It is often essential to create datasets with foreseeable characteristics. For the design and testing of advanced spatiotemporal pattern mining algorithms,adaptable and large datasets are needed. In this paper, we present a synthetic dataset generator, ERMO-DG, that is intended for creating spatiotemporal patterns. Generated datasets consist of spatiotemporal object instances of different feature types, where these instances are represented by spatial regions evolving over time. The generator allows researchers to systematically create spatiotemporal datasets with predictable characteristics such as number of patterns, cardinality of patterns, velocity, acceleration, lifetime and spatialareas of instances.

Strategy mining is a new area of research about discovering strategies for decision-making. It is motivated by how similarity is assessed in retrospect in law. In the legal domain, when both case facts and court decisions are present, it is often useful to assess similarity by accounting for both case facts and case outcomes. In this paper, we formulate the strategy mining problem as a clustering problem with the goal of finding clusters that represent disparate conditional dependency of decision labels on other features. Existing clustering algorithms are inappropriate to cluster dependency because they either assume feature independence, such as K-means, or only consider the co-occurrence of features without explicitly modeling the special dependency of the decision label on other features, such as Latent Dirichlet Allocation (LDA). We propose an Expectation Maximization (EM) style unsupervised learning algorithm for dependency clustering. Like EM, our algorithm is grounded in statistical learning theory. It minimizes the empirical risk of decision tree learning. Unlike other clustering algorithms, our algorithm is irrelevant-feature resistant, and its learned clusters modeled by decision trees are strongly interpretable and predictive. We systematically evaluate both the convergence property and solution quality of our algorithm using a common law dataset comprised of actual cases. Experimental results show that our algorithm significantly outperforms K-means and LDA on clustering dependency

In the era of Big Data, knowledge integration is key for tasks such as social media aggregation, opinion mining, and cyber-issue detection. The integration of different kinds of knowledge coming from multiple sources, however, is often a problematic issue as it either requires a lot of manual effort in defining aggregation rules or suffers from noise generated by automatic integration techniques. In this work, we propose a method based on conceptual primitives for efficiently integrating pieces of knowledge coming from different common and common-sense resources, which we test in the field of concept-level sentiment analysis.

Learning rules is a common way of extracting usefulinformation from knowledge or data bases. Many ofsuch data sets contain numerical attributes. However,approaches like ILP or association rule mining are optimizedfor data with categorical values, and consideringnumerical attributes is expensive. In this paper,we present an extension to the top-down ILP algorithm,which enables an efficient discovery of datalogrules from data with both numerical and categorical attributes.Our approach comprises a preprocessing phasefor computing the correlations between numerical andcategorical attributes, as well as an extension to the ILPrefinement step, which enables us to detect interestingcandidate rules and to suggest refinements with relevantattribute combinations. We report on experiments withU.S. Census data, Freebase and DBpedia, and show thatour approach helps to efficiently discover rules with numericalintervals.

Endowing the semantically-oblivious Internet with Intelligence would advance the Internet capability to learn traffic behavior and to predict future events. In this paper, we propose a hybrid intelligence memory system, or NetMem, for network-semantics reasoning and targeting Internet intelligence. NetMem provides a memory structure, mimicking the human memory functionalities, via short-term memory (StM) and long-term memory (LtM). NetMem has the capability to build runtime accessible dynamic network-concept ontology (DNCO) at different levels of granularity. We integrate Latent Dirichlet Allocation (LDA) and Hidden Markov Models (HMM) to extract network-semantics based on learning patterns and recognizing features with syntax and semantic dependencies. Due to the large scale and high-dimensionality of Internet data, we utilize the Locality Sensitive Hashing (LSH) algorithm for data dimensionality reduction. Simulation results using real network traffic show that NetMem with hybrid intelligence learn traffic data semantics effectively and efficiently even with significant reduction in volume and dimensionality of data, thus enhancing Internet intelligence for self-/situation-awareness and event/behavior prediction.

We propose a technique of training models for feature extraction using prior expectation of regions of importance in an instrument's timbre. Over a dataset of training examples, we extract significant spectral peaks, calculate their ratio to fundamental frequency, and use $k$-means clustering to identify a set of windows of spectral prominence for each instrument. These windows are used to extract amplitude values from training data to use as features in classification tasks. We test this approach on two databases of 17 instruments, cross evaluate between datasets, and compare with MFCC features.

Efficient data indexing and exact k-nearest-neighbor (kNN) retrieval are still challenging tasks in high-dimensional spaces. This work highlights the difficulties of indexing in high-dimensional and tightly-clustered dataspaces by exploring several important tunable parameters for optimizing kNN query performance using the iDistance and iDStar algorithms. We experiment on real and synthetic datasets of varying size, cluster density, and dimensionality, and compare performance primarily through filter-and-refine efficiency and execution time. Results show great variability over parameter values and provide new insights and justifications in support of prior best-use practices. Local segmentation with iDStar consistently outperforms iDistance in any clustered space below 256 dimensions, setting a new benchmark for efficient and exact kNN retrieval in high-dimensional spaces. We propose several directions of future work to further increase performance in high-dimensional real-world settings.

Research in the field of Affective Computing on affect recognition through speech has used a “fishing expedition” approach. Although some frameworks could achieve certain success rates, many of these approaches missed the theory behind the underlying voice and speech production mechanism. In this work, we found some correlation among the acoustic parameters (paralinguistic/non-verbal speech content) in the physiological mechanism of voice production. Furthermore, we also found some correlation when analyzing their relationships statistically. Aligned with this finding, we implemented our framework using the K-Nearest Neighbors (KNN) algorithm. Although our work is still in its infancy, we believe this context-free approach will bring us forward toward creating an intelligent agent with affect recognition ability. This paper describes the problem, our approach and our results.

This study explores correlations between human ratings of essay quality and component scores based on similar natural language processing indices and weighted through a principal component analysis. The results demonstrate that such component scores show small to large effects with human ratings and thus may be suitable to providing both summative and formative feedback in an automatic writing evaluation systems such as those found in Writing-Pal.

List of conference committees for the Twenty-Seventh International Florida Artificial Intelligence Research Society Conference (FLAIRS-27).