Speech communication refers to the process of transferring information from one person to another by speaking in a specific language. It is a highly inter-disciplinary subject, which is related to physiology, linguistics, phonetics, signal processing, and computer science. Many interesting and impactful computer applications have been developed to enable and improve human-computer and human-human speech communications. In this course, students will learn the scientific fundamentals underlying human speech communication, the basic techniques of computer speech and language processing, state-of-the-art spoken language technologies and their applications. Advanced research topics and future directions will also be discussed.

Tsinghua University

Machine learning is a scientific discipline concerned with the design and development of algorithms that allow computers to evolve behaviors like human being. It is the core of the artificial intelligence and is rapidly applied to various areas nowadays, such as search and social network. This course will introduce the fundamental of machine learning techniques and explain its implication in the machine translation. We aim to allow students to master key skills and to be able to develop novel algorithms in machine learning.

Tsinghua University

This course is set for graduate and senior undergraduate students, who have a basic background in statistics. The course teaches the fundamental theory and algorithms behind machine learning (ML), a key issue in the artificial intelligence. It aims to equip students with a deeper understanding and adept skills in the field of ML, to solve various pragmatic questions in broader areas. ML has a wide range of applications, including web search, social network, speech recognition, image processing, robotics and finance. We choose the application of statistical machine translation to demonstrate how the ML study is motivated and applied. Machine translation is one of the hottest area nowadays, with governments and large companies such as Microsoft, Baidu and Google investing huge resources to the task. The syllabus contains both theory and practice: First, we describe the training criteria (Bayesian decision rule, maximum likelihood), statistical models (CRF, SVM, HMM and neural network) and training algorithms (EM and Gibbs Sampling); Then, advanced machine translation methods are introduced based on the statistical approaches (word alignment, phrase training, decoding and optimization). The course is mainly conducted through lectures and discussions. The students are encouraged to join research projects to realize and to develop novel ideas.

Tsinghua University

Course Contents
Block 1: Machine Translation
Rule-based machine translation Statistical/hybrid machine translation
Block 2: Natural language human-computer interaction
Natural language human-computer interaction: Dialogue phenomena Approaches to computational dialogue modeling and dialogue management Comparison of dialogue systems for various applications User modeling, usability testing and evaluation Learning dialogue strategies

University of Saarland