The Global Language Network.

From the very beginning, languages have played a major role in human society. The languages have evolved from tongue movements, which use to mimic the sounds of beasts and other animals, to very structured and sophisticated use of words. Around the world, various civilizations have emerged and each has its own unique way to expressing information in the form of language. With the various wars,trades and conquests various civilizations annexed their territories, population and resources, this also resulted in the popularizing their language and bringing diversity to their language. Languages take birth, they evolve and they die. Many languages exists in the world at present and each has it own significance in the society. There is no particular way of determining the influence of a particular language in the global network. Some of the people in MIT media labs have tried to model the effect of the language in the global network.
Basic Assumptions-
In the words of linguist David Crystal, “Why a language becomes a global language has little to do with the number of people who speak it. It is much more to do with who those speakers are.”  The basic assumptions for finding the global influence of a particular language is who speaks that language and how are the people speaking that language are connected. In past , Latin was pan-European language because it was official language of Roman Empire. In present world scenario if we want to focus on those connecting international elites, then we should look at the multilingual coexpressions of these individuals.
Gathering Data-
People at MIT have found out three ways to find the multilingual coexpressions of the elites.The nodes in the network are various languages and edges are created in following manner with given database.
1) Translated Books-  More than 2.2 million book translations compiled by UNESCO’s Index Translationum project is used for this study. An outward edge in book network signifies that a book is translated from current language to another.
2) Wikipedia Edits- Two Languages are connected if a person who has edited a wikipedia article in one language has also edited article in another language.
3) Tweets-  Two languages are connected when a person tweets in two or more languages.
In all the above ways we can create a network and study various network properties. Visual representation of the network under various settings follows.




 If you want to have a good visualization of these network you should go to this website- http://language.media.mit.edu/visualizations/books .
Now when we have a connected weighted graph we can find the eigenvector centrality of each node and rank languages according to it. The one having higher eigenvalue is considered to be more prominent global language.

Based on the books translated network top 10 languages are-

Language Centrality Translations From Translations To Speakers (M) English 0.89803531 1225237 146294 1500 French 0.29695532 216624 238463 200 German 0.26334749 201718 292124 185 Italian 0.09374308 66453 59830 70 Russian 0.08565274 101395 82772 278 Spanish 0.08539987 52955 228910 500 Japanese 0.04398496 26921 130893 132 Dutch 0.03955701 18978 111371 27 Latin 0.03404642 19240 362 0 Swedish 0.03363697 39192 71688 10
Based on the Wikipedia Edits Top 10 languages are- Language Centrality Edits Editors Multilinguals # Speakers (M) English 0.65929841 198361048 1589250 218320 1500 German 0.4787209 33977378 224215 70281 185 French 0.33783444 23070757 142795 45439 200 Spanish 0.28746349 13645596 145487 41700 500 Italian 0.15703838 11923658 72981 21393 70 Russian 0.1516584 12445887 81925 23828 278 Dutch 0.13486947 6393791 46951 18050 27 Japanese 0.12399229 16149315 102857 18578 132 Portuguese 0.10986822 5168734 60487 15213 290 Polish 0.0938191 6589015 47015 13932 43
Based on Twitter Data Top 10 languages are-
Language Centrality Tweets Users Multilinguals Speakers (M) English 0.69329476 255351176 10859465 3129456 1500 Malay 0.48559568 49546710 1651705 1033758 300 Portuguese 0.35208674 46520572 1617409 708142 290 Spanish 0.34811446 44195979 2043468 769887 500 Filipino 0.13316405 1905619 257611 250959 90 Dutch 0.10582998 10526980 435128 231807 27 Arabic 0.05422551 9993172 366643 119350 530 Japanese 0.04418507 91669691 2602426 124833 132 French 0.03696668 3434065 147843 84199 200 Italian 0.02517954 1586225 89242 52558 70
The key inference we make from this network is that English is the most prominent language in all networks. French and German are top ranked in books and wikipedia but not in the case of Twitter data. Indian languages like Hindi, Bengali, Gujrati, Tamil, Telgu, Marathi, Punjabi has good number of speakers but they are nowhere in rankings. Their eigenvector centrality is close to zero. This is because of the fact that the intellectual elites of the countries do not converse in these languages though these being their mother tongues. These results shows that though languages which are popular have very less global impact. With this network we can clearly sense that the languages with high popularity and low centrality are dying, as intellectual elites are preferring other languages over these. As the time will pass these languages will have higher chances of becoming extinct and more central languages taking their places.
Refrence - http://language.media.mit.edu/