Tables

Table 1: Dataset

Data collection
Date range	Nov 1, 2019 - Feb 28, 2021
Days collected	486
Handles collected	46
Diplomacy	34
Media	12
Datasets	Number of cases
1. Total original tweets and retweets collected (network analysis)	343.148
1a. Diplomacy original tweets	37.344
1b. Diplomacy retweets	23.512
1c. Media original tweets	253.578
1d. Media retweets	28.714
2. Subsample of original tweets in English (topic analysis)	239.943
2a. Diplomacy	25.830
2.b Media	214.113
3. Coded sample of diplomacy original tweets in English (discourse analysis)	4.879

Table 2: Diplomacy

@handle	User	Followers	Followings	Original tweets in period	Original tweets in English	Retweets by handle in period	Retweets in English	Hashtags in original tweets	Hashtags in retweets	Total tweets in period	Total tweets since created	Date created
@Amb_ChenXu	CHEN Xu, Ambassador, Permant Representative of the P.R.C. to the U.N. office in Geneva, Switzerland	8036	68	299	298	49	46	22	41	348	414	2019.12
@AmbassadeChine	Embassy of the P.R.C. in Paris, France	36568	1136	5485	136	1365	263	2349	2290	6850	8350	2019.08
@AmbCina	Embassy of the P.R.C. in Rome, Italy	35146	151	1553	0	137	95	3715	160	1690	2866	2018.05
@AmbCuiTiankai	CUI Tiankai, former Chinese Ambassador to the U.S.	135871	43	260	250	29	28	150	33	289	416	2019.06
@AmbLiuXiaoMing	LIU Xiaoming, former Chinese Ambassador to the U.K.	121828	46	4606	4465	11	11	2865	10	4617	4745	2019.10
@CCGBelfast	Consulate General of the P.R.C. in Belfast, U.K.	964	8	2	2	2	2	0	0	4	4	2020.03
@China_Lyon	Consulate General of the P.R.C. in Lyon, France	721	323	25	2	179	8	10	164	204	300	2020.03
@ChinaAmbUN	ZHANG Jun, Ambassador, Permanent Representative of the P.R.C. to the U.N.	10395	231	406	381	294	290	174	218	700	1390	2020.02
@ChinaCGCalgary	Consulate General of the P.R.C. in Calgary, Canada	2993*	196*	1292	1260	16	16	49	5	1308	2305	2019.12
@chinacgedi	Consulate General of the P.R.C. in Edinburgh, U.K.	1463	23	37	37	89	86	8	115	126	241	2020.02
@ChinaCGMTL	Consulate General of the P.R.C. in Montreal, Canada	484*	68*	103	26	86	57	39	100	189	1430	2020.01
@ChinaConsulate	Consulate General of the P.R.C. in Chicago, U.S.	3676	146	442	428	81	81	139	61	523	1309	2017.02
@ChinaConSydney	Consulate General of the P.R.C. in Sydney, Australia	5851*	340*	2081	2066	333	321	2285	329	2414	4746	2020.04
@ChinaEmbGermany	Embassy of the P.R.C. in Berlin, Germany	4409	214	929	50	449	354	1217	431	1378	1820	2019.12
@ChinaEmbOttawa	Embassy of the P.R.C. in Ottawa, Canada	13176*	335*	559	543	687	680	27	768	1246	4723	2014.06
@ChinaEUMission	Mission of the P.R.C. to the E.U.	21131	1913	1865	1675	292	286	1427	293	2157	10000	2013.09
@ChinaInDenmark	Embassy of the P.R.C. in Copenhagen, Denmark	1255	511	488	470	300	291	108	361	788	2099	2017.05
@ChinainVan	Consulate General of the P.R.C. in Vancouver, Canada	716*	57*	1	1	2	2	0	0	3	727	2021.02
@Chinamission2un	Mission of the P.R.C. to the U.N.	63414	619	1292	1263	1637	1617	796	1613	2929	4744	2015.04
@ChinaMissionGva	Mission of the P.R.C. to the U.N. office in Geneva, Switzerland	3482	175	1146	1069	1401	1379	584	1076	2547	4526	2015.05
@ChinaMissionVie	Mission of the P.R.C. to the U.N. office in Vienna, Austria	3976	396	655	654	149	148	924	123	804	1081	2019.10
@chinascio	State Council Information Office of the P.R.C.	47975	172	2478	2465	102	100	4062	103	2580	16800	2015.09
@ChineseEmbinUK	Embassy of the P.R.C. in London, U.K.	28965	40	1670	1395	139	138	846	110	1809	2456	2019.11
@ChineseEmbinUS	Embassy of the P.R.C. in Washington, D.C., U.S.	86782	255	1347	1262	648	630	1256	681	1995	2375	2019.06
@CHN_UN_NY	Spokesperson of Mission of the P.R.C. to the U.N.	1438	54	254	253	969	942	24	673	1223	1906	2020.05
@ChnConsul_osaka	Consulate General of the P.R.C. in Osaka, Japan	18171*	741*	477	1	673	59	99	344	1150	3125	2019.09
@ChnEmbassy_jp	Embassy of the P.R.C. in Tokyo, Japan	91103*	774*	1110	7	500	103	27	168	1610	5172	2014.04
@ChnMission	LIU Yuyin, Spokesperson, Permanent Representative of the P.R.C. to the U.N. office in Geneva, Switzerland	1011	97	11	10	189	178	0	22	200	1158	2020.01
@consulat_de	Consulate General of the P.R.C. in Strasbourg, France	955	361	565	6	1412	513	1631	1642	1977	2482	2020.02
@GeneralkonsulDu	DU Xiaohui, Consul General, Consulate General of the P.R.C. to Hamburg, Germany	1581	85	291	23	325	102	292	405	616	690	2020.02
@MFA_China	Ministry of Foreign Affairs, Beijing, P.R.C.	298033	160	1630	1595	1754	1234	636	1222	3384	4177	2019.10
@SpokespersonCHN	HUA Chunying, Spokesperson & Director General, Information Department, Ministry of Foreign Affairs, Beijing, P.R.C.	894507	159	2223	2042	132	122	2136	100	2355	3522	2019.10
@SpokespersonHZM	HU Zhaoming, Spokesperson & Director General, Bureau of Public Information and Communication, International Department, C.P.C. Central Committee, Beijing, P.R.C.	7707	35	97	97	0	0	13	0	97	150	2020.04
@zlj517	ZHAO Lijian, Spokesperson & Deputy Director General, Information Department, Ministry of Foreign Affairs, Beijing, P.R.C.	960093	174492	1665	1598	9081	8423	326	8506	10746	65400	2010.05

