact Award === The SIGEVO Impact Award looks back to the GECCO conference ten years earlier and recognizes up to three papers a year which are considered by the current ACM SIGEVO Executive Committee to have had significant impact over the period since their first publication at the GECCO conference. An example (originally published in GECCO 2010) received this award in 2020. === GECCO Best Paper Award === The ACM SIGEVO sponsors awards for the best papers presented at the GECCO conference. Because GECCO conferences have very many parallel tracks there are multiple awards recognising presentations in the different tracks. At GECCO 2025 Best Paper Awards were presented across 12 tracks. === FOGA Best Paper Award === The ACM SIGEVO sponsors awards for the best papers presented at the FOGA conference. Because FOGA operates on a single track, it is easier to compare papers. Since 2019 this Award has been made (suggesting only four awards up to the latest conference in 2025). ACM SIGEVO records the 2019 award. === Humie Award === The Humies Awards are rewards for the best form of human-competitive results using evolutionary computation or related algorithms and published in the wider literature (they do not need to be published at a conference or in a journal sponsored by ACM SIGEVO or even the ACM.) They were established through a gift from John Koza and have been in operation from 2004 to the present. The link with ACM SIGEVO is that the winners of the competition (submissions are evaluated in advance) are presented with Humie Awards at GECCO conferences. The Humie Awards website provides full details for the rules and how to submit entries to the competition. == Journals == ACM SIGEVO sponsors the main journal covering evolutionary computation published by the ACM: ACM Transactions on Evolutionary Learning and Optimization. ACM SIGEVO refers to the preceding ISGEC organisation (see History above) as sponsoring two other important journals in the field: The Evolutionary Computation journal. Genetic Programming and Evolvable Machines. While these journals continue to be important in the field, the wording on the website of ACM SIGEVO suggests that ACM SIGEVO is not involved in their publication. == References and notes ==
Mata v. Avianca, Inc.
Mata v. Avianca, Inc. was a U.S. District Court for the Southern District of New York case in which the Court dismissed a personal injury case against the airline Avianca and issued a $5,000 fine to the plaintiffs' lawyers who had submitted fake precedents generated by ChatGPT in their legal briefs. == Background == In February 2022, Roberto Mata filed a personal injury lawsuit in the U.S. District Court for the Southern District of New York against Avianca, alleging that he was injured when a metal serving cart struck his knee during an international flight. The plaintiff's lawyers used ChatGPT to generate a legal motion, which contained numerous fake legal cases involving fictitious airlines with fabricated quotations and internal citations. Avianca's lawyers notified the Court that they had been "unable to locate" a few legal cases cited in the legal motion. The Court could not locate the cases either and ordered the plaintiff's lawyers to provide copies of the cited legal cases. Mata's lawyers provided copies of documents purportedly containing all but one of the legal cases, after ChatGPT assured that the cases "indeed exist" and "can be found in reputable legal databases such as LexisNexis and Westlaw." == Opinion == In May 2023, Judge P. Kevin Castel dismissed the personal injury case against Avianca and ordered the plaintiff's attorneys to pay a $5,000 fine. Judge Castel noted numerous inconsistencies in the opinion summaries, describing one of the legal analyses as "gibberish." Judge Castel held that Mata's lawyers had acted with "subjective bad faith" sufficient for sanctions under Federal Rule of Civil Procedure Rule 11. == Impact == In July 2024, the American Bar Association issued its first formal ethics opinion on the responsibilities of lawyers using generative AI (GAI). The 15-page opinion outlines how the Rules of Professional Conduct apply to the use of GAI in the practice of law. Experts caution that lawyers cannot reasonably rely on the accuracy, completeness, or validity of content generated by GAI tools. Due to the continued usage of GAI in the practice of law, Mata has been described as a landmark case by legal professionals, as it is frequently cited by courts in cases where usage of GAI during the course of proceedings leads to the creation and citation of nonexistent caselaw.
Minne Atairu
Minne Atairu is a Nigerian interdisciplinary artist, a recipient of the 2021 Global South Award Lumen Prize for Art and Technology. She generates synthetic Benin Bronzes through recombination of historical fragments, sculptures, texts, images, and sounds. == Early life and education == Atairu was born in Benin, Nigeria. She holds a bachelor's degree in art history from the University of Maiduguri in Maiduguri, Nigeria; a master's degree in museum studies from the George Washington University in Washington, D.C.; and a doctorate in art education from Teachers College, Columbia University in New York City. Her academic research integrates artificial intelligence, art/museum education and hip-hop based education. == Works == Atairu's artmaking involves using artificial intelligence (AI; such as StyleGAN, GPT-3) to make artwork. She uses tools such as Midjourney and Blender software to develop her works. === Mami Wata === Her first work is a Yoruba goddess called Mami Wata where she used Midjourney in generating the images. === To the Hand === For her 2023 installation To the Hand at The Shed arts center, she worked with Blender to convert text into 3D-printed sculptures made of corn starch or sugarcane infused with bronze. The rings of ground terra-cotta that surround the sculpture represent the walls and deep moats of Benin. == Publications == Atairu, Minne (February 1, 2024). "Reimagining Benin Bronzes using generative adversarial networks". AI & Society. 39 (1): 91–102. doi:10.1007/s00146-023-01761-7. ISSN 1435-5655.
