Visualizing The Network Structure Of Multilingualism In 2.5 Million Hours Of Global Television News

Yesterday we unveiled a first glimpse at the linguistic landscape of the Internet Archive TV News Archive's 2.5-million-hour uncaptioned global archive spanning 100 channels from 50 countries over portions of the last 24 years. One of the most fascinating results of that analysis was the identification of 9.44 million language transitions across the 2.5 million hours in which speech in one language was followed by speech in a different language. To explore the network structure of the multilingualism of the TV News Archive in more detail, we examined those 9.44 million transitions, filtering them to language pairs with at least 11 transitions in either direction (A->B or B->A), yielding 2,247 transitions, which we visualize below in Gephi using Force Atlas 2 for layout, modularity for coloration and PageRank for node size.

You can download the CSV file to analyze yourself.