Tiziana Ligorio

Postmodernism and fuzzy systems

Affinities and shared ideas may be found at the foundations of both fuzzy set theory and the western philosophical movement commonly referred to as Postmodernism. The interest in highlighting these affinities is driven by the fact that the ideas and beliefs that are questioned by both these movements are exactly those which stand (or used to stand) as foundations of both western philosophy and science, and have their roots in an Aristotelian view of the world. By direct comparison between concepts drawn from passages of the main philosophers of the Postmodern movement and fundamentals of fuzzy set theory, parallelisms and commonalities are brought to the surface.

Wizards' dialogue strategies to handle noisy speech recognition

This paper reports on a novel approach to the design and implementation of a spoken dialogue system. A human subject, or wizard, is presented with input of the sort intended for the dialogue system, and selects from among a set of pre-defined actions. The wizard has access to hypotheses generated by noisy automated speech recognition and queries a database with them using partial matching. During the ambitious study reported here, different wizards exhibited different behaviors, elicited different degrees of caller affinity for the system, and achieved different degrees of accuracy on retrieval of the requested items. Our data illustrates that wizards did not trust automated speech recognition hypotheses when they could not lead to a correct database match, and instead asked informed questions. The wealth of data and the richness of the interactions are a valuable resource with which to model expert wizard behavior.

Data mining to support human-machine dialogue for autonomous agents

Next-generation autonomous agents will be expected to converse with people to achieve their mutual goals. Human-machine dialogue, however, is challenged by noisy acoustic data, and by peoples preference for more natural interaction. This paper describes an ambitious project that embeds human subjects in a spoken dialogue system. It collects a rich and novel data set, including spoken dialogue, human behavior, and system features. During data collection, subjects were restricted to the same databases, action choices, and noisy automated speech recognition output as a spoken dialogue system. This paper mines that data to learn how people manage the problems that arise during dialogue under such restrictions. Two different approaches to successful, goal-directed dialogue are identified this way, from which supervised learning can predict appropriate dialogue choices. The resultant models can then be incorporated into an autonomous agent that seeks to assist its user.

Partial redundant modeling

Redundant modeling combines different models of the same problem using channeling constraints [1]. Channeling constraints allow different formulations of a problem to interact, propagating the constraints between different formulations. This can result in a significant improvement in performance. Originally, work on redundant modeling assumed that redundant models must fully characterize the problem [1]. Later, Smith argued that only the primal model need fully characterize the problem, while the dual model need only have all the dual variables and channeling constraints between the two models (a minimal combined model)[2]. This paper proposes partial redundant modeling, an extension of the minimal combined model that encourages more than two models, omits some dual variables and omits all the dual constraints. Partial redundant models originate in problems with a categorical structure, where the variables may be subdivided into categories. Often these categories can be identified as groups of variables that fall under n-ary constraints that partition the variables into disjoint sets. Real world problems, such as scheduling and rostering, may also have categorical structure. Logic puzzles are a class of problems with a simplified version of categorical structure. A logic puzzle consists of a set of objects, a set of categories (same-size disjoint subsets of those objects) that must take on different values, and a set of semantic relations, which specify the relationships that hold between the categories. In a logic puzzle, each category can be viewed as a subset of CSP variables under an all-diff constraint. Different CSP models can be obtained by selecting the objects in a different category as the domain values, taking all other objects in the other categories to be the variables. We propose to maintain multiple partial redundant models of problems with categorical structure, adding channeling constraints between the n-ary constraints in the redundant partial models. These channeling constraints make certain that value assignments under an n-ary constraint in one partial model are reflected under that n-ary constraint in some other partial model. We call these categorical channeling constraints.