Some future work on decision-making with Rough Sets

-Doctor Nidhika Yadav

#decision-making

Firstly, in my Ph.D. work, I incepted the concept of novel Rough Set based uncertainty measure, ‘span’, and the definition of special Rough subsets of the Universe, ‘spanning sets’.

I elaborated on how ‘span’ measures the salience of any subset of the Universe, with the ‘spanning set’ being the subset that maximizes this measure. This set represents the key elements of a problem.

I note that while there has been substantial work on Rough Sets in different applications, none have considered a stochastic Rough subset (of the Universe U) and examined its suitability as a representative of the entire Universe under consideration.

Therefore, I define these concepts from the perspective of generic decision-making and subsequently apply them to determine extracts of text documents.

This approach views the task of extractive summarization as a problem of generating a subset dynamically from the Universe of sentences that is the best representation of the Universe, with a reduced cardinality and without losing essential information.

The fundamental idea of our work is to determine the most suitable subset(s) of the Universe of sentences in the text document(s) under consideration.

I formulated an optimization problem to generate the extract for the text under consideration using the proposed uncertainty measure of ‘span ‘.

We applied this work and modeled unsupervised keyword extraction using the span concept for Soft Rough Sets.

We also defined the concept of complete span to address the drawback of sparse indiscernibility relation.

The complete span measures the cumulative effect of each attribute of the Information System to produce a spanning set.

For my problem of text summarization, a spanning set of a collection of sentences under certain constraints represents a set of distinct sentences having maximum information.

I also noted that computing a spanning set is a computationally complex problem and that the objective function does not have a well-formed mathematical formulation;

Hence, I solved the modeled optimization problem of finding a spanning set using the evolutionary algorithm of Particle Swarm Optimization.

I evaluated standardized DUC datasets (http://www-nlpir.nist.gov/projects/duc/) and Enron Email datasets.

It follows from extensive experiments that the developed Rough Set based Span for Text Summarization model performed best among several proposed Information Systems and problem-modeling schemes.

We applied it to the domain of extractive text summarization, but the novelty of the proposed uncertainty measure is its versatility in solving various real-life applications.

I pursued many of these applications after completing my Ph.D.

Thanks for scrolling down,

Stay Well,

~~~