Linking GDELT 2.0 GKG Data to Real-World Events

Figure 11: Monthly average tensions in the South China Sea from March 2015 to November  2017 (based on GDELT 2.0 GKG)

Note: Tensions are monthly aggregates based on the inverse of tone values on a negative–positive  spectrum theoretically from -100 to +100 but typically no more than -10 to +10 in practice. Higher  values represent higher tensions (i.e., more negative tone); lower values represent lower tensions (i.e.,  more positive tone). 

3.2.5 Linking GDELT 2.0 GKG Data to Real-World Events 

As with the GDELT 1.0 Event Database, the general validity of the tensions dataset  derived from the GDELT 2.0 GKG can be confirmed by manually comparing the  events occurring within a given month to the level of tensions for that month  ( Tensions ). In terms of time periods with lower tensions, it is clear from the time  series that, in November 2016, South China Sea tensions were at their lowest (i.e.,  more positive tone) in recent years. This corresponds precisely with the findings  described in {3.2.2 Linking GDELT 1.0 Event Database to Real World Events}. In that  month, a series of events are likely to have contributed to the reduction in tensions,  including an agreement between China and Malaysia on South China Sea 

cooperation, an agreement between China and the Philippines concerning bilateral  mechanisms for dispute management, Philippine statements on setting aside the  disputes and Beijing’s willingness to allow fishing in the Scarborough Shoal area, 

‧

國

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

and suggestions that the US could decrease its military involvement in the South  China Sea. These events are covered in greater detail in {3.2.2 Linking GDELT 1.0  Event Database to Real World Events}. 

It should be reiterated here that tensions derived from the GDELT 2.0 GKG  are based on tone values, and these have varying levels. Higher tensions imply a more  negative tone in relevant news articles, whereas lower tensions imply a more positive  tone. These values represent the difference in the overall percentage of words that  are negative or positive, respectively. For example, if an event record contains 5% 

negative words and 3% positive words, the resulting tone would be -2. These are then  inverted and aggregated into monthly averages representing tensions so that a 

month with tensions of +1.6 has higher tensions than one with tensions of +1.4.  

As for periods of heightened tensions, the dataset suggests that South China  Sea tensions reached a peak in August 2017. That month began with reports of  Vietnamese foreign ministry spokeswoman Le Thi Thu Hang condemning the  opening of a movie theater on China-occupied Woody Island in the Paracel Islands  following a Xinhua report on the issue. The following day, Philippine Senior ¹²⁸ Associate Justice Antonio Carpio was quoted as stating that a joint venture in the  South China Sea between the Philippine government and China as a sovereign entity  would be unconstitutional. On August 7, Australian Minister for Foreign Affairs ¹²⁹ Julie Bishop, Japanese Minister for Foreign Affairs Taro Kono, and US Secretary of  State Rex Tillerson, a er meeting in Manila, released a joint statement, in which  they “expressed serious concerns over maritime disputes in the South China Sea” and 

“reiterated that the three countries will continue to fly, sail, and operate wherever 

128  “Vietnam criticizes China’s cinema on disputed South China Sea island,” Reuters, August 1, 2017, 

<https://www.reuters.com/article/us-vietnam-china-southchinasea/vietnam-criticizes-chinas-cinem a-on-disputed-south-china-sea-island-idUSKBN1AH46E>. 

129  Edu Punay and Alexis Romero, “‘Joint venture with China on South China Sea violates Charter’,” 

The Philippine Star, August 2, 2017, 

<http://www.philstar.com/headlines/2017/08/02/1723769/joint-venture-china-south-china-sea-violate s-charter>. 

‧

國

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

international law allows.” Shortly therea er, ASEAN member states and China ¹³⁰ adopted a framework for moving forward with a Code of Conduct in the South China  Sea a er fi een years of minimal progress since the signing of the Declaration on  the Conduct of Parties in the South China Sea in 2002, but the framework was  immediately criticized as being heavily influenced by Beijing and unlikely to make a  meaningful contribution to dispute management in the future. Later in the month, ¹³¹ the Chinese Ministry of Foreign Affairs submitted an official complaint to the 

Indonesian embassy in Beijing on August 25, demanding that the Indonesian 

government rescind its decision from the previous month to rename its claimed part  of the South China Sea as the North Natuna Sea. Taken together, these events ¹³² serve to confirm the validity of the data by demonstrating that a series of occurrences  would likely have contributed to a measurement of heightened tensions (i.e., more  negative tone) based on the dataset. 

The tensions dataset based on the GDELT 2.0 GKG can also be compared to  the more recent months in the dataset derived from the GDELT 1.0 Event Database. 

In many instances, such as November 2016 covered above, similar observations can  be made when looking at corresponding time periods. In some instances, the 

datasets differ. Both these similarities and differences are to be expected. Similarities  result from the fact that this dissertation uses both of the two sources as 

interpretations of South China Sea tensions and some of the time periods overlap  with one another. Differences can be attributed to the different ways that tensions  are defined using the two datasets, with Goldstein values being used in the GDELT  1.0 Event Database and tone values being used in the GDELT 2.0 GKG. Whereas 

130  “Australia-Japan-United States Trilateral Strategic Dialogue Ministerial Joint Statement,” Office of  the Spokesperson, US Deparment of State, August 6, 2017, 

<https://www.state.gov/r/pa/prs/ps/2017/08/273216.htm>. 

131  Christian Shepherd and Manuel Mogato, “ASEAN, China adopt framework for cra ing code on  South China Sea,” Reuters, August 6, 2017, 

<https://www.reuters.com/article/us-asean-philippines-southchinasea/asean-china-adopt-framewor k-for-cra ing-code-on-south-china-sea-idUSKBN1AM0AY>. 

132  Saifulbahri Ismail, “China demands Indonesia rescind decision to rename part of South China  Sea,” Channel NewsAsia, September 2, 2017, 

evidence from the GDELT 2.0 GKG suggests that South China Sea tensions peaked  in August 2017, that peak was seen in the GDELT 1.0 Event Database in the 

following month. Because Goldstein values ( GoldsteinScale ) measure explicit  actions (e.g., +1.0 “Meet with”; +3.0 “Agree to future action or procedure, to meet or  to negotiate”; +3.4 “Praise, hail, applaud, extend condolences”; +6.5 “Make 

substantive agreement”) between two countries, events in such months could ¹³³ suggest lower tensions (i.e., more cooperative actions). In contrast, because tone  values ( V2Tone ) measure the overall percentage of words that are positive or negative  related to a given event record, events in the same month could suggest higher  tensions (i.e., more negative tone). In the news media, the cooperative events of  August 2016 were o en presented in the broader context and with further analysis of  the South China Sea disputes. Such similarities and differences should be taken into  consideration when making conclusions based on the findings presented in this  dissertation. 

3.2.6 State Involvement Data Selection from GDELT 2.0 GKG for RQ1