The most active governments in the Internet Governance Forum (IGF) process since 2006 are five hosts (Greece, Brazil, Egypt, Lithuania, and Kenya) and five other countries (the United Kingdom, Switzerland, France, the USA, and Sweden). Only 28% of countries spoke at the IGF (54 out of 193 UN member states). Egypt is both the most active country in the IGF and the regional leader, contributing 52% to all verbal inputs from Africa. These are some of the results from the pilot study of the language corpus of the IGF (2006–2011) conducted by DiploFoundation.
The study is based on official statements delivered by government representatives at the IGF and during the preparatory process. Due to the illegibility of some transcripts, 53% of the statements were attributed in the pilot study. The full study should increase attribution to 90%. Diplo is in the process of establishing a consortium which will conduct the full study of the IGF language as well as a comparative study of other Internet governance policy processes that have transcription records (ICANN and ITU meetings).
This analysis is concerned with the distribution of word frequency among the representatives of national governments at the IGF. We first study the volume of contribution on behalf of the representatives of different governments in order to find out which where the most talkative over the period covered. Then we try to figure out the most important factors that influence the volume of contribution. We also discuss the differences in volume of verbal contribution to the IGF between world regions based on the UN region classification.
Next, we use word frequency distributions for the representatives of different countries as a tool to describe the overall semantics of their specific Internet diplomacy discourse. The latter enables us to develop a semantic map, similar to the one already presented in the section on multistakeholderism, a map where governments are represented by points and the similarity among respective word frequency distributions by distance. We show how meaningful patterns emerge in the language of Internet diplomacy, opening possibilities for more elaborate, in-depth research.
These analyses are based on word frequency distributions that were drawn from the pilot version of the IGF text corpus, where approximately 50% of the complete verbal contribution to the IGF was automatically tagged and mapped onto relevant actors and events. From the analysis of this initial version of the text corpus, we find that representatives of only 54 countries verbally contributed to the IGF. The analysis encompasses the verbal contribution of government representatives no matter whether they intervened during the IGF main sessions – where they represented their governments directly – or whether their interventions originated at different events alongside the main sessions (at workshops and/or preparatory meetings, perhaps).
The most active governments
Figure 1 presents the volume of contribution on the behalf of the representatives of these 54 countries for six IGF years 2006–2011 overall. Egypt was the most active, followed by the United Kingdom, Brazil, Kenya, Switzerland, Greece, France, the USA, Lithuania, and Sweden in the top ten. It is not surprising that among the top ten most verbally productive governments at the IGF 2006–2011 were the governments of host countries (Greece – IGF Athens 2006, Brazil – IGF Rio de Janeiro 2007, Egypt – IGF Sharm el Sheikh 2009, Lithuania – IGF Vilnius 2010, and Kenya – IGF Nairobi 2011). Only India, the country that hosted the IGF 2008. in Hyderabad, is missing from the top ten (it ranked 14th, after Latvia, China, and Canada).
The host countries made more intensive contributions in preparatory meetings and official parts of each IGF (welcome addresses, opening speeches). Figure 2 presents the volume of contribution for the six IGF host countries 2006–2011; it can be readily seen that the total number of words contributed on the behalf of the respective country’s government officials is significantly higher compared to their contribution at the IGF meetings that were not hosted in their countries.
Figure 1. Total verbal contribution to the IGF 2006–2011 for 54 countries.
The scale is given in thousands of word occurrences.
IGF 2012 Baku update: The following figure presents the volume of contribution for the countries whose representatives spoke at the IGF 2012 in Baku.
Figure 1b. Total verbal contribution to the IGF 2012 Baku.
The scale is given in thousands of word occurrences.
Figure 2. Volume of contribution to the IGF 2006–2011 for the six IGF host countries. The pattern is certainly not absolutely consistent, with Brazil, Greece, and Egypt providing a steadily high contribution over the years, but the pattern of notably high verbal contribution on the behalf of the government representatives in the year in which a particular country hosted the IGF is obvious.
Figure 3. Volume of contribution for the top contributors who never hosted the IGF themselves.
