This paper surveys significant research on the problem of rendering expressive music by means of AI techniques with an emphasis on Case-Based Reasoning. Following a brief overview discussing why we prefer listening to expressive music instead of lifeless synthesized music, we examine a representative selection of well-known approaches to expressive computer music performance with an emphasis on AI-related approaches. In the main part of the paper we focus on the existing CBR approaches to the problem of synthesizing expressive music, and particularly on TempoExpress, a case-based reasoning system developed at our Institute, for applying musically acceptable tempo transformations to monophonic audio recordings of musical performances. Finally we briefly describe an ongoing extension of our previous work consisting on complementing audio information with information of the gestures of the musician. Music is played through our bodies, therefore capturing the gesture of the performer is a fundamental aspect that has to be taken into account in future expressive music renderings. This paper is based on the “2011 Robert S. Engelmore Memorial Lecture” given by the first author at AAAI/IAAI 2011.

Given a current news event, we tackle the problem of generating plausible predictions of future events it might cause. We present a new methodology for modeling and predicting such future news events using machine learning and data mining techniques. Our Pundit algorithm generalizes examples of causality pairs to infer a causality predictor. To obtain precisely labeled causality examples, we mine 150 years of news articles and apply semantic natural language modeling techniques to headlines containing certain predefined causality patterns. For generalization, the model uses a vast number of world knowledge ontologies. Empirical evaluation on real news articles shows that our Pundit algorithm performs as well as non-expert humans.

The humour found in short jokes and their often equivalent newspaper cartoons graphic representations are often de­pendent on the results of ambiguity in human speech. The ambiguities can be unexpected and funny. Sometimes well-known ambiguities cooperatively repeated can also be funny. Captioned cartoons often derive their humour from an unexpected ambiguity that can be understood by a lis­tener who can automatically use world knowledge to re­solve the ambiguity. The question considered here is whether the listener can be a computational device as well as a human and the pragmatic difficulty of applying lin­guistic pragmatics to do so. Computational analysis of nat­ural language statements needs to successfully resolve am­biguous statements. Computerized understanding of dia­logue must not only include syntactic and semantic analy­sis, but also pragmatic analysis. Pragmatics includes an un­der­standing of the speaker’s intentions, the context of the utter­ance, and social implications of human communica­tion, both polite and hostile. Computational techniques can use restricted world knowledge in re­solving ambiguous lan­guage use. This paper considers the prag­matic difficulties in recognizing humour in short jokes as well as their repre­sentation in cartoons.

Collaborative filtering (CF) aims to predict users' ratings on items according to historical user-item preference data. In many real-world applications, preference data are usually sparse, which would make models overfit and fail to give accurate predictions. Recently, several research works show that by transferring knowledge from some manually selected source domains, the data sparseness problem could be mitigated. However for most cases, parts of source domain data are not consistent with the observations in the target domain, which may misguide the target domain model building. In this paper, we propose a novel criterion based on empirical prediction error and its variance to better capture the consistency across domains in CF settings. Consequently, we embed this criterion into a boosting framework to perform selective knowledge transfer. Comparing to several state-of-the-art methods, we show that our proposed selective transfer learning framework can significantly improve the accuracy of rating prediction tasks on several real-world recommendation tasks.

The hidden Markov model (HMM) is a widely-used generative model that copes with sequential data, assuming that each observation is conditioned on the state of a hidden Markov chain. In this paper, we derive a novel algorithm to cluster HMMs based on the hierarchical EM (HEM) algorithm. The proposed algorithm i) clusters a given collection of HMMs into groups of HMMs that are similar, in terms of the distributions they represent, and ii) characterizes each group by a "cluster center", i.e., a novel HMM that is representative for the group, in a manner that is consistent with the underlying generative model of the HMM. To cope with intractable inference in the E-step, the HEM algorithm is formulated as a variational optimization problem, and efficiently solved for the HMM case by leveraging an appropriate variational approximation. The benefits of the proposed algorithm, which we call variational HEM (VHEM), are demonstrated on several tasks involving time-series data, such as hierarchical clustering of motion capture sequences, and automatic annotation and retrieval of music and of online hand-writing data, showing improvements over current methods. In particular, our variational HEM algorithm effectively leverages large amounts of data when learning annotation models by using an efficient hierarchical estimation procedure, which reduces learning times and memory requirements, while improving model robustness through better regularization.

