He is presently the lead guest editor of "Sustainability," focusing on the applications of machine learning and artificial intelligence in …

Entity Linking is the task of matching a mention to an entity in a given knowledge base (KB). It contributes to annotating a massive amount of documents existing on the Web to harness new facts about their matched entities. However, existing Entity Linking systems focus on developing models that are typically domain-dependent and robust only to a particular knowledge base on which they have been trained. The performance is not as adequate when being evaluated on documents and knowledge bases from different domains. Approaches based on pre-trained language models, such as Wu et al. (2020), attempt to solve the problem using a zero-shot setup, illustrating some potential when evaluated on a general-domain KB. Nevertheless, the performance is not equivalent when evaluated on a domain-specific KB. To allow for more accurate Entity Linking across different domains, we propose our framework: Cross-Domain Neural Entity Linking (CDNEL). Our objective is to have a single system that enables simultaneous linking to both the general-domain KB and the domain-specific KB. CDNEL works by learning a joint representation space for these knowledge bases from different domains. It is evaluated using the external Entity Linking dataset (Zeshel) constructed by Logeswaran et al. (2019) and the Reddit dataset collected by Botzer et al. (2021), to compare our proposed method with the state-of-the-art results. The proposed framework uses different types of datasets for fine-tuning, resulting in different model variants of CDNEL. When evaluated on four domains included in the Zeshel dataset, these variants achieve an average precision gain of 9%.

Data is the lifeblood of modern machine learning systems, including for those in Music Information Retrieval (MIR). However, MIR has long been mired by small datasets and unreliable labels. In this work, we propose to break this bottleneck using generative modeling. By pipelining a generative model of notes (Coconet trained on Bach Chorales) with a structured synthesis model of chamber ensembles (MIDI-DDSP trained on URMP), we demonstrate a system capable of producing unlimited amounts of realistic chorale music with rich annotations including mixes, stems, MIDI, note-level performance attributes (staccato, vibrato, etc.), and even fine-grained synthesis parameters (pitch, amplitude, etc.). We call this system the Chamber Ensemble Generator (CEG), and use it to generate a large dataset of chorales from four different chamber ensembles (CocoChorales). We demonstrate that data generated using our approach improves state-of-the-art models for music transcription and source separation, and we release both the system and the dataset as an open-source foundation for future work in the MIR community.

Recent models can generate fluent and grammatical synthetic reviews while accurately predicting user ratings. The generated reviews, expressing users' estimated opinions towards related products, are often viewed as natural language 'rationales' for the jointly predicted rating. However, previous studies found that existing models often generate repetitive, universally applicable, and generic explanations, resulting in uninformative rationales. Further, our analysis shows that previous models' generated content often contain factual hallucinations. These issues call for novel solutions that could generate both informative and factually grounded explanations. Inspired by recent success in using retrieved content in addition to parametric knowledge for generation, we propose to augment the generator with a personalized retriever, where the retriever's output serves as external knowledge for enhancing the generator. Experiments on Yelp, TripAdvisor, and Amazon Movie Reviews dataset show our model could generate explanations that more reliably entail existing reviews, are more diverse, and are rated more informative by human evaluators.

Unmanned Aerial Vehicles grow more popular by the day and applications for them are crossing boundaries of science and industry, with everything from aerial photography to package delivery to disaster management benefiting from the technology. But before they become commonplace, there are challenges to be solved to make them reliable and safe. The following paper discusses the challenges associated with the precision landing of an Unmanned Aerial Vehicle, including methods for sensing and control and their merits and shortcomings for various applications.

As UAV technology improves, more uses have been found for these versatile autonomous vehicles, from surveillance to aerial photography, to package delivery, and each of these applications poses unique challenges. This paper implements a solution for one such challenge: To land on a moving target. This problem has been addressed before with varying degrees of success, however, most implementations focus on indoor applications. Outdoor poses greater challenges in the form of variables such as wind and lighting, and outdoor drones are heavier and more susceptible to inertial effects. Our approach is purely vision based, using a monocular camera and fiducial markers to localize the drone and a PID control to follow and land on the platform.

Creative artificial intelligence is the latest and, in some ways, most surprising and exhilarating art form in the world. It also isn't fully formed yet. That tension is causing some confusion. If you're familiar at all with the use of creative artificial intelligence, you probably know it through one of the popular text-to-image AI applications, which use sprawling databases of existing imagery to convert a written prompt into a new picture. DALL-E 2 from OpenAI is the best known, but more recent and arguably cooler applications include Midjourney and Stable Diffusion.

Kee Riches, a 23-year-old L.A. rapper, was shot and killed in Compton on Saturday night. Riches, whose real name is Kian Nellum, was shot along with another man -- 29-year-old Robert Leflore Jr. -- around 9:40 p.m. on the 1500 block of S. Chester Avenue in Compton, according to the L.A. County Sheriff's Department and L.A. County Medical Examiner-Coroner records. Tributes poured in across the artist's social media accounts upon word of his death. The "2 Live" and "Westside Lady" rapper was known in the area for his love of his community and drive to build it up, much like slain rapper Nipsey Hussle, who was gunned down in 2019. Riches previously told L.A. Taco that the Crenshaw hero, whom he described as "the embodiment of a street soldier, a real hustler," left a similar impact on his own life.

Artificial intelligence developed in Kyiv is taking over one of the most treasured roles in film, as James Earl Jones steps back as the voice of Darth Vader. The Star Wars actor, 91, was helped to reach the chilling heights of his performance 45 years ago by the Ukrainian startup Respeecher in the recent Obi-Wan Kenobi series as the company worked with Jones and clips of his past performances. The AI "clone" of his voice could then be layered over lines read by another actor to create a Darth Vader who sounds more real than the real thing. "Lucasfilm came to us, essentially through word of mouth,", said Dmytro Bielievtsov, Respeecher's chief technical officer. "Someone posted about our tech in an internal sound engineering chat, and it got picked up."

Darth Vader, the villain of the Star Wars franchise, is now voiced by artificial intelligence after the retirement of actor James Earl Jones. Jones, who is 91 years old, has voiced the helmeted menace since 1977's Star Wars: Episode IV – A New Hope (originally titled Star Wars). His voice as Darth Vader was last heard in the 2019 film The Rise of Skywalker. The space opera will now use an AI replication of Jones's voice, created by Ukrainian start-up Respeecher. The voice was first heard in the show Obi Wan Kenobi.