(*) metadata retrieved on 25.02.2022, whereas the rest were retrieved on 21.06.2021

Table 3: Media

@handle	User	Followers	Followings	Original tweets in period	Original tweets in English	Retweets by handle in period	Retweets in English	Hashtags in original tweets	Hashtags in retweets	Total tweets in period	Total tweets since created	Date created
@CGTNOfficial	China Global Television Network (CGTN)	13528250	70	44640	43984	12883	12825	44498	7598	57523	174500	2013.01
@chenweihua	CHEN Weihua, China Daily E.U. Bureau Chief and columnist	98677	2814	10015	9417	12768	12349	0	4436	22783	38400	2009.11
@ChinaDaily	China Daily	4284437	537	38612	38412	1658	1649	69934	3343	40270	152400	2009.11
@CNS1952	China News Service	475273	146	20204	0	3	0	4877	1	20207	59000	2013.07
@globaltimesnews	Global Times	1870039	520	59120	58646	573	568	80108	794	59693	191100	2009.06
@HuXijin_GT	HU Xijin, Global Times Editor-in-chief	439720	670	880	880	14	14	2	18	894	2551	2014.08
@PDChina	People's Daily	6928270	4360	17972	17933	93	93	18144	111	18065	99400	2011.05
@PDChinese	People's Daily (Chinese)	753245	332	13577	0	0	0	3851	0	13577	52300	2013.06
@QiushiJournal	Qiushi Journal	1691	158	133	128	0	0	352	0	133	388	2020.05
@shen_shiwei	SHEN Shiwei, CGTN News Producer	36485	4956	4922	4554	377	357	9740	398	5299	6902	2012.05
@XHNews	Xinhua News	12395089	65	40199	40019	29	29	21828	10	40228	202300	2012.02
@XinWen_Ch	Voice of China	4242	1221	3304	140	316	26	208	79	3620	3793	2019.12