Many latent (factorized) models have been proposed for recommendation tasks like collaborative filtering and for ranking tasks like document or image retrieval and annotation. Common to all those methods is that during inference the items are scored independently by their similarity to the query in the latent embedding space. The structure of the ranked list (i.e. considering the set of items returned as a whole) is not taken into account. This can be a problem because the set of top predictions can be either too diverse (contain results that contradict each other) or are not diverse enough. In this paper we introduce a method for learning latent structured rankings that improves over existing methods by providing the right blend of predictions at the top of the ranked list. Particular emphasis is put on making this method scalable. Empirical results on large scale image annotation and music recommendation tasks show improvements over existing approaches.

Previous work on recommender systems mainly focus on fitting the ratings provided by users. However, the response patterns, i.e., some items are rated while others not, are generally ignored. We argue that failing to observe such response patterns can lead to biased parameter estimation and sub-optimal model performance. Although several pieces of work have tried to model users' response patterns, they miss the effectiveness and interpretability of the successful matrix factorization collaborative filtering approaches. To bridge the gap, in this paper, we unify explicit response models and PMF to establish the Response Aware Probabilistic Matrix Factorization (RAPMF) framework. We show that RAPMF subsumes PMF as a special case. Empirically we demonstrate the merits of RAPMF from various aspects.

This paper considers stochastic bandits with side observations, a model that accounts for both the exploration/exploitation dilemma and relationships between arms. In this setting, after pulling an arm i, the decision maker also observes the rewards for some other actions related to i. We will see that this model is suited to content recommendation in social networks, where users' reactions may be endorsed or not by their friends. We provide efficient algorithms based on upper confidence bounds (UCBs) to leverage this additional information and derive new bounds improving on standard regret guarantees. We also evaluate these policies in the context of movie recommendation in social networks: experiments on real datasets show substantial learning rate speedups ranging from 2.2x to 14x on dense networks.

Creating a musical fitness function is largely subjective and can be critically affected by the designer's biases. Previous attempts to create such functions for use in genetic algorithms lack scope or are prejudiced to a certain genre of music. They also are limited to producing music strictly in the style determined by the programmer. We show in this paper that musical feature extractors, which avoid the challenges of qualitative judgment, enable creation of a multi-objective function for direct music production. The main result is that the multi-objective fitness function enables creation of music with varying identifiable styles. To demonstrate this, we use three different multi-objective fitness functions to create three distinct sets of musical melodies. We then evaluate the distinctness of these sets using three different approaches: a set of traditional computational clustering metrics; a survey of non-musicians; and analysis by three trained musicians.

The word autonomy is often used in the discussion of software-based music-generating systems. Whilst the term conveys a very clear concept — the sense of self-determination of a system — attempts to formalise autonomy are at an early stage, and the term is subject to a range of interpretations when practically applied. We consider how the evaluation of music-generating systems will be enhanced by a clearer understanding of autonomy and its application to music. We discuss existing definitions and approaches to quantifying autonomy and consider, through a series of examples, the information that is required in order to make precise formal judgements about autonomy, and the identification of relevant levels at which the principle of autonomy applies in music. We conclude that automated measures can supplement human evaluation of autonomy, but that (a) automated measures must be supported by sound reasoning about the features and timescales used in the measurement, and (b) they are improved by a having knowledge of the internal working of the system, rather than taking a black box approach. We consider multi-dimensional representations of system behaviour that may capture a richer sense of the notion of autonomy. Finally, we propose an approach to automatically probing music systems as a means of determining an autonomy `portrait'.