Accepted Manuscript Properties of language networks and language systems Shuiyuan Yu, Chunshan Xu
PII: DOI: Reference:
S1571-0645(14)00094-3 10.1016/j.plrev.2014.06.003 PLREV 501
To appear in:
Physics of Life Reviews
Received date: 14 May 2014 Accepted date: 21 May 2014
Please cite this article in press as: Yu S, Xu C. Properties of language networks and language systems. Phys Life Rev (2014), http://dx.doi.org/10.1016/j.plrev.2014.06.003
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Comment
Properties of language networks and language systems Comment on ‘‘Approaching human language with complex networks’’ by Cong and Liu Shuiyuan Yu 5EJQQNQH%QORWVGT5EKGPEG%QOOWPKECVKQP7PKXGTUKV[QH%JKPC$GKLKPI%JKPC 'OCKN UJWK[WCP["IOCKNEQO
ChunshanXu 5EJQQNQH(QTGKIPUVWFKGU#PJWK,KCP\JW7PKXGTUKV[*GHGK%JKPC 'OCKN CFKPZW"EQO Language is generally considered a defining feature of human beings, a key medium for interpersonal communication, a fundamental tool for human thinking and an important vehicle for culture transmission. For the anthropoids to evolve into human being, the emergence of linguistic system is a vital step. Then, how can language serve functions so complicated and so important? To answer this question, it is necessary to probe into a central topic in linguistics: the structure of language, which has been inevitably involved in various fields of linguistic research-----the functions of languages, the evolution of languages, the typology of languages, etc. The study of a structure means an investigation from a systematic perspective, an enquiry into its components and the interrelations among them. Network, also mathematically termed as graph, consists of a set of nodes (object) and edges (relations) among them, and thus makes a quite suitable tool for modeling systematic structures. Now that languages are held by many linguists as systems [1], it is not surprising to find network approaches widely applied to modeling language structures. In fact, network approaches have been adopted in many scientific disciplines, social or natural, such as sociology, communication, economy, finance, internet, transportation, molecular biology, etc [2]. The findings in these fields may well contribute to the study of language networks, and provide a basis for the comparison between language networks and other social or natural networks. From a network, certain parameters can be calculated, reflecting properties of the network concerned. The interpretation of these parameters requires the nodes in a network be homogeneous. When it comes to human language, node of network can be linguistic units at various levels: phoneme, syllable, word, sentence, etc., and thus language networks at different linguistic levels can be constructed.
In normal cases, with one network we can only ‘‘observe’’ a language in one
single‘‘ section’’, or rather, model it at one single linguistic level, unable to grasp the holistic picture of that language. As a result, the normal network approaches to language modeling, which focus on one single linguistic level at one time, are somewhat inadequate for the investigation into the overall structure of languages: the holistic properties of a language as a complicated system cannot be directly predicted from its composing components, but consist in, to a considerable degree, the interrelations
among its linguistic subsystems. To solve this problem, Liu and Cong suggested a revised approach which put emphasis on multi-level networks [3], which is held as the first major line of network researches devoted to the understanding of human language and the development of linguistics [4].Based on one text collection, networks at different linguistic levels may be constructed so that network parameters at different linguistic levels can be compared so as to discern the relations among these levels. This can be explained by an analogy with tomography scanning in medical field, which enables doctors to compare scanning images at different levels, or sections, in order to render a diagnosis grounded in a full and integrated knowledge of the situation. As can be seen, such an network approach, instead of committing itself to any isolated level of language, is mainly concerned with the interrelations among different levels, and may thus contribute to revealing the holistic properties of language, serving, at least potentially, as one partial solution to the tough problem that ‘‘a language as a whole is much greater than the simple sum of its parts’’. Apparently, the strength of language networks lies in capturing holistic properties of a language. Therefore, a comprehensive knowledge of a language perhaps demands not only network methods but other approaches, such as those based on the theory of information or synergetics. From the perspective of information theory, what is most interesting in a language is its capacity in communication and its coding(decoding) features, such as the information content and distribution of entropy[5], while in terms of synergetics, what is fascinating is how a language could be systematically modeled, how the dynamic relations could be established among different components of a language when it is in real use, and how subsystems in a language are coordinated into a larger system with holistic structure and holistic functions[6]. From a structural perspective, it seems that the approaches based on synergetics may render better complement to complex networks methods in the investigation into properties of languages. References [1] Saussure FD. 1959. Course in General Linguistics (translated by Baskin W) New York: Philosophical Library. [2] L. da F. Costa, O. N. Oliveira Jr, G. Travieso, F. A. Rodrigues, P. R. V. Boas, L. Antiqueira, M. P. Viana, and L. E. C. da Rocha, 2007.Analyzing and modeling real-world phenomena with complex networks: A survey of applications, ArXiv e-prints, vol. 0711, no. 3199. [3] Liu H, Cong J. 2014. Empirical characterization of modern Chinese as a multi-level system from the complex network approach. Journal of Chinese Linguistics, 42: 1-38 [4] Cong J, Liu H. Approaching human language with complex networks. Physics of Life Reviews 2014; [in this issue]. [5] Ferrer-i-Cancho, R., DĊbowski, Ł. & Moscosodel Prado Martín, F. 2013. Constant conditional entropy and related hypotheses.Journal of Statistical Mechanics, L07001. [6] Köhler, Reinhard. 2005. Synergetic Linguistics. In: Köhler, Gabriel Altmann & Rajmund G. Piotrowski (Eds). Quantitative Linguistics. An International Handbook (pp.760---775). Berlin, New York: Walter de Gruyter.
