Concept evolution analysis based on the Dissipative Structure of Concept Semantic Space
Introduction
In the domain of text semantic analysis and application, compared with a keyword, a concept has bigger semantic granularity and holds more semantic information [[1], [2]], which is used in ontology construction [3], text semantic representation [4], text semantic annotation [5], semantic search [6], event detection and tracking [7], etc., in order to improve the efficiency of text semantic processing. In the above applications, the accuracy of concept analysis plays a crucial role in improving the precision of applications.
The accuracy of concept analysis is mainly reflected in the following aspects. (1) When a new concept appears, the system can detect it timely. (2) The system can express the current semantic of a concept accurately. (3) The system can perceive the change of a concept semantic timely. To achieve these goals, we have to analyze the phenomenon of concept evolution and its process from dynamic perspective.
Concept semantic evolution is a common phenomenon. Concept semantic, in the physical world or in the Web virtual space, is constantly evolving. Concept semantic evolution is a continuous process accompanying a concept’s formation and development at different stages. Concept semantic evolution usually includes concept generation, concept disappearance, concept semantic transference, concept semantic diversity.
At present, concept learning methods can come down to three types: concept extraction in linguistics [[8], [9]], concept extraction in statistics [10] and the combination of the two [11]. But these methods have their obvious deficiencies in analyzing the above phenomenon of concept evolution. It is difficult to determine whether concept semantic changes just by means of the changes of the attributive words or the relations between attributive words. It is also difficult to determine when concept semantic is comparatively stable and when it evolves.
Therefore, it is likely to analyze the process of concept evolution accurately from the global and dynamic perspective. The general process of a text semantic analysis system is constantly acquiring texts (i.e. web pages), extracting keywords of texts, mining semantic relations between the keywords and extracting concepts based on semantic relations between the keywords or syntactic analysis. In the above process, the continuous addition of texts leads to the constant changes of keywords and their semantic relations, which causes concept changes, namely, semantic evolution phenomena of concept semantic. Concept Semantic Space (CSS) can be regarded as surroundings where the above concepts evolve [12]. In CSS, concept evolutions are seen as the results of the interaction of the keywords by semantic relations.
Many similarities between CSS and thermodynamic system can found when we make an analogy. The keywords environment in CSS is analogous to the molecules in the thermodynamic system, the interaction between the keywords to the thermal operation of the molecules, the stable relation between the keywords to the ordered inter-molecules structure, concepts in CSS to subsystem of thermodynamic system. The theory of Dissipative Structure reveals the rules of molecule movement that the system mutates from the disordered state to the ordered state under the external influence. Concept Semantic Space (CSS) has a similar rule to molecule thermal operation, especially the process of keywords conversion from the disordered to the ordered, which is analogous to the Dissipative Structure of thermodynamics. So the theory of Dissipative Structure can be employed to study the Dissipative Structure of CSS. The ordered state of CSS is a macro-reflection of an ordered organization established by interactions of keywords and concepts in the CSS. The inter-conversion between the ordered state and the disordered state presents the concept semantic changes, namely, concept semantic evolution. Therefore, the paper proposes a method of studying concept semantic evolution by means of CSS evolution based on the theory of Dissipative Structure, which is expected to tackle the deficiencies in the above methods of concept evolution analysis.
The paper is organized as follows: the first section is introduction. Section 2 presents basic theories and related work. In Section 3, we construct a dissipative structure model of CSS and discuss discrimination of the Dissipative Structure of CSS. Section 4 analyzes concept evolution based on the evolution of CSS. In Section 5, the early detection method of emergency is introduced based on concept evolution analysis. Section 6 is mainly about experiments and conclusion is drawn in the last section.
Section snippets
Concept
The definitions of concept in philosophy, linguistics, logic, psychology, cognitive informatics, software engineering and knowledge engineering are not all the same [13]. Philosophically, concept is the basic unit of thinking. In artificial intelligence, concept is used to model the knowledge of human. In linguistics, concept is a noun or noun phrase as the subject of to-be structure [14]. In cognitive informatics, concept is an abstract structure with exact semantic of cognitive process, such
Dissipative Structure of CSS
Making an analogy with Dissipative Structure of thermodynamics system, the paper proposes the Dissipative Structure of Concept Semantic Space (CSS) shown in Fig. 1. From the perspective of Web-oriented text semantic analysis, the external surroundings in the figure is the whole Web, and the center is CSS of a certain field. Whether the space satisfies the requirements of Dissipative Structure is determined by the following conditions:
(1) Open system
In the Web, CSS of a certain domain is open
Relation between concept evolution and CSS evolution
In CSS, concept evolution and CSS evolution are not independent but correlative. Concept evolution refers to the generation of a new concept, the disappearance of an old concept, changes of concept semantics or others. CSS evolution means that CSS converts from an ordered state to the next ordered state. In the process, CSS may experience one or more middle states (partly disordered state). Compared with CSS in the old ordered state, some concepts in the new one evolved. The above evolutions
Application
We apply CSS Dissipative Structure to the detection of web emergency and achieve a good result at the early detection of the event.