Table 4: Followings within the network

	@Amb_ChenXu follows…	@AmbassadeChine follows…	@AmbCina follows…	@AmbCuiTiankai follows…	@AmbLiuXiaoMing follows…	@CCGBelfast follows…	@China_Lyon follows…	@ChinaAmbUN follows…	@ChinaCGCalgary follows…	@chinacgedi follows…	@ChinaCGMTL follows…	@ChinaConsulate follows…	@ChinaConSydney follows…	@ChinaEmbGermany follows…	@ChinaEmbOttawa follows…	@ChinaEUMission follows…	@ChinaInDenmark follows…	@ChinainVan follows…	@Chinamission2un follows…	@ChinaMissionGva follows…	@ChinaMissionVie follows…	@chinascio follows…	@ChineseEmbinUK follows…	@ChineseEmbinUS follows…	@CHN_UN_NY follows…	@ChnConsul_osaka follows…	@ChnEmbassy_jp follows…	@ChnMission follows…	@consulat_de follows…	@GeneralkonsulDu (deleted account)	@MFA_China follows…	@SpokespersonCHN follows…	@SpokespersonHZM follows…	@zlj517 follows…	@CGTNOfficial follows…	@chenweihua follows…	@ChinaDaily follows…	@CNS1952 follows…	@globaltimesnews follows…	@HuXijin_GT follows…	@PDChina follows…	@PDChinese follows…	@QiushiJournal follows…	@shen_shiwei follows…	@XHNews follows…	@XinWen_Ch follows…
@Amb_ChenXu	N/A	x					x	x	x	x		x	x	x		x	x	x	x	x	x			x		x	x	x	x		x	x		x		x							x	x			25
@AmbassadeChine	x	N/A	x			x	x		x		x	x	x	x	x	x					x	x		x		x	x	x	x		x	x		x		x	x			x	x		x	x		x	28
@AmbCina		x	N/A			x						x				x			x		x			x		x	x		x		x	x		x		x			x					x			16
@AmbCuiTiankai	x	x	x	N/A		x	x	x	x			x	x	x	x	x	x	x	x	x	x	x		x	x	x	x	x	x		x	x	x	x	x	x	x	x	x	x	x		x	x		x	38
@AmbLiuXiaoMing	x	x	x		N/A	x	x	x	x	x		x	x	x		x	x	x	x	x	x	x	x	x	x	x	x	x	x		x	x	x	x	x	x	x	x	x	x	x		x	x		x	39
@CCGBelfast						N/A	x							x		x					x		x			x					x	x		x										x			10
@China_Lyon		x					N/A				x		x			x					x			x		x			x					x		x				x			x	x			13
@ChinaAmbUN	x	x				x	x	N/A	x			x	x	x	x	x	x		x	x	x	x		x	x	x	x	x			x	x		x		x	x	x	x				x	x		x	30
@ChinaCGCalgary		x				x	x		N/A	x	x	x	x	x		x		x			x	x		x		x	x		x		x	x		x									x	x			21
@chinacgedi						x	x		x	N/A						x			x		x		x			x					x	x		x										x			12
@ChinaCGMTL							x		x		N/A	x				x					x								x					x										x			8
@ChinaConsulate			x	x		x	x		x		x	N/A			x	x			x		x	x		x		x	x		x		x	x		x		x							x	x			21
@ChinaConSydney		x				x	x		x				N/A	x		x				x	x			x		x			x		x	x		x		x				x			x	x			18
@ChinaEmbGermany		x				x	x						x	N/A		x					x		x	x		x			x		x	x		x	x	x							x	x			17
@ChinaEmbOttawa		x	x			x	x		x		x	x	x		N/A	x	x	x	x		x	x		x		x	x		x		x	x		x		x	x		x	x	x		x	x			28
@ChinaEUMission	x	x	x			x	x	x	x		x	x	x	x	x	N/A	x	x	x	x	x	x	x	x		x	x	x	x		x	x		x		x	x		x		x		x	x			33
@ChinaInDenmark		x					x		x							x	N/A					x		x		x					x	x		x		x											11
@ChinainVan									x		x				x			N/A													x			x						x			x	x			8
@Chinamission2un	x	x	x			x	x	x	x	x		x	x	x	x	x	x	x	N/A	x	x	x	x	x	x	x	x	x			x	x		x		x	x	x	x	x	x	x	x	x		x	37
@ChinaMissionGva	x	x					x	x					x			x			x	N/A	x		x	x		x	x	x			x	x		x		x								x			18
@ChinaMissionVie	x	x	x			x	x	x	x	x	x	x	x	x	x	x			x	x	N/A	x	x	x	x	x	x	x	x		x	x		x		x							x	x			30
@chinascio		x	x				x	x	x	x		x	x	x	x	x	x	x	x	x	x	N/A		x		x	x	x	x		x	x	x	x	x	x	x	x	x		x		x		x		33
@ChineseEmbinUK	x	x	x		x	x	x	x	x	x		x	x	x	x	x	x	x	x	x	x	x	N/A	x		x	x				x	x		x		x	x	x	x		x		x	x			33
@ChineseEmbinUS	x	x	x	x		x	x	x	x			x	x	x	x	x	x	x	x	x	x	x	x	N/A	x	x	x	x	x		x	x		x		x	x	x	x	x	x		x	x		x	37
@CHN_UN_NY							x	x											x		x			x	N/A	x					x	x		x		x											10
@ChnConsul_osaka		x					x		x			x	x	x		x					x	x		x		N/A	x				x	x		x			x							x			16
@ChnEmbassy_jp		x	x			x	x		x			x			x	x	x		x	x	x	x		x		x	N/A		x		x	x		x		x	x	x	x	x	x	x		x			27
@ChnMission		x				x		x								x				x	x					x		N/A						x		x							x	x			11
@consulat_de		x					x		x		x					x					x					x			N/A		x	x		x										x			11
@GeneralkonsulDu (deleted account)																														N/A																	0
@MFA_China	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x		N/A	x	x	x	x	x	x	x	x	x	x		x	x			41
@SpokespersonCHN	x	x	x	x		x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x		x	x	x	x		x	N/A	x	x	x	x	x	x	x		x	x	x	x	x	x	41
@SpokespersonHZM		x				x	x			x			x	x		x				x	x	x		x		x			x				N/A	x		x	x		x				x	x			19
@zlj517	x	x	x	x		x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x		x	x	x	N/A		x	x	x	x	x			x	x	x	x	40
@CGTNOfficial		x	x	x	x			x	x		x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x		x		x	x	x	x	N/A	x	x	x	x	x	x		x	x	x		37
@chenweihua		x				x	x	x					x	x		x			x	x	x			x			x	x						x		N/A	x			x	x			x		x	19
@ChinaDaily	x	x			x	x	x	x	x	x	x	x	x	x	x	x	x	x	x		x	x	x	x	x	x	x	x	x		x	x	x	x		x	N/A	x	x	x	x		x	x	x		38
@CNS1952					x														x			x				x	x					x	x	x		x		N/A				x				x	11
@globaltimesnews		x	x		x	x	x	x	x	x	x	x	x	x	x	x		x	x	x	x	x	x	x		x	x		x		x	x	x	x		x	x	x	N/A	x	x	x	x	x	x		37
@HuXijin_GT	x	x					x						x	x	x	x		x	x					x		x	x	x	x					x		x			x	N/A	x		x	x		x	21
@PDChina		x	x	x	x		x	x	x	x		x	x	x	x	x	x	x	x	x	x	x	x	x		x	x	x	x		x	x	x	x	x	x	x	x	x	x	N/A	x	x	x	x		39
@PDChinese		x	x		x	x	x							x	x	x			x		x			x		x	x		x			x		x		x		x	x		x	N/A		x		x	22
@QiushiJournal		x				x	x				x		x	x	x	x		x			x								x				x	x			x						N/A	x			15
@shen_shiwei		x				x	x	x				x	x	x		x										x	x	x	x					x		x							x	N/A		x	16
@XHNews		x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x		x	x		x	x	x	x		x	x	x	x	x	x		x	x	N/A		39
@XinWen_Ch																																				x										N/A	1
	15	36	20	8	9	29	38	22	29	15	17	26	30	29	23	40	18	20	28	22	38	25	17	34	11	39	29	20	30	0	33	34	13	43	7	36	22	17	21	18	19	6	30	39	7	13

