top of page

Inter-edition agreement

We investigate, if national historical timelines are consistent across languages. For that, we compute a measure of their divergence across Wikipedia editions. Based on the extracted probability distributions of years, for each country we compute a matrix of pairwise inter-language dissimilarities, using the Jensen-Shannon (J-S) divergence.

Inter-language consensus in Wikipedia articles on national histories, based on pairwise Jensen-Shannon divergence values. Countries in the lower left of the plot show the highest consensus across editions. Stars represent data centroids for countries of the same region. The plot shows a high inter-language consensus on average, though the descriptions are not identical across editions. European countries exhibit the highest amount of consensus.

Data points in the lower left quarter correspond to countries with the lowest medians and the narrowest distributions of J-S scores (i.e. the smallest differences between the most similar and the most different language pair), and thus, with the highest inter-lingual consensus.

 

Overall, J-S scores are centered around very low values (medians between .06 and .16), which indicates a high average agreement across language editions. Their spread covers a higher range (up to .35), implying the presence of large differences between some language pairs.


Based on the location of data centroids (stars), we observe higher interlingual consensus on the history of European and African countries, compared to Americas, Asia, and Oceania. The largest interlingual disagreement is found in the articles on the history of Australia, Malawi, Madagascar, China, Japan, and the Netherlands; some with the highest consensus are Liechtenstein, Belgium, the Democratic Republic of Congo, and Malaysia.

In case of China , for example, high disagreement is partially driven by the differences between Russian and other language editions. This is especially evident during the period of Sino-Soviet split in 1960-80s, which is less densely covered in the Russian language Wikipedia.

Timelines of history of Belgium, on the other hand, are almost identical across all 30 language editions (we present only 6 largest editions in order not to obstruct the view).

Inter-lingual consensus on histories of selected countries. To illustrate cases with very high (Belgium) and very low (China) inter-lingual consensus, we present parts of probability distributions of dates zoomed into 1750-2010s, for 6 large editions. Chinese timeline in Russian Wikipedia differs noticeably from the timelines in all other editions, while for Belgium all timelines are almost identical.

Illustration of countries with large and small inter-edition consensus
bottom of page