Linguistic Network

1
Linguistic Network

• Word Adjacency
• Anthony Strathman
• Yilin Wu

2
What are linguistic networks?

• Introduction

3
Motivation

• The study of word connections is important for
• The structure and evolution of a language.
• Cognitive science the transition from
  non-syntactic to syntactic communication involved
  the co-evolution of language and brain.

4
History of Studies

• Zipfs law the frequency of words follows
  power-law distribution with respect to the rank
  in the frequency table.
• (Cf http//en.wikipedia.org/wiki/Zipf's_law)
• It does not involve the interactions between
  words. So does not provide deep insight to the
  organization of a language.

5
Network Approach

• A convenient tool to study the interaction of
  words, thus providing the organization and
  dynamics of a language
• People have study word network from various
  perspectives (synonym, polysemy, word
  cooccurrence, etc. ). Most results show small
  world and scale-free properties of language. (see
  references)
• For example, in the word cooccurrence studies,
  Cancho and Sole (2001) showed that (1).the
  average distance between any two two words is d
  2 3 (2).a scale-free distribution of degrees.

6
Word Adjacency Network

• Project Description

7
Network Definition

• We take our nodes to be individual words
• The nodes are said to be linked if they appear
  adjacent to each other in a sentence
• All links are taken to be bidirectional

Excerpt from The Prince Such dominions thus
acquired are either accustomed to live under a
prince, or to live in freedom and are acquired
either by the arms of the prince himself, or of
others, or else by fortune or by ability.
8
Motivation

• Studying linguistics networks can give insight
  into the structure of natural language
• Looking for any structural drift over time

9
Source Material

• Four different texts
• The Prince (Niccolo Machiavelli)
• A Tale of Two Cities (Charles Dickens)
• Gone with the Wind (Margaret Mitchell)
• Crime and Punishment (Fyodor Dostoyevsky)

10
The Resultant Networks

• Results

11
Example Network

• Trimmed version of network from The Prince

12
General Characteristics
13
Degree Distribution
14
Clustering
15
Clustering
16
Clustering
17
Assortativity
18
Shortest Path Length and Diameter

• The measured diameter of all four networks was 2
• The distribution of path lengths is remarkably
  narrow (1.99 with standard deviation of around
  0.01)
• Given the difference in size of these networks,
  (ranging from 5000 to 20000 nodes) we see no
  change in diameter with size.

19
Conclusions

• The word adjacency networks of all four sources
  are very similar in structure
• By this measure, grammatical structure seems
  fairly stable in time
• The structure seems to be independent of the
  length of the text.

20
References

• Dorogovtsev, S. N. and Mendes, J. F. F., Language
  as an evolving word web, Proc. R. Soc. London B
  268, 26032606 (2001).
• Ferrer i Cancho, R. and Sole, R. V., The small
  world of human language, Proc. R. Soc. London B
  268, 22612265 (2001).
• Sigman, M. and Cecchi, G. A., Global organization
  of the Wordnet lexicon, Proc. Natl. Acad. Sci.
  USA 99,17421747 (2002).
• Steyvers, M. and Tenenbaum, J. B., The largescale
  structure of semantic networks Statistical
  analyses and a model for semantic growth,
  Preprint cond-mat/0110012 (2001).
• Corominas, B. and Sole, R. V. (2006) Network
  topology and self-consistency in language games.
  Journal of Theoretical Biology, 241(2)438--441.
• Motter, A. E., de Moura, A. P. S., Lai, Y-C., and
  Dasgupta, P. (2002) Topology of the conceptual
  network of language. Physical Review E,
  65(065102).
• Text samples are provided from the database
  search
• http//onlinebooks.library.upenn.edu/search.html
• Project Gutenberg