The paper discusses the issue of hallucinations in natural language processing (NLP) and speech recognition. With the advancement of deep learning models like transformers, performance in NLP tasks such as question answering, machine translation, and summarization has improved significantly. However, these models can generate text that is unfaithful to the input or contains information not supported by the source. This phenomenon is called hallucination. Similar issues are observed in speech recognition systems, where the output text may not match the corresponding speech signal. The paper introduces the problem of hallucinations in various NLP and speech tasks, categorizes them into intrinsic and extrinsic types, and describes techniques to quantify them. It also presents recent methods to mitigate hallucinations in these tasks. The paper highlights the importance of addressing hallucinations in NLP and speech, as they can lead to incorrect or unverified information being generated. The tutorial covers the causes of hallucinations in NLP and speech tasks, provides examples, and discusses strategies for mitigating them. It also presents a generic framework for understanding and addressing hallucinations in different tasks. The paper emphasizes the need for further research and development to improve the accuracy and reliability of NLP and speech systems. The tutorial is aimed at researchers and practitioners in the field of NLP and speech processing, who are interested in understanding and mitigating hallucinations in their work. The authors are researchers at IBM Research with expertise in NLP and speech processing. The tutorial is structured to cover the problem of hallucinations in NLP and speech, the causes and effects of hallucinations, and strategies for mitigating them. The tutorial is designed to be informative and practical, providing a comprehensive overview of the topic.The paper discusses the issue of hallucinations in natural language processing (NLP) and speech recognition. With the advancement of deep learning models like transformers, performance in NLP tasks such as question answering, machine translation, and summarization has improved significantly. However, these models can generate text that is unfaithful to the input or contains information not supported by the source. This phenomenon is called hallucination. Similar issues are observed in speech recognition systems, where the output text may not match the corresponding speech signal. The paper introduces the problem of hallucinations in various NLP and speech tasks, categorizes them into intrinsic and extrinsic types, and describes techniques to quantify them. It also presents recent methods to mitigate hallucinations in these tasks. The paper highlights the importance of addressing hallucinations in NLP and speech, as they can lead to incorrect or unverified information being generated. The tutorial covers the causes of hallucinations in NLP and speech tasks, provides examples, and discusses strategies for mitigating them. It also presents a generic framework for understanding and addressing hallucinations in different tasks. The paper emphasizes the need for further research and development to improve the accuracy and reliability of NLP and speech systems. The tutorial is aimed at researchers and practitioners in the field of NLP and speech processing, who are interested in understanding and mitigating hallucinations in their work. The authors are researchers at IBM Research with expertise in NLP and speech processing. The tutorial is structured to cover the problem of hallucinations in NLP and speech, the causes and effects of hallucinations, and strategies for mitigating them. The tutorial is designed to be informative and practical, providing a comprehensive overview of the topic.