Multi-Dimensional Summarization in Cyber-Physical Society

Hai Zhuge

• Publisher: Elsevier
• Year: 2016
• Print ISBN: 9780128034552
• eBook ISBN: 9780128034644
• Topic: Computer Science
• Subtopic: Artificial Intelligence (AI) & Semantics
• Index: Computer Science

Text summarization has been studied for over a half century, but traditional methods process texts empirically and neglect the fundamental characteristics and principles of language use and understanding. Automatic summarization is a desirable technique for processing big data. This reference summarizes previous text summarization approaches in a multi-dimensional category space, introduces a multi-dimensional methodology for research and development, unveils the basic characteristics and principles of language use and understanding, investigates some fundamental mechanisms of summarization, studies dimensions on representations, and proposes a multi-dimensional evaluation mechanism. Investigation extends to incorporating pictures into summary and to the summarization of videos, graphs and pictures, and converges to a general summarization method. Further, some basic behaviors of summarization are studied in the complex cyber-physical-social space. Finally, a creative summarization mechanism is proposed as an effort toward the creative summarization of things, which is an open process of interactions among physical objects, data, people, and systems in cyber-physical-social space through a multi-dimensional lens of semantic computing. The author's insights can inspire research and development of many computing areas. - The first book that proposes the method for the summarization of things in cyber-physical society through a multi-dimensional lens of semantic computing. - A transformation from the traditional application-driven research paradigm into a data-driven research paradigm for creative summarization through information modeling, cognitive modeling and knowledge modeling. - A multi-dimensional methodology for studying, managing, creating and applying methods.

1

Introduction

Abstract

Summarization plays an important role in understanding and representing natural languages. With the rapid and continual expansion of texts, pictures and videos in cyberspace, automatic summarization becomes more and more desirable. Text summarization has been studied for more than half century, but it is still hard to automatically generate fluent texts that satisfy readers. Existing approaches follow the traditional computing paradigm to process texts empirically. The fundamental characteristics and principles of understanding and using languages are neglected. It is time for a shifting research paradigm to make a breakthrough. This book summarizes previous text summarization approaches with a multi-dimensional category space, explores the principles of emerging structure in texts, introduces a multi-dimensional methodology for research, and development, unveils the basic characteristics and principles of using and understanding languages, investigates some fundamental mechanisms of summarization, studies the dimensions and the forms of representations, and suggests a multi-dimensional evaluation mechanisms. Research extends to the incorporation of pictures into summary and to the summarization of videos, graphs and pictures, and then reaches a general framework of summarization. Further, some basic structure and behaviors of summarization are studied in the complex space consisting of cyberspace, physical space and social space. The limitation of summarization is unveiled and the notion of innovative summarization is proposed. The basic viewpoints include the following aspects: (1) The structure of a representation emerges with operations on it, and a complex structure of understandable representation is near decomposable. The structure can improve summarization but not great. Make a better summary needs practical semantic representation. (2) A representation suitable for summarization should render a core that reflects motivation. (3) Intelligent summarization is an open process of various interactions, involved in various explicit and implicit semantic links. (4) The form of summary is diverse and summarization can carry out from multiple dimensions. (5) Automatic summarization has a limitation, and linking summarization to cyberspace, physical space, and social space to establish a human-machine-nature symbiotic environment is a way to approach and even break the limitation. This work serves to contribute to the formation of the new paradigm of summarization research.

Keywords

Artificial intelligence; citation; Cyber-Physical Society; dimension; graph summarization; natural language processing; picture summarization; semantic link; text summarization; video summarization; methodology

To realize an expert-level automatic summarization is a challenge to computing research [49]. Automatic summarization has been studied for over a half century, following the traditional paradigm of computing. It is now critical to shift research paradigm with the emerging Cyber-Physical Society and the revolution of sciences, technologies, and industries. Observing various human summaries and rethinking how human make summaries is a way to inspire fundamental research.

Versatile summaries accompany our daily life. People have created many forms of summary such as the abstracts of scientific papers, the prefaces of books, the tables of contents, personal curriculum vitae, the headlines of news, webpages with hyperlinks, book reviews, Wikipedia, and the results of Web search. Some summaries incorporate pictures, videos, graphs, or tables into texts. Applications include Web portals such as Yahoo, YouTube, posters, slides, medical certificates, TV guides, advertisements, and conference programs. A good summary should represent the core meaning of the original text, quickly attract attention, and effectively convey meaning with regard to interests. The current summaries that people often read are designed by professional people.

