Mini Workshop on Music and Language Models
QMUL School of Electronic Engineering and Computer Science
Centre for Digital Music Seminar Series
Special Lectures by Keiji Hirata and Satoshi Tojo
Date/time: 3pm-5pm, Thursday, 7 December 2017
Location: BR (Bancroft Road) 3.02
Open to academics, students, alumni, public; all welcome. Admission is FREE, no pre-booking required.
We are extremely pleased to be hosting a visit by three visitors from Japan, Professor Keiji Hirata of Future University Hakodate, Professor Satoshi Tojo of the Japan Advanced Institute of Science and Technology, and Dr. Masatoshi Hamanaka of RIKEN. In a sequence of lectures at Queen Mary University of London on Thursday, 7 December 2017, Professors Hirata and Tojo will speak about the semantic and linguistic structure of music, and work at the forefront of machine learning, artificial intelligence, music and language, and computational music theory.
Towards Formalisation of Music Theory
Speaker: Professor Keiji Hirata, Future University Hakodate
Abstract: Driven by the recent big success of the machine learning techniques in the natural language processing and computer vision fields, many music information researchers are currently attempting to naïvely apply such techniques to the tasks of classifying, understanding, and generating music information. However, we believe that music as a communication media is essentially different from language and vision from the semiotics point of view. Hence, to manipulate symbols in music at the superficial level, we first have to reveal the semantic structure of music then apply such techniques to the tasks in a proper manner. In this talk, we will present a brief summary of our research on formalising the time-span tree of GTTM, in which we hypothesize that the time-span tree represents the semantic structure of music in a sense. Then, I shall discuss how we can partly restructure music theory in the context of machine learning, and find a clue to computational music theory.
Biosketch: Keiji Hirata received the Doctor of Engineering degree from the University of Tokyo in 1987. He joined NTT Basic Research Laboratories in 1987 (later renamed to NTT Communication Science Laboratories) and Future University Hakodate as professor in 2011. From 1990 to 1993, under the 5th Generation Computer Project, he joined the Institute for New Generation Computer Technology (ICOT) and was actively engaged in the research and development of parallel inference machines. His research interests include computational music theory, algebraic frameworks for music creation, and music mobility as a service (MaaS).
More about Keiji Hirata at www.fun.ac.jp/~hirata.
Linguistic Model of Music
Speaker: Professor Satoshi Tojo, Japan Advanced Institute of Science and Technology
Abstract: Music and language are considered to share the common biological origin. Although the meaningful part became independent of music and resulted in language, why do we human beings still possess music? In this talk, we illustrate the similarity in our recognition process of music and language, referring to automata theory. Furthermore, we show some examples of long-distance dependencies in chord sequences and contend that there exists Chomskian context-free grammar in music. The talk will include two digressions: one is what is agent communication. If music does not contribute to meaning, what is transmitted among agents? We show a logical paradigm of agent communication and discuss what is the understanding of music. The other digression concerns the iterated learning model by Simon Kirby. If we can find context-free grammar rules statistically, what are the roles of intermediate categories, that is, nodes in tree structure. Also, if computers can find rules, can they compose music? We place these as open questions for future issues of linguistic musicology.
Biosketch: Satoshi Tojo received a Bachelor of Engineering, Master of Engineering, and Doctor of Engineering degrees from the University of Tokyo, Japan. He joined Mitsubishi Research Institute, Inc. (MRI) in 1983, and the Japan Advanced Institute of Science and Technology (JAIST), Ishikawa, Japan, as associate professor in 1995 and became professor in 2000. His research interest is centered on formal semantics of natural language, and logic in artificial intelligence, including knowledge and belief of rational agents, and grammar acquisition, as well as linguistic models of music.
More about Satoshi Tojo at www.jaist.ac.jp/~tojo
Further information on the mini-workshop at http://mupae.blogspot.co.uk/2017/11/music-and-language-models.html