In the last 10-15 years, NLP techniques have proven effective in a number of forensic applications, such as author profiling and deception detection. The tutorial will show the experience of DeCour - DEception in COURts -, a corpus constituted by transcripts of hearings held in four Italian courts, which was employed for a typical task of text classification carried out thorough stylometric techniques, in order to distinguish false from truthful statements. The process will be examined with particular attention to the methods employed, from the data collection, through the preprocessing and the feature selection. In the end, the data analysis and the results will be discussed.

Biography
Tommaso Fornaciari is an Investigative Psychologist with the Italian National Police. Since obtaining his Ph.D. at the University of Trento, he has carried out research activities in forensic linguistics, publishing studies in which computational methods are employed with the aim of detecting deception in text and in transcripts of spoken language from criminal proceedings. He presently works at the Department of Public Security of the Italian Ministry of the Interior, engaged in research and technological innovation for public security. Prior to that, he worked at the Forensic Science Police Service, where he dealt with criminal analysis, mostly regarding violent murders.

Machine translation has a long and bumpy history. Automatic translation of human languages has been one of the goals since the beginning of the development of computers and digital technology. Starting initially with ideas of deciphering foreign languages, computational linguists quickly moved on with translation engines based on handwritten grammars. Exaggerated expectations lead to major disappointments mocking machine translation as an object of ridicule. The advent of statistical machine translation was received with skepticism until it lead to scientific and commercial success. Today, translation services are accepted and widely used gisting tools but progress does not stop here. The wave of deep learning and neural machine translation swipes over language technology and promises another jump forward. Constantly, new ideas and extensions are proposed making today's MT research an exciting enterprise with a bright future.

In this tutorial, I will present the basic concepts of neural MT as the new state-of-the-art in automatic translation. We will look at the common architecture of attention-based sequence-to-se quence models and include an overview of useful extensions and practical tricks. The tutorial will mention available tools and resources to make it easy to get started with hands-on experience on real-world data. We do not require any deep understanding of neural networks and machine learning as a background but focus on a gentle introduction of models and techniques.

Biography
Jörg Tiedemann works as a professor of language technology at the University of Helsinki since August 2015. He received his Ph.D. in computational linguistics from Uppsala University in 2003. His work is mainly focused on machine translation, question answering and data mining from multilingual resources. During his Ph.D. he spent a year at the University of Edinburgh and worked as a post-doctoral researcher at the University of Groningen. He maintains OPUS, the world's largest open collection of parallel corpora and more details about his research can be found at http://blogs.helsinki.fi/tiedeman/