PII: DOI: Reference:
S1571-0645(14)00094-3 10.1016/j.plrev.2014.06.003 PLREV 501
To appear in:
Physics of Life Reviews
Received date: 14 May 2014 Accepted date: 21 May 2014
Please cite this article in press as: Yu S, Xu C. Properties of language networks and language systems. Phys Life Rev (2014), http://dx.doi.org/10.1016/j.plrev.2014.06.003
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Comment
Properties of language networks and language systems Comment on ‘‘Approaching human language with complex networks’’ by Cong and Liu Shuiyuan Yu 5EJQQNQH%QORWVGT5EKGPEG%QOOWPKECVKQP7PKXGTUKV[QH%JKPC$GKLKPI%JKPC 'OCKN UJWK[WCP["IOCKNEQO
ChunshanXu 5EJQQNQH(QTGKIPUVWFKGU#PJWK,KCP\JW7PKXGTUKV[*GHGK%JKPC 'OCKN CFKPZW"EQO Language is generally considered a defining feature of human beings, a key medium for interpersonal communication, a fundamental tool for human thinking and an important vehicle for culture transmission. For the anthropoids to evolve into human being, the emergence of linguistic system is a vital step. Then, how can language serve functions so complicated and so important? To answer this question, it is necessary to probe into a central topic in linguistics: the structure of language, which has been inevitably involved in various fields of linguistic research-----the functions of languages, the evolution of languages, the typology of languages, etc. The study of a structure means an investigation from a systematic perspective, an enquiry into its components and the interrelations among them. Network, also mathematically termed as graph, consists of a set of nodes (object) and edges (relations) among them, and thus makes a quite suitable tool for modeling systematic structures. Now that languages are held by many linguists as systems [1], it is not surprising to find network approaches widely applied to modeling language structures. In fact, network approaches have been adopted in many scientific disciplines, social or natural, such as sociology, communication, economy, finance, internet, transportation, molecular biology, etc [2]. The findings in these fields may well contribute to the study of language networks, and provide a basis for the comparison between language networks and other social or natural networks. From a network, certain parameters can be calculated, reflecting properties of the network concerned. The interpretation of these parameters requires the nodes in a network be homogeneous. When it comes to human language, node of network can be linguistic units at various levels: phoneme, syllable, word, sentence, etc., and thus language networks at different linguistic levels can be constructed.
In normal cases, with one network we can only ‘‘observe’’ a language in one
single‘‘ section’’, or rather, model it at one single linguistic level, unable to grasp the holistic picture of that language. As a result, the normal network approaches to language modeling, which focus on one single linguistic level at one time, are somewhat inadequate for the investigation into the overall structure of languages: the holistic properties of a language as a complicated system cannot be directly predicted from its composing components, but consist in, to a considerable degree, the interrelations
among its linguistic subsystems. To solve this problem, Liu and Cong suggested a revised approach which put emphasis on multi-level networks [3], which is held as the first major line of network researches devoted to the understanding of human language and the development of linguistics [4].Based on one text collection, networks at different linguistic levels may be constructed so that network parameters at different linguistic levels can be compared so as to discern the relations among these levels. This can be explained by an analogy with tomography scanning in medical field, which enables doctors to compare scanning images at different levels, or sections, in order to render a diagnosis grounded in a full and integrated knowledge of the situation. As can be seen, such an network approach, instead of committing itself to any isolated level of language, is mainly concerned with the interrelations among different levels, and may thus contribute to revealing the holistic properties of language, serving, at least potentially, as one partial solution to the tough problem that ‘‘a language as a whole is much greater than the simple sum of its parts’’. Apparently, the strength of language networks lies in capturing holistic properties of a language. Therefore, a comprehensive knowledge of a language perhaps demands not only network methods but other approaches, such as those based on the theory of information or synergetics. From the perspective of information theory, what is most interesting in a language is its capacity in communication and its coding(decoding) features, such as the information content and distribution of entropy[5], while in terms of synergetics, what is fascinating is how a language could be systematically modeled, how the dynamic relations could be established among different components of a language when it is in real use, and how subsystems in a language are coordinated into a larger system with holistic structure and holistic functions[6]. From a structural perspective, it seems that the approaches based on synergetics may render better complement to complex networks methods in the investigation into properties of languages. References [1] Saussure FD. 1959. Course in General Linguistics (translated by Baskin W) New York: Philosophical Library. [2] L. da F. Costa, O. N. Oliveira Jr, G. Travieso, F. A. Rodrigues, P. R. V. Boas, L. Antiqueira, M. P. Viana, and L. E. C. da Rocha, 2007.Analyzing and modeling real-world phenomena with complex networks: A survey of applications, ArXiv e-prints, vol. 0711, no. 3199. [3] Liu H, Cong J. 2014. Empirical characterization of modern Chinese as a multi-level system from the complex network approach. Journal of Chinese Linguistics, 42: 1-38 [4] Cong J, Liu H. Approaching human language with complex networks. Physics of Life Reviews 2014; [in this issue]. [5] Ferrer-i-Cancho, R., DĊbowski, Ł. & Moscosodel Prado Martín, F. 2013. Constant conditional entropy and related hypotheses.Journal of Statistical Mechanics, L07001. [6] Köhler, Reinhard. 2005. Synergetic Linguistics. In: Köhler, Gabriel Altmann & Rajmund G. Piotrowski (Eds). Quantitative Linguistics. An International Handbook (pp.760---775). Berlin, New York: Walter de Gruyter.
