Gustavo Arroyo

Improving offline narrowing-driven partial evaluation using size-change graphs

An offline approach to narrowing-driven partial evaluation (a partial evaluation scheme for first-order functional and functional logic programs) has recently been introduced. In this approach, program annotations (i.e., the expressions that should be generalised at partial evaluation time to ensure termination) are based on a simple syntactic characterisation of quasi-terminating programs. This work extends the previous offline scheme by introducing a new annotation strategy which is based on a combination of size-change graphs and binding-time analysis. Preliminary experiments point out that the number of program annotations is significantly reduced compared to the previous approach, which means that faster residual programs are often produced.

Two simple and effective feature selection methods for continuous attributes with discrete multi-class

We present two feature selection methods, inspired in the Shannons entropy and the Information Gain measures, that are easy to implement. These methods apply when we have a database with continuous attributes and discrete multi- class. The first method applies when attributes are independent among them given the class. The second method is useful when we suspect that interdependencies among the attributes exist. In the experiments that we realized, with synthetic and real databases, the proposed methods are shown to be fast and to produce near optimum solutions, with a good feature reduction ratio.

Non-Myopic Feature Selection Method for Continuous Attributes and Discrete Class

Currently there exist diverse feature selection ranking methods and metrics for databases with pure discrete data (attributes and class), or pure continuous data. However, little work has been done for the case of continuous attributes with discrete class, and at the same time evaluating attribute subsets in a non-myopic fashion, considering its inter-dependencies or interactions. Normally what we can do is perform discretization, and then apply some traditional feature selection method; nevertheless the results vary depending on the discretization method that we utilized. Additionally, if we only evaluate isolated attributes, we probably obtain poor results, because we are not considering attribute inter-dependencies. We propose a metric and method for feature selection on continuous data with discrete class, inspired in the Shannons entropy and the information gain, which overcomes the above problems. In the experiments that we realized, with synthetic and real databases, the proposed method has shown to be fast and produce near optimum solutions, selecting few attributes.CCP (Center for Counter Plagiarism) is a tool designed to aid in the process of verifying digital documents in order to ensure their originality and appropriate usage of references. In this paper we describe the design of an interface for CCP that aims to reduce the burden of determining whether plagiarism exists by locating identical copies in the web as well as noting text that seems original. The main contribution of CCP is a simple yet meaningful graphical user interface that helps in the decision making process when drawing the line between fair use and plagiarism. We also analyze and discuss the preliminary results of performance and usability tests run on CCP.

A Transformational Approach to Polyvariant BTA of Higher-Order Functional Programs

We introduce a transformational approach to improve the first stage of offline partial evaluation of functional programs, the so called binding-time analysis (BTA). For this purpose, we first introduce an improved defunctionalization algorithm that transforms higher-order functions into first-order ones, so that existing techniques for termination analysis and propagation of binding-times of first-order programs can be applied. Then, we define another transformation (tailored to defunctionalized programs) that allows us to get the accuracy of a polyvariant BTA from a monovariant BTA over the transformed program. Finally, we show a summary of experimental results that demonstrate the usefulness of our approach.

A technique for information retrieval from microformatted websites

In this work, we introduce a new method for information extraction from the semantic web. The fundamental idea is to model the semantic information contained in the microformats of a set of web pages, by using a data structure called semantic network. Then, we introduce a novel technique for information extraction from semantic networks. In particular, the technique allows us to extract a portion—a slice—of the semantic network with respect to some criterion of interest. The slice obtained represents relevant information retrieved from the semantic network and thus from the semantic web. Our approach can be used to design novel tools for information retrieval and presentation, and for information filtering that was distributed along the semantic web.

Obtaining Teachers' Expertise to Refine an Affective Model in an Intelligent Tutor for Learning Robotics

Emotions are a ubiquitous component of motivation and learning; therefore it is desirable for intelligent tutors the incorporation of affective models. We developed an affective behavior model for intelligent tutoring systems with base on intuition, literature and teachers’ expertise. The model selects a tutorial action based on the knowledge and affective state of the student. We conducted a study to ask teachers how they deal with affective aspects when they are teaching; and we used the results to refine our model. In the study the teachers saw a video of students interacting with the learning environment, they rated the affective and knowledge state of the students and selected the tutor actions according with the student state. The tutorial action is given via an animated pedagogical agent. Nine teachers from different scholar levels participated in the study. In this paper, we present the teachers study to refine the affective behavior model.

Intelligent learning system based on SCORM learning objects

This paper shows the creation of the adaptive SCORM sequencing models, taking advantage of the latest developments offered by the artificial intelligence field, to provide the best choice to the student, based in learning objects, using a tutor model in self learning. The Tutor uses decision networks also called influence diagrams, to improve the development of resources and learning materials in a learning content management system, to offer students the best pedagogical decision according to their performance. The intelligent learning system is validated in an online environment. The results of the evaluation process in undergraduate engineering courses are encouraging because they show improvements in students learning who used this approach, compared to those who did not use it. The paper also shows the potential application of this learning approach for power systems operators.

Towards CNC programming using haskell

Recent advances in Computerized Numeric Control (CNC) have allowed the manufacturing of products with high quality standards. Since CNC programs consist of a series of assembler-like instructions, several high-level languages (e.g., AutoLISP, APL, OMAC) have been proposed to raise the programming abstraction level. Unfortunately, the lack of a clean semantics prevents the development of formal tools for the analysis and manipulation of programs. In this work, we propose the use of Haskell for CNC programming. The declarative nature of Haskell provides an excellent basis to develop program analysis and manipulation tools and, most importantly, to formally prove their correctness.