New platform enhancements automate and significantly accelerate the production of high-quality speech, audio, text, natural language, image and computer vision data, with multimodal and multi-domain support, for the faster deployment of mission-critical AI models supporting federal and enterprise applications.
AppTek, a leader in Artificial Intelligence (AI), Machine Learning (ML), Automatic Speech Recognition (ASR), Neural Machine Translation (NMT), Text-to-Speech (TTS) and Natural Language Processing / Understanding (NLP/U) technologies, today announced the expansion of its Workbench data labeling and annotation platform to include video labeling capabilities for computer vision models, in addition to its industry-leading ASR, NMT, TTS, and NLP/U data services.
AppTek is trusted by some of the world’s leading cloud and technology companies to power their conversational speech, text and image-based AI models through its managed data services program. To create the robust data sets required for these high-performing AI models, AppTek employs its proprietary Workbench, a secure, SaaS-based and cloud-based data annotation and labeling tool, which plugs into any data supply chain and fuses the human-machine relationship to produce high quality data while improving operational efficiencies up to 80% over manual processes. The platform is utilized by AppTek’s distributed workforce, which spans across 70+ countries, to generate custom ML data sets for bespoke AI models supporting both enterprise and federal customers. Workbench integrates AppTek’s advanced ASR and NMT engines to produce ‘tried and tested’ data sets through a customizable human-in-the-loop workflow while incorporating a collection of quality assurance tools for the scientific validation of models to help enterprises generate high quality data at scale.
Now AppTek expands its portfolio of data offerings to include the labelling of data for video tasks and the creation of multi-modal speech/video models with the addition of computer vision (CV) engines powered by partner IDenTV. The semi-automation of these engines dramatically accelerates the time to complete CV annotation tasks versus manual-only labeling processes, allowing for the faster deployment of vision models such as object detection, activity detection, scene detection, facial detection, video segmentation, rich metadata tagging, facial blurring, redaction of sensitive information, optical character recognition (OCR), and more. These computer vision classifiers combine with AppTek’s automatic speech recognition AI models to optimize and scale the data labelling and annotation workflow. By combining large amounts of multi-format and multi-domain audio, text and image data, customers can deploy bespoke models for new and innovative AI offerings.
“As federal and enterprise customers move rapidly to adopt AI, time and cost efficiencies are critical,” said Katie Nguyen, SVP of Data Operations at AppTek. “The complexity involved to ingest, label and annotate large amounts of multi-format audio, text and video data makes the process costly and time consuming. By supplementing AppTek’s high-performing speech and language models with video annotation capabilities, we can now deliver a new portfolio of scientifically tested and validated data sets at a fraction of the time it would take for manual annotation and at significantly less cost.”
“AppTek has teamed with system integrators and leading industry partners to optimize data science workflows with backend AI models for computer vision classifiers and automatic speech recognition to drive the Workbench platform,” said AppTek CEO Mudar Yaghi. “We continue to focus on speed to value for our customers, and these new enhancements to the Workbench will drive even more efficiencies and cost savings.”