Edits (app)
Edits is an American photo and short form video editing software service owned by Meta Platforms. It allows users to create videos and edit them by using features like green screens, and AI animation, and also provides real-time statistics to Instagram creators to track their accounts. Accounts directly from Instagram can be imported, and videos can be exported vice-versa. It is available solely on iOS and Android. On Apple, it supports over 32 different languages, including French, Spanish, and Chinese. It has been noted by critics as a direct competitor for apps like CapCut, owned by Chinese brand ByteDance. The Instagram head, Adam Mosseri, also acknowledged these similarities. Launched on April 22 for both iOS and Android. It received over 5M+ users on Apple and Android combined in its first 4 days since its launch. == History == On January 19, 2025, following the ban of all ByteDance Apps from the Google Play Store, and App Store, Instagram head Adam Mosseri announced on Threads that they would be launching the app in February for iOS, followed by an Android counterpart. He said the app is working with select people to test its features. In a separate post, he emphasized that the app is "more for creators than casual video makers". == Features == Edits contains many similar features to other competition of video editors like KineMaster, Inshot, and CapCut. When creating a video, users have the option to export in resolution of HD, 4K, and 2K, along with having HDR and SDR support. Like many traditional video editing software, it includes a timeline, and basic undo-redo buttons. On the bottom bar, 7 tabs for editing exist, namely the Split, Volume, Adjust, Speed, Delete, Filters, Green Screen, Voice FX, Extract Audio, Mirror, Slip, Replace and Duplicate bars. Basic features, like splitting, and adjusting speed and volume of clips are present, along with more advanced Green Screens, and AI features. Being a mobile video editor app, Edits also has drag-and-drop features to ease customer usage. Users have the ability to record videos directly within the app. This feature allows users to create content without needing extra software or devices. They can choose from several focal lengths, which affect how close or wide the shot appears. The app also supports different frame rates. Users have the ability to record videos directly within the app. This feature allows users to create content without needing extra software or devices. Once users are done filming your clips, they can simply transfer them into a project to start editing immediately. Upcoming features for the app include Keyframes, AI-powered modification, Collaboration, and Enhanced creativity. == Reception == Since its release, it received over 5 million downloads in 4 days. Critically, the app received great rankings from many. From users, the app received an average of 4.45 stars over Google Play Store and App Store in the first few days, with Google Play Store receiving the least stars. As in reviews, it was received mixed by the public. Many people praised the smoothness and intuivity of the app. "The app is more than just a basic editor, offering a full suite of creative tools, including a dedicated tab for inspiration and trending audio, as well as a tab for managing drafts," said a blogger. Some users were disappointed with the range of editing tools, some users have noted that it could benefit from more transition options between clips. Some even reported crashing between clips.
Whisper (speech recognition system)
Whisper is a machine learning model for speech recognition and transcription, created by OpenAI and first released as open-source software in September 2022. It is capable of transcribing speech in English and multiple other languages, and can translate several non-English languages into English. Whisper is a weakly-supervised deep learning acoustic model, made using an encoder-decoder transformer architecture. OpenAI claims that the combination of different training data and post-training filtering used in its development has led to improved recognition of accents, background noise, and jargon compared to previous approaches. While the model does not outperform larger, more specialized models and still experiences AI hallucination, it has been showed to be useful for general sound recognition and has many applications across different industries. == Background == Speech recognition has had a long history in research; the first approaches made use of statistical methods, such as dynamic time warping, and later hidden Markov models. At around the 2010s, deep neural network approaches became more common for speech recognition models, which were enabled by the availability of large datasets ("big data") and increased computational performance. Early approaches to deep learning in speech recognition included convolutional neural networks, which were limited due to their inability to capture sequential data, which later led to developments of Seq2seq approaches, which include recurrent neural networks, which made use of long short-term memory. Transformers, introduced in 2017 by Google, displaced many prior state-of-the-art approaches across a wide range in machine learning, and started becoming the core neural architecture in fields such as language modeling and computer vision. Weakly-supervised approaches to training acoustic models were recognized in the early 2020s as promising for speech recognition approaches using deep neural networks. According to a NYT report, in 2021 OpenAI believed they exhausted sources of higher-quality data to train their large language models and decided to complement scraped web text with transcriptions of YouTube videos and podcasts, and developed Whisper to solve this task. Whisper Large V2 was released on December 8, 2022, followed by Whisper Large V3 being released in November 2023, during the OpenAI Dev Day. In March 2025, OpenAI released new transcription models based on GPT-4o and GPT-4o mini, both of which have lower error rates than Whisper. == Architecture == The Whisper architecture is based on an encoder-decoder transformer. Input audio is resampled