COA Beat Assistant is a customized GPT model designed to read and analyze audit reports. It supports watchdog journalism by helping journalists interpret complex financial documents and extract relevant information for investigative stories.
Quispe Chequea verifies content in Spanish and Peruvian indigenous languages like Quechua, Aymara, and Awajún. It converts fact-checked information into audio for radio dissemination across nine regions. The platform leverages large language models for analysis and Tacotron 2 for voice synthesis, allowing indigenous interpreters to create voices that deliver credible news in local languages. It's used by trained regional broadcasters to ensure factual communication for Andean and Amazonian communities.
UAlter is a news reading assistant designed to reformat and condense information from Clarín’s news articles. It provides readers with six alternative formats: a summary of the article, a chronological timeline, highlights, a data table, a names index, and an FAQ. Each format offers a specific lens for readers to better understand complex news content, making data more digestible and accessible by breaking down key information and ordering it in ways that support efficient comprehension and engagement.
Lede AI is a company that uses artificial intelligence to assist news organizations, particularly local and independent newsrooms, by automating content creation for specific, routine topics. By parsing and transforming structured data (e.g., sports scores, public records) into readable, informative articles, Lede AI allows newsrooms to cover more ground with fewer resources, especially for stories with a high volume but relatively low complexity. The AI operates as a content production tool, optimizing time and labor for newsrooms with limited staff, making it especially useful for small to mid-sized publishers.
Weather Bot is an automation tool designed to gather data from the National Hurricane Center and the National Weather Service, identifying relevant weather alerts and generating reports. This AI-driven application aims to enhance the efficiency and accuracy of weather reporting, ensuring that critical alerts are delivered promptly to the community. By continuously monitoring data feeds, the bot can detect weather events and produce comprehensive reports that highlight essential details. Additionally, it streamlines the process by translating reports between languages, making information accessible to a wider audience. The generated reports are then published to a content management system and notifications are sent to relevant parties for further dissemination.
Clip2Story utilizes OpenAI's GPT-3.5 Turbo to transcribe news video clips and summarize the transcripts into initial story drafts. This prototype web application aims to reduce the workload for journalists by automating the creation of digital content from video sources, thus increasing efficiency and allowing reporters to focus on other important stories. The process begins with video uploads that are transcribed using the Trint API, followed by journalist review for accuracy. Once approved, the transcript is summarized, and relevant keywords are generated for the CMS. This system not only speeds up the reporting process but also enhances local news coverage by providing a quick turnaround for breaking stories.
The Minutes application scrapes, transcribes, and summarizes city council meetings from various video sources. It provides a searchable database of meeting content, allowing users to easily access and navigate important public discussions, thereby enhancing transparency and civic engagement.
TensorFlow is an open-source machine learning platform that offers a wide range of tools for developing, training, and deploying machine learning models. With capabilities for neural networks, natural language processing, computer vision, and more, it supports model building and deployment across devices. TensorFlow includes libraries and community resources, making it suitable for scalable applications in research and industry alike. Its flexible architecture is optimized for both large-scale and small-scale machine learning projects.
PyTorch is an open-source machine learning library primarily used for developing and training deep learning models. It provides a flexible platform for building neural networks using dynamic computation graphs, allowing developers to design and modify models on the fly. PyTorch supports both CPU and GPU-based computing, making it scalable for training large models. Its ecosystem also includes tools for automatic differentiation, optimization, and deployment of machine learning models.
Deepware is an AI-driven tool designed for deepfake detection, capable of analyzing visual and audio content to identify manipulation. It uses advanced machine learning algorithms to assess the authenticity of media, such as images, videos, and audio recordings, by scanning for inconsistencies, metadata anomalies, and compression traces. Deepware also offers real-time monitoring to automatically scan incoming media and forensic analysis to detect signs of tampering or editing.
User Guide
© IRCAI and UNESCO 2024