In this paper, we propose a new approach for recommender systems based on target tracking by Kalman filtering. We assume that users and their seen resources are vectors in the multidimensional space of the categories of the resources. Knowing this space, we propose an algorithm based on a Kalman filter to track users and to predict the best prediction of their future position in the recommendation space.

While traditional research on text clustering has largely focused on grouping documents by topic, it is conceivable that a user may want to cluster documents along other dimensions, such as the author's mood, gender, age, or sentiment. Without knowing the user's intention, a clustering algorithm will only group documents along the most prominent dimension, which may not be the one the user desires. To address the problem of clustering documents along the user-desired dimension, previous work has focused on learning a similarity metric from data manually annotated with the user's intention or having a human construct a feature space in an interactive manner during the clustering process. With the goal of reducing reliance on human knowledge for fine-tuning the similarity function or selecting the relevant features required by these approaches, we propose a novel active clustering algorithm, which allows a user to easily select the dimension along which she wants to cluster the documents by inspecting only a small number of words. We demonstrate the viability of our algorithm on a variety of commonly-used sentiment datasets.

Missing data estimation is an important challenge with high-dimensional data arranged in the form of a matrix. Typically this data matrix is transposable, meaning that either the rows, columns or both can be treated as features. To model transposable data, we present a modification of the matrix-variate normal, the mean-restricted matrix-variate normal, in which the rows and columns each have a separate mean vector and covariance matrix. By placing additive penalties on the inverse covariance matrices of the rows and columns, these so-called transposable regularized covariance models allow for maximum likelihood estimation of the mean and nonsingular covariance matrices. Using these models, we formulate EM-type algorithms for missing data imputation in both the multivariate and transposable frameworks. We present theoretical results exploiting the structure of our transposable models that allow these models and imputation methods to be applied to high-dimensional data. Simulations and results on microarray data and the Netflix data show that these imputation techniques often outperform existing methods and offer a greater degree of flexibility.

Journalists and editors work under pressure to collect relevant details and background information about specific events. They spend a significant amount of time sifting through documents and finding new information such as facts, opinions or stakeholders (i.e. people, places and organizations that have a stake in the news). Spotting them is a tedious and cognitively intense process. One task, essential to this process, is to find and keep track of stakeholders. This task is taxing cognitively and in terms of memory. Tell Me More offers an automatic aid to this task. Tell Me More is a system that, given a seed story, mines the web for similar stories reported by different sources and selects only those stories which offer new information with respect to that original seed story. Much like a journalist, the task of detecting named entities is central to its success. In this paper we briefly describe Tell Me More and, in particular, we focus on Tell Me More's entity detection component. We describe an approach that combines off-the-shelf named entity recognizers (NERs) with WPED, an in-house publicly available NER that uses Wikipedia as its knowledge base. We show significant increase in precision scores with respect to traditional NERs. Lastly, we present an overall evaluation of Tell Me More using this approach.

In this paper we propose a middleware to reuse multimedia resources in order to produce new types of multimedia artifacts. In this work we adopt some basic concepts of film theory, such as the notions of plot, fabula and, in particular, diegetic time. The techniques we use are located within the area of artificial intelligence, using an explicit representation of time. The middleware consists of several modules, some devoted to the semantic annotation of multimedia components, and others to their visualization. Some modules regard the analysis of temporal connectivity and consistency of events. From a methodological point of view, an important module of the middleware contains the representation of a story (time of the narration and time of the story) and the temporal reasoning services, which are both implemented using a logic programming language (Flora2). Finally, there is a module in the middleware that translates the logical representation (in Flora2 language) into SMIL language, which allows the use of the final composition by a standard player.

A method is presented for converting between Guitar Hero difficulty levels by treating the problem as one of machine translation, with the different difficulties as different languages. The Guitar Hero I and II discs provide aligned corpora with which to train bigram-based language models and translation models. Given an Expert sequence, the model can create sequences of Hard, Medium, or Easy difficulty that retain the feel of the original, while obeying heuristics typical of those difficulties. Training the model requires a single pass through the corpus, while translation is quadratic in the length of the Expert sequence. The method outperforms a recurrent neural network in producing sequences that match the hand-designed levels. The method may make it easier for amateurs to produce content for the Rock Band Network.

Chris Jurney (Lead Programmer, Double Fine Productions) Sumit Basu (Microsoft Research) Chris Jurney is a rock and roll experimental game For those who can play an instrument or have a respectable programmer at Double Fine Productions, with 11 singing voice, music can be a wonderful years experience in games and simulation. He has means of creative expression, social engagement, shipped 4 titles in the games industry: Company of and fun. For many others, though, it can be frustrating Heroes, Frontline: Fuel of War, Dawn of War 2, and and inaccessible: even if an inspired youth Brutal Legend. Jurney frequently speaks on the topic has great musical ideas, she may not have the of game AI, having presented at the Game Developers knowledge or ability to get her latest song out from Conference (GDC), GDC China, Columbia her head and into her MP3 player. In this talk, Basu will show three vignettes of how he and his colleagues University, the University of Pennsylvania, and the have used interactive machine learning to New Jersey and Philadelphia chapters of the International extend the creative reach of aspiring musicians: a Game Developers Association (IGDA).

This paper describes our current approach in implementing computational improvisational micro-agents. This approach is intended to foster bottom-up research to better understand how to build more complex agent behaviors in a theatrical improvisational setting. Micro-agent designs are based on our current findings in a multi-year study focused on studying real life theatrical improvisers with an aim towards better understanding the cognition employed inimprovisation at the individual and group level. It also introduces a key architectural component from the domain of fuzzy logic that enables us to clearly represent some of our current findings.

Narrative, and in particular storytelling, is an important part of the human experience. Consequently, computational systems that can reason about narrative can be more effective communicators, entertainers, educators, and trainers. One of the central challenges in computational narrative reasoning is narrative generation, the automated creation of meaningful event sequences. There are many factors -- logical and aesthetic -- that contribute to the success of a narrative artifact. Central to this success is its understandability. We argue that the following two attributes of narratives are universal: (a) the logical causal progression of plot, and (b) character believability. Character believability is the perception by the audience that the actions performed by characters do not negatively impact the audience's suspension of disbelief. Specifically, characters must be perceived by the audience to be intentional agents. In this article, we explore the use of refinement search as a technique for solving the narrative generation problem -- to find a sound and believable sequence of character actions that transforms an initial world state into a world state in which goal propositions hold. We describe a novel refinement search planning algorithm -- the Intent-based Partial Order Causal Link (IPOCL) planner -- that, in addition to creating causally sound plot progression, reasons about character intentionality by identifying possible character goals that explain their actions and creating plan structures that explain why those characters commit to their goals. We present the results of an empirical evaluation that demonstrates that narrative plans generated by the IPOCL algorithm support audience comprehension of character intentions better than plans generated by conventional partial-order planners.

In many applications involving multi-media data, the definition of similarity between items is integral to several key tasks, e.g., nearest-neighbor retrieval, classification, and recommendation. Data in such regimes typically exhibits multiple modalities, such as acoustic and visual content of video. Integrating such heterogeneous data to form a holistic similarity space is therefore a key challenge to be overcome in many real-world applications. We present a novel multiple kernel learning technique for integrating heterogeneous data into a single, unified similarity space. Our algorithm learns an optimal ensemble of kernel transfor- mations which conform to measurements of human perceptual similarity, as expressed by relative comparisons. To cope with the ubiquitous problems of subjectivity and inconsistency in multi- media similarity, we develop graph-based techniques to filter similarity measurements, resulting in a simplified and robust training procedure.