Weight learning is a challenging problem in Markov Logic Networks (MLNs) due to the large size of the ground propositional probabilistic graphical model that underlies the first-order representation of MLNs. Though more sophisticated weight learning methods that use lifted inference have been proposed, such methods can typically scale up only in the absence of evidence, namely in generative weight learning. In discriminative learning, where the evidence typically destroys symmetries, existing approaches are lacking in scalability. In this paper, we propose a novel, intuitive approach for learning MLNs discriminatively by utilizing approximate symmetries. Specifically, we reduce the size of the training database by clustering approximately symmetric atoms together and selecting a representative atom from each cluster. However, each choice made from the clusters induces a different distribution, increasing the uncertainty in our learned model. To reduce this uncertainty, we learn a finite mixture model by stacking the different distributions, where the parameters of the model are learned using an EM approach. Our results on several benchmarks show that our approach is much more scalable and accurate as compared to existing state-of-the-art MLN learning methods.

Report on the Thirtieth International Florida Artificial Intelligence Research Society Conference (FLAIRS-30)

The 30th International Florida Artificial Intelligence Research Society Conference (FLAIRS-30) was held May 22–24, 2017, at the Hilton Marco Island Beach Resort and Spa in Marco Island, Florida, USA. The conference events included invited speakers, special tracks, and presentations of papers, posters, and awards. The conference chair was Ingrid Russell from the University of Hartford. The program cochairs were Vasile Rus from The University of Memphis and Zdravko Markov from Central Connecticut State University. The special tracks were coordinated by Keith Brawner from the Army Research Laboratory.

We present in this paper a study on negation in dialogues. In particular, we analyze the peculiarities of negation in dialogues and propose a new method to detect intra-sentential and inter-sentential negation scope and focus in dialogue context. A key element of the solution is to use dialogue context in the form of previous utterances, which is often needed for proper interpretation of negation in dialogue compared to literary, non-dialogue texts. We have modeled the negation scope and focus detection tasks as a sequence labeling tasks and used Conditional Random Field models to label each token in an utterance as being within the scope/focus of negation or not. The proposed negation scope and focus detection method is evaluated on a newly created corpus (called the DeepTutor Negation corpus; DT-Neg). This dataset was created from actual tutorial dialogue interactions between high school students and a state-of-the-art intelligent tutoring system.

Vector based word representation models are often developed from very large corpora. However, we often encounter words in real world applications that are not available in a single vector model. In this paper, we present a novel Neural Network (NN) based approach for obtaining representations for words in a target model from another model, called the source model, where representations for the words are available, effectively pooling together their vocabularies. Our experiments show that the transformed vectors are well correlated with the native target model representations and that an extrinsic evaluation based on a word-to-word similarity task using the Simlex-999 dataset leads to results close to those obtained using native model representations.

We present in this paper an innovative solution to the challenge of building effective educational technologies that offer tailored instruction to each individual learner. The proposed solution in the form of a conversational intelligent tutoring system, called DeepTutor, has been developed as a web application that is accessible 24/7 through a browser from any device connected to the Internet. The success of several large scale experiments with high-school students using DeepTutor is a solid proof that conversational intelligent tutoring at scale over the web is possible.

Cochlear implants (CIs) are an effective intervention for individuals with severe-to-profound sensorineural hearing loss. Currently, no tuning procedure exists that can fully exploit the technology. We propose online unsupervised algorithms to learn features from the speech of a severely-to-profoundly hearing-impaired patient round-the-clock and compare the features to those learned from the normal hearing population using a set of neurophysiological metrics. Experimental results are presented. The information from comparison can be exploited to modify the signal processing in a patient’s CI to enhance his audibility of speech.

There is no denying the fact that AI's original aim of reproducing human-level intelligence has taken a back seat in favor of the development of practical and efficient systems for important but narrow domains under numerous umbrellas. While the importance of perception for intelligence is now well-understood, a major hurdle is to discover the appropriate representation and processes that can seamlessly support low-level perception and high-level cognition in a computational architecture. There is no shortage of cognitive architectures (see Samsonovich 2010 for a catalog), however, principled design is scarce. In this paper, we explicate our position and report on our ongoing investigations on representation and processes for developing an intelligent agent from first principles.

We report recent advances in intelligent tutoring systems with conversational dialogue. Macroadaptivity refers to a system's capability to select appropriate instructional tasks for the learner to work on. Microadaptivity refers to a system's capability to adapt its scaffolding while the learner is working on a particular task. Learning progressions and deeper dialogue and natural language processing techniques are key features of DeepTutor, the first intelligent tutoring system based on learning progressions.

We report recent advances in intelligent tutoring systems with conversational dialogue. We highlight progress in terms of macro and microadaptivity. Macroadaptivity refers to a system’s capability to select appropriate instructional tasks for the learner to work on. Microadaptivity refers to a system’s capability to adapt its scaffolding while the learner is working on a particular task. The advances in macro and microadaptivity that are presented here were made possible by the use of learning progressions, deeper dialogue and natural language processing techniques, and by the use of affect-enabled components. Learning progressions and deeper dialogue and natural language processing techniques are key features of DeepTutor, the first intelligent tutoring system based on learning progressions. These improvements extend the bandwidth of possibilities for tailoring instruction to each individual student which is needed for maximizing engagement and ultimately learning.