Hemant Joshi

Local and Global Recency Weighting Approach to Bug Prediction

Finding and fixing software bugs is a challenging maintenance task, and a significant amount of effort is invested by software development companies on this issue. In this paper, we use the Eclipse projects recorded software bug history to predict occurrence of future bugs. The history contains information on when bugs have been reported and subsequently fixed.

A Dynamic Approach to Software Bug Estimation

There is an increase in the number of software projects developed in a globally distributed environment. As a consequence of such dispersion, it becomes much harder for project managers to make resource estimations with limited information. In order to assist in the process of project resource estimations as well as support management planning for such distributed software projects, we present a methodology using software bug history data to model and predict future bug occurrences. The algorithm works in a two-step analysis mode: Local and Global Analysis. Local analysis leverages the past history of bug counts for a specific month. On the other hand, global analysis considers the bug counts over time for each individual component. The bug prediction achieved by the algorithm is close to actual bug counts for individual components of Eclipse software.

Automatic Dissemination of Text Information using the EBOTS system

World Wide Web contains 170 Terabytes of information [1] and storage estimates show that the new information is growing at a rate of over 30% a year. With the quanta of information growing exponentially, it is important to understand the information semantically to know what concepts are relevant and what are irrelevant. The Evolutionary Behavior Of Textual Semantics (EBOTS) system being developed at University of Arkansas at Little Rock [2] aims at the quantitative reasoning aspect of textual information. In the automatic decision-making mode, the EBOTS system can distinguish between relevant and irrelevant information, discarding irrelevant documents and accepting only relevant information to develop expertise in a particular field. This paper discusses the usefulness of Information Theory in the development of relevance criteria and the results obtained in the context of textual information.

Who Stole the Bat? Deception Detection on the Basis of Actions

In work very relevant to national defense and homeland security, this paper describes software that performs symbolic reasoning (in particular., logical inference) to detect deception on the basis of actions. This is in sharp contrast with present approaches that detect deception based on physiological factors, as well as on verbal and non-verbal cues. Our approach attempts to model agents and their actions. This is achieved in a logic programming framework using a theory of agents, a theory of actions, and a theory of reasoning with respect to time. As a test case, a childrens mystery is analyzed and implemented. The software correctly reasons about who the potential suspects are, and ultimately, correctly identifies the chief culprit. Further, it can correctly introspect with regard to previously held beliefs. Our approach is novel in that it attempts to identify deception on the basis of actions, and it does so in a high level, symbolic reasoning framework

A child’s story to illustrate automated reasoning systems using opportunity and history

The primary author has performed previous work to create generalized, high-level reasoning software to identify deception on the basis of actions. The work here is to apply that software architecture to an entirely different domain. The original domain was an espionage scenario, and detected deception when communications were in conflict. This present domain is reasoning about a child’s story to determine who is lying about the theft of some objects. Applying the previous work to a different domain is an attempt to demonstrate the generality of the architecture. It is also an attempt to further generalize the software, and to formalize additional “common sense” strategies in the detection of deception. This software detects deception on the basis of actions. This is in sharp contrast with present approaches that detect deception based on physiological factors, as well as on verbal and non-verbal cues. Our approach models agents and their actions in a logic programming framework using a theory of agents, a theory of actions, and a theory of reasoning with respect to time. As a test case, a children’s mystery is analyzed and implemented. The goal of the story is to identify who stole some items. The software correctly reasons about who the potential suspects are, and ultimately, correctly identifies the chief culprit. Further, it can correctly introspect with regard to previously held beliefs. The program we have developed is able to mimic the thought processes and conclusions of a police investigation.