Nowadays, the Internet is highly developed, so emergency in the real world will be reported on the Web and keep fermenting with the help of Web, such as news, forum, blog, microblog. In the whole process of the event, the Web spreads information and also retains the transitional information simultaneously, so it becomes an effective source of information to detect
Experiment on discriminating the Dissipative Structure of CSS
(1) Experimental purpose. The purpose is to verify the relationship between changes of space information entropy and network features when CSS forms a Dissipative Structure.
(2) Data set. The experiment considers news webpages of the news portal in Chinese mainland as basic data, including about 6,500,000 news pages from July 1, 2009 to June 30, 2013. In the experiment, the pretreatments include text extraction, word segment, keyword extraction, concept extraction etc. in the data set. To repeat
Conclusions
To solve the current problem that the deficiency of concept semantic analysis in analyzing the phenomenon of concept evolution, including the difficulty in identifying the concept semantic changes, the paper proposes the method of concept evolution analysis based on the Dissipative Structure of Concept Semantic Space.
The main work is presented as follows.
(1) Propose the Dissipative Structure of CSS based on the theory of Dissipative Structure used in the thermodynamic system, depicting the
Acknowledgments
Research work reported in this paper was partly supported by the Science Foundation of Shanghai under grant no. 16ZR1435500, and by the Foundation for Innovative Research Groups of the National Natural Science Foundation of China under grant no. 71621002.
Xiao Wei received the B.S. degree from the Shandong University, China, and the Ph.D. degree from Shanghai University, China, all in computer science. He is currently an Associate Professor with the Shanghai University, Shanghai, China, and the Postdoctoral Researcher with the Institute of Automation Chinese Academy of Sciences, Beijing, China. His research interests include web content analysis, semantic search, and e-learning.
References (35)
- et al.
Generic ontology of datatypes
Inform. Sci.
(2016) - et al.
Knowle: A semantic link network based system for organizing large scale online news events
Future Gener. Comput. Syst.
(2015) - et al.
Learning ontologies from natural language texts
Int. J. Human-Comput. Stud.
(2004) Concept space connected to knowledge processing for supporting creative design
Knowl.-Based Syst.
(1997)- et al.
Dissipative structure: An explanation and an ecological example
J. Theoret. Biol.
(1972) Learning concept hierarchies from text corpora using formal concept analysis
J. Artificial Intelligence Res.
(2005)- et al.
Measuring algebraic complexity of text understanding Based on Human Concept Learning
IEEE Trans. Human-Mach. Syst.
(2014) - et al.
Learning semantic representation with neural networks for community question answering retrieval
Knowl.-Based Syst.
(2015) - et al.
Recent automatic text summarization techniques: A survey
Artif. Intell. Rev.
(2017) - et al.
On-line new event detection and tracking
Learning domain ontologies for web service descriptions: an experiment in bioinformatics
Ontology learning and its application to automated terminology translation
IEEE Intell. Syst.
Sherlock: A semi-automatic framework for quiz generation using a hybrid semantic similarity measure
Cogn. Comput.
Building the concept semantic space for large text database
Comput. Syst. Sci. Eng.
On concept algebra: A denotational mathematical structure for knowledge and software modeling
Int. J. Cogn. Inform. Natural. Intell.
A Concise Introduction to Logic
The theoretical framework of cognitive informatics
Int. J. Cogn. Inform. Natural Intell.
Cited by (4)
Special section on intelligent sensing and applications for cyber-physical systems
2018, Future Generation Computer SystemsSocial Mechanisms of the Subject Area Formation. The Case of “Digital Economy”
2020, Springer Proceedings in Mathematics and StatisticsA cognitive-based AES model towards learning written English
2019, Journal of Ambient Intelligence and Humanized ComputingSentiment-enhanced learning model for online language learning system
2018, Electronic Commerce Research
Xiao Wei received the B.S. degree from the Shandong University, China, and the Ph.D. degree from Shanghai University, China, all in computer science. He is currently an Associate Professor with the Shanghai University, Shanghai, China, and the Postdoctoral Researcher with the Institute of Automation Chinese Academy of Sciences, Beijing, China. His research interests include web content analysis, semantic search, and e-learning.
Daniel Dajun Zeng received his Ph.D. degree from Carnegie Mellon University. He is a Researcher in the Institution of Automation, Chinese Academy of Sciences. His main research interests are Web computing, agent modeling, and security informatics.
Xiangfeng Luo is a professor in the School of Computer Engineering and Science, Shanghai University, China. He received the master’s and Ph.D. degrees from the Hefei University of Technology in 2000 and 2003, respectively. His main research interests include Web Wisdom, Cognitive Informatics, and Text Understanding.