Shriphani Palakodety – författare

The ebook edition of this title is Open Access and freely available to read online.Artificial intelligence (AI) depends on data. In sensitive domains – such as healthcare, security, finance, and many more – there is therefore tension between unleashing the power of AI and maintaining the confidentiality and security of the relevant data.This book – intended for researchers in academia and R&D engineers in industry – explains how advances in three areas—AI, privacy-preserving techniques, and acceleration—allow us to achieve the dream of high performance privacy-preserving AI. It also discusses applications enabled by this emerging interplay.The book covers techniques, specifically secure multi-party computation and homomorphic encryption, that provide complexity theoretic security guarantees even with a single data point. These techniques have traditionally been too slow for real-world usage, and the challenge is heightened with the large sizes of today's state-of-the-art neural networks, including large language models (LLMs). This book does not cover techniques like differential privacy that only concern statistical anonymization of data points.

This book focuses on methods that are unsupervised or require minimal supervision—vital in the low-resource domain. Over the past few years, rapid growth in Internet access across the globe has resulted in an explosion in user-generated text content in social media platforms. This effect is significantly pronounced in linguistically diverse areas of the world like South Asia, where over 400 million people regularly access social media platforms. YouTube, Facebook, and Twitter report a monthly active user base in excess of 200 million from this region. Natural language processing (NLP) research and publicly available resources such as models and corpora prioritize Web content authored primarily by a Western user base. Such content is authored in English by a user base fluent in the language and can be processed by a broad range of off-the-shelf NLP tools. In contrast, text from linguistically diverse regions features high levels of multilinguality, code-switching, and varied languageskill levels. Resources like corpora and models are also scarce. Due to these factors, newer methods are needed to process such text.This book is designed for NLP practitioners well versed in recent advances in the field but unfamiliar with the landscape of low-resource multilingual NLP. The contents of this book introduce the various challenges associated with social media content, quantify these issues, and provide solutions and intuition. When possible, the methods discussed are evaluated on real-world social media data sets to emphasize their robustness to the noisy nature of the social media environment.On completion of the book, the reader will be well-versed with the complexity of text-mining in multilingual, low-resource environments; will be aware of a broad set of off-the-shelf tools that can be applied to various problems; and will be able to conduct sophisticated analyses of such text.