to 16,000 Hertz (Hz) and converted to an 80-channel Log-magnitude Mel spectrogram using 25 ms windows with a 10 ms stride. The spectrogram is then normalized to a [-1, 1] range with near-zero mean. The encoder takes this Mel spectrogram as input and processes it. It first passes through two convolutional layers. Sinusoidal positional embeddings are added. It is then processed by a series of Transformer encoder blocks (with pre-activation residual connections). The encoder's output is layer normalized. The decoder is a standard transformer decoder. It has the same width and Transformer blocks as the encoder. It uses learned positional embeddings and tied input-output token representations (using the same weight matrix for both the input and output embeddings). It uses a byte-pair encoding tokenizer, of the same kind as used in GPT-2. English-only models use the GPT-2 vocabulary, while multilingual models employ a re-trained multilingual vocabulary with the same number of words. Special tokens are used to allow the decoder to perform multiple tasks: Tokens that denote language (one unique token per language). Tokens that specify task (<|transcribe|> or <|translate|>). Tokens that specify if no timestamps are present (<|notimestamps|>). If the token is not present, then the decoder predicts timestamps relative to the segment, and quantized to 20 ms intervals. <|nospeech|> for voice activity detection. <|startoftranscript|>, and <|endoftranscript|> . Any text that appears before <|startoftranscript|> is not generated by the decoder, but given to the decoder as context. Loss is only computed over non-contextual parts of the sequence, i.e. tokens between these two special tokens. == Training data == The training dataset consists of 680,000 hours of labeled audio-transcript pairs sourced from the internet using semi-supervised learning. This includes 117,000 hours in 96 non-English languages and 125,000 hours of X→English translation data, where X stands for any non-English language. Preprocessing involved standardization of transcripts, filtering to remove machine-generated transcripts using heuristics (e.g., punctuation, capitalization), language identification and matching with transcripts, fuzzy deduplication, and deduplication with evaluation datasets to avoid data contamination. Speechless segments were also included to allow voice activity detection training. For the files still remaining after the filtering process, audio files were then broken into 30-second segments paired with the subset of the transcript that occurs within that time. If this predicted spoken language differed from the language of the text transcript associated with the audio, that audio-transcript pair was not used for training the speech recognition models, but instead for training translation. The model was trained using the AdamW optimizer with gradient norm clipping and a linear learning rate decay with warmup, with batch size 256 segments. Training proceeded for 1 million updates (approximately 2-3 epochs). No data augmentation or regularization, except for the Large V2 model, which used SpecAugment, Stochastic Depth, and BPE Dropout. The training used data parallelism with float16, dynamic loss scaling, and activation checkpointing. === Post-training filtering === After training the first model, researchers ran it on different subsets of the training data, each representing a distinct source. Data sources were ranked by a combination of their error rate and size. Manual inspection of the top-ranked sources (high error, large size) helped determine if the source was low quality (e.g., partial transcriptions, inaccurate alignment). After training, it was fine-tuned to suppress the prediction of speaker names and low-quality sources were then removed. == Capacity == While Whisper does not outperform models which specialize in the LibriSpeech dataset, when tested across many datasets, it is more robust and makes 55.2% fewer errors than other models. Whisper has a differing error rate with respect to transcribing different languages, with a higher word error rate in languages not well-represented in the training data. The authors found that multi-task learning improved overall performance compared to models specialized to one task. They conjectured that the best Whisper model trained is still underfitting the dataset, and larger models and longer training can result in better models. Third-party evaluations have found varying levels of AI hallucination. A study of transcripts of public meetings found hallucinations in eight out of every 10 transcripts, while an engineer discovered hallucinations in "about half" of 100 hours of transcriptions and a developer identified them in "nearly every one" of 26,000 transcripts. A study of 13,140 short audio segments (averaging 10 seconds) found 187 hallucinations (1.4%), 38% of which generated text that could be harmful because it inserted false references to things like race, non-existent medications, or violent events that were not in the audio. == Applications == The model has been used as the base for many applications, such as a unified model for speech recognition and more general sound recognition. Whisper has also been integrated into the workflow of biomedical research. In 2025, a study on Alzheimer's disease detection used the model to transcribe spontaneous speech recordings. The transcripts that were generated by the model were combined with LLM vector embeddings and traditional classifiers to help classify the patients' health. Another application is when OVALYTICS incorporated Whisper to transcribe YouTube videos and automate content moderation systems, which improved its detection of offensive content. The model has also been used in academic libraries and cultral heritage institutions to generate transcripts and captions for their digitized audiovisual collections. In a 2025 case study, Emory University Libraries found that Whisper reduced the labor used in transcription by around 30-35%, shifting work from text creation to text correction. However, human review is still necessary to make sure accuracy, formatting, and accessibility are all standard.