In order to get a better insight into the most active governments at the IGF, we provide in Figure 3 a breakdown per year of verbal contribution for the representatives of another five national governments. Two interesting trends are noticeable. Switzerland, as the host of the World Summit on the Information Society (WSIS) Geneva in 2003 and a key player in the WSIS process, was among the most active participants in the first four IGFs. The USA, who had reservations about the establishment of the IGF at WSIS Tunisia in 2005, increased its involvement in the IGF process in the last two years. This is an indication of US support for the IGF.
Figure 4 presents the overall 2006–2011 volume of verbal contribution from five world regions. We have used the UN classification schema assigning each country to the one of the following five regions: Latin America and the Caribbean, Asia-Pacific, Eastern Europe, Africa, and Western Europe and Others (including the United States and Australia). The analysis shows regional IGF leaders. In Africa, it is Egypt with 52% contributions from Africa. The United Kingdom made 20% statements in the most active region of Western Europe and Others.
Figure 4. Volume of contribution by UN world region.
The semantic geography of the IGF
The semantic geography is based on word frequency distributions and used to derive the overall similarity between the patterns of word usage by governments. The analysis starts with identifying the 100 most frequently used words on the behalf of government representatives in the IGF 2006–2011, followed by analysis of usage of all thus identified words by each government. By using the attributive multidimensional scaling methodology , country positions were described by points in space (we represent the countries by points in space exactly as we have represented different stakeholders in different IGF years in the analysis of multistakeholderism). We have used the obtained distributions of word frequency to compute the distances between the countries such that the closer the two points representing two particular countries stand, the more similar their patterns of word usage are.
Figure 5a presents a two-dimensional semantic space – a map, better say – with all 54 countries from our sample. Remember, the closer they stand on the semantic map, the more similar their patterns of word usage tend to be. Compare the relations depicted in Figure 5a with your real-world perception, experience and expertise in IG-related issues and draw a conclusion as to whether the map corresponds – at least partly – to the overall, realistic pattern of similarity among the ideas proposed, the concepts used, and the topics discussed on behalf of the representatives of national governments at the IGF 2006–2011.
Figure 5a. Semantic map of the IGF 2006–2011 representing the structure of similarity among the patterns of word usage on the behalf of the representatives of national governments. The dense region on the left is magnified in Figure 5b.
Figure 5b. The magnified dense region found on the left of the semantic map presented in Figure 5a.
Remember, the farther away from each other two countries are (two points) found, the less similar their specific patterns of word usage seem to be. By examining Figure 5a, we conclude that only those governments with the highest total volume of contributions were able to distinguish themselves from the overall highly similar group of word patterns clustered in the dense region on the left of the map. These countries – practically, the only ones whose names can be read from Figure 5a – are exactly those countries that were verbally the most productive during the six years of the IGF up to 2011. The volume of verbal contribution triggered linguistic diversity.
In order to keep the technical side of this analysis as simple as possible, we are not telling you the full story here. We have actually produced a ten-dimensional semantic space in order to represent the similarities and differences between the verbal contributions representatives of different governments, but the previous two figures reveal the configuration in only two dimensions. In the following four figures (6a–6d), we present the configuration in the first three dimensions of the complete solution, given in different perspectives in order to ease the interpretation. Now the structure of similarity among the patterns of word usage reveals even more detail, as an additional dimension of representation becomes available.
Figures 6a–b. A three-dimensional semantic space of word usage patterns.
Figures 6c-d. A three-dimensional semantic space of word usage patterns.
 Each country was represented by a word frequency vector on the 100 most frequently used words on the behalf of the representatives of national governments in the IGF in general. The Euclidean distance matrix was computed from such descriptions of all 54 countries, and an algorithm for non-metric multidimensional scaling (MDS) was then used to produce the configuration. Aside from small differences in the relative positions of the countries represented by the dense region of Figure 5a (the region magnified in Figure 5b), there was not much difference between the configurations in the first two or three dimensions whether they were derived from two-, three- or ten-dimensional MDS solutions. Also, the whole procedure was initially performed by using the complete word frequency distribution, relying on word frequency vectors of length more than 6000 words to describe each country’s pattern of word usage; again, we were not able to detect any qualitatively important differences when we have used the 100 most frequently words only compared to the MDS solution derived from the complete word frequency distribution.