Hcss report

rhetorical side of assertiveness
, we used three main tools: 
• The first one is 
GDELT
, the open-source Global Database of Events, Language, 
and Tone that was first released in 2013. It covers millions of full-text newspaper 
articles published since 1979 and applies various coding and natural language 
processing tools to them in order to automatically extract events (almost half a 

---

HCSS REPORT
15
billion categorized events, with 120.000 being added daily), actors, geographical 
locations and ‘tone’ (negative/positive). For example, if an article contains the line 
“Sudanese students and police fought in the Egyptian capital” it codes the event 
as “SUDEDU fought COP”. Next, the system finds the nearest mention of a city or 
locality in the text – in this case Cairo – and adds its latitude and longitude to the 
event data. One of the important coding distinctions GDELT provides is between 
‘verbal’ and ‘material’ instances of either cooperation or conflict. The HCSS team 
thus was able to identify the codes that could be classified as ‘assertive’ and to 
examine those events that involved China and Russia and were coded as evidence 
for rhetorical assertiveness.
35
• The second one is the 
HCSS Off-Base
36
which contains all webpages of the official 
websites of (so far only) the Ministries of Foreign Affairs of 7 important powers, 
including China and Russia. These websites represent one the most authoritative 
(and in our opinion underexplored) sources for foreign policy analysis, as they 
reflect how countries want to present their foreign policy positions to the rest of 
the world. The HCSS team also applied a number of mostly open-source software 
tools to extract the main topics from those websites and to zoom in on those 
N-grams (single words or combinations of 2 or 3 words) that are associated with 
‘assertiveness’. We first identified ‘baskets’ of expressions (e.g., “China demands”) 
that seemed to clearly zoom in on those countries’ own assertiveness (and not 
that of others) and then identified all N-grams that co-occurred within two 
sentences of the occurrence of any part of that basket (i.e., in two sentences 
before and after the sentence in which the term “China demands” occurs).
• We also scanned all