* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download INF 5820 Language technological applications
Existential risk from artificial general intelligence wikipedia , lookup
Machine learning wikipedia , lookup
Ethics of artificial intelligence wikipedia , lookup
Person of Interest (TV series) wikipedia , lookup
Philosophy of artificial intelligence wikipedia , lookup
Knowledge representation and reasoning wikipedia , lookup
Wizard of Oz experiment wikipedia , lookup
INF 5820 LANGUAGE TECHNOLOGICAL APPLICATIONS Fall 2012 Today 1. Hour: Course overview General Machine translation Spoken Dialogue Systems 2. Hour: Starting Spoken Dialogue Systems INF5820 http://www.uio.no/studier/emner/matnat/ifi/INF58 20/index-eng.xml Builds on INF4820 Algorithms for Artificial Intelligence and Natural Language Processing, might be taken in parallel Alternates with INF5830 Natural Language Processing Two applications Spoken Dialogue Systems First part of semester Pierre Lison (plison) Machine translation Second part Jan Tore Lønning (jtl) Language Technology Group (LTG), 7. floor Classes Fridays 10.15-12 Lectures OJD 2458 Postscript Mondays 12.15-14 Group/lectures OJD 2458 Postscript Obligatory assignments 3 obligatory assignments Dialogue 1: 26 September Dialogue 2: 17 October Translation: final,15 November PhDs: Paper presentation Exam Written exam 18 December at 0900 Spoken Dialogue Systems Pierre Lison Machine Translation Jan Tore Lønning Machine Translation Active research field since 1949, In the 1950s MT was not only the most important NLP/computational linguistics field, it was the only one IBM 1954 press release Interest, results and funding have varied over time Today: Fully-automatic text-translation: Systran, Google Speech-translation: commercial Aid for professional translators: trados Two types of approaches to NLP Rule-based Empirical Build a declarative model using Linguistics Logic Algorithms How does it fit data? Start with naturally occurring text What information can we get? Statistics/Machine learning Use this to reproduce the examples Applied to MT Rule-based Which linguistic information should be included, syntax? semantics? Approaches Direct translation Syntax-based transfer Semantic-based transfer .. Empirical Example-based translation Statistical machine translation (SMT) Word-based Phrase-based Syntactic Figure 25.8 What we will study 1. 2. 3. MT overview MT evaluation Statistical MT, 4. 5. The main part Rule-based MT with semantic transfer Hybrid methods Literature J&M, ch. 25 Koehn, in particular, Part II Core methods: ch. 4-8 A few papers Recommended prior knowledge INF4820 - Algorithms for artificial intelligence and natural language processing Some knowledge in statistics is an advantage (Wrong text on English page) In particular: Probability theory N-grams Hidden Markov Models Dynamic Programming Assignments Building an SMT system with open-source software Experiment with different settings and parameters Evaluation One obligatory assignment with several parts