1.1 Open collaborative human summarization

Wikipedia can be regarded as an open collaborative human summarization environment. One Wikipedia page is a summary of one thing. Figure 1.1 shows a Wikipedia page that summarizes John McCarthy, the pioneer of artificial intelligence. It looks like a Web-style curriculum vitae. From the other pages on events and concepts, we can see that Wikipedia uses a set of structures similar to book (e.g., table of contents) and paper (e.g., begin with a definition, followed by historical review, and ended with references, further readings and external links) to unify various content contributions and help readers to understand the entries with reading experience.

Figure 1.1 Wikipedia: a Web-based platform for collaborative human summarization.

The advantages of Wikipedia can be summarized as follows:

1. The representation adopts the language structure that appears in books, papers, and webpages. It is in line with human reading experience. The table of contents with hyperlinks guides readers to read the interested parts quickly. The reference links provide supportive or extended reading materials for readers. Different from the structures of the hardcopy newspapers, books and papers, its structure is suitable for Web browsing.

2. The content of its page is divided by sections (with title), which helps readers’ understanding through the structure and divides the interests of contributors. Sections are relatively independent from each other, which enables different contributors to focus on different sections at one time of contribution and to represent various viewpoints from different aspects of understandings.

3. It opens to all users to read and edit. This openness enables the contents to evolve to reflect various viewpoints and understandings from different contributors at different times.

4. It has a category network on pages. A category with attributes id, page-id, name, in-links, out-links, and pages provides the links for browsing the pages within a category or through categories. Figure 1.2 is an example of the category network. Different from the traditional category hierarchy that makes accurate classification and represents a single abstraction, the category network has loops so that the links between categories do not just represent single abstraction. This reflects diverse understandings of contributors and can help establish more possible links between terms in other texts if terms can be mapped into categories.

5. It provides an evolving common content base and indicates the ways to make abstraction for interpreting terms. The contents are for human to read while the category network guides people to browse relevant contents. The category network provides indicators for various application systems to predict and suggest abstraction approach.

Figure 1.2 A category network in Wikipedia, where the arrows point to super-categories.

Wikipedia has the following shortcomings, especially from the computing point of view:

1. The size and display of the content cannot adapt to the interests of different readers.

2. It usually takes a long time to evolve into a complete content of a page. This leads to incomplete content and redundant links and categories.

3. It relies on human contribution and labor-intensive editing work.

4. The category structure is not strong enough in structure and semantics (e.g., the current attributes of category do not reflect the nature of the category in question) to explain representation and reflect abstraction. A well-structured and semantics-rich category hierarchy can better explain representations from different abstraction levels, e.g., explaining two representations with the following keyword sets “I, like, apple” and “He, like, orange” by a representation at the higher abstraction level “People, like, fruit.”

5. It can help readers to understand some terms explained in natural language, but it is limited in ability to interpret diverse representations, and it is also hard to interpret a summarization result and the process of summarization since the purpose of designing the category structure is to provide navigational links to all of its pages in a hierarchy of categories which enables readers who know “essential—defining—characteristics” of a topic to browse and quickly find the sets of pages on topics with the characteristics.

6. It lacks objective evaluation on the contents and categories, which can provide significant guidance for contributors and readers.

The Wikipedia is a platform that enables people to make and share a collaborative summarization on the World Wide Web. It records users’ understandings and opinions and evolves with operations. The following chapters will discuss the applications of Wikipedia.

A way to obtain the interpretation of representation (e.g., a term in text) is to map the representation into the categories in Wikipedia by using one language representation to interpret another language representation. One advantage of the mapping is that it can save people’s time for searching the interpretations of a representation. The limitation is that it is not formalized for machines to carry out reasoning to prove a solution or verify an assumption strictly. Therefore, the interpretation is empirical and incomplete. On the other hand, Wikipedia mainly contains general and mature categories, so it is not able to interpret specific categories, especially the new concepts appeared in research like the concept of Cyber-Physical Society. From the evolution point of view, Wikipedia will include more and more categories. New concepts will be included in Wikipedia later when they become common-interest concepts. However the expres...

Table of contents

• Cover image
• Title page
• Table of Contents
• Copyright
• Dedication
• About the Author
• Foreword
• Preface
• Acknowledgment
• 1. Introduction
• 2. The emerging structures
• 3. Patterns in representation and understanding
• 4. The think lens
• 5. Multi-dimensional methodology
• 6. Characteristics and principles of understanding and representation
• 7. Implicit links in multi-dimensional space
• 8. General citation
• 9. Dimensions of summary
• 10. Multi-dimensional evaluation
• 11. Incorporating pictures into a summary
• 12. Summarizing videos, graphs and pictures
• 13. General framework of summarization
• 14. Summarization of things in Cyber-Physical Society
• 15. Limitations and challenges
• 16. Creative summarization
• 17. Conclusion
• Appendix A: Human–machine–nature symbiosis
• References