Table 5a: Popular hashtags in original tweets and retweets from all diplomats

Hashtags	Number of cases
#covid19	4783
#china	3537
#coronavirus	1355
#xinjiang	1102
#us	994
#chine	855
#hongkong	753
#wuhan	572
#cina	498
#xijinping	478
#cpec	347
#beijing	340
#bri	294
#ciie	254
#5g	246
#covid	245
#covid_19	240
#who	189
#shanghai	188
#un	180
#poverty	177
#畅游友城	176
#tibet	174
#voyagezdanslesvilleschinoisesjumelées	171
#pompeo	163
#pressroomhighlights	161
#chinese	158
#multilateralism	155
#vaccine	144
#unsc	143

Table 5b: Popular hashtags in original tweets from MFA and SCIO

Hashtags	Number of cases
#covid19	772
#china	484
#us	342
#xinjiang	162
#pressroomhighlights	159
#coronavirus	141
#poverty	116
#wuhan	98
#hongkong	91
#beijing	71
#trade	67
#5g	60
#tibet	59
#pompeo	56
#beltandroad	54
#vaccines	51
#economy	47
#springfestival	44
#xijinping	41
#un	38
#twosessions	37
#cpc	36
#ciie	34
#humanrights	34
#nationalsecurity	33
#who	33
#vaccine	32
#economic	31
#gdp	31
#multilateralism	31

Table 5c: Popular hashtags in original tweets from embassies, ambassadors and missions

Hashtags	Number of cases
#covid19	2612
#china	1935
#xinjiang	633
#coronavirus	599
#chine	514
#cina	485
#hongkong	450
#wuhan	278
#xijinping	247
#畅游友城	176
#covid	172
#voyagezdanslesvilleschinoisesjumelées	171
#us	159
#covid_19	152
#beijing	138
#5g	136
#ciie	131
#xizang	124
#shanghai	115
#nationalsecuritylaw	114
#魅力疆藏	102
#who	100
#uk	98
#covidー19	90
#jiangsu	90
#chinese	89
#multilateralism	89
#climatechange	85
#hksar	85
#forzacinaitalia	83

Table 6: Foreign influencers mentioning Chinese diplomats and media

@handle	User
@AndyBxxx	Andy Boreham (Reports On China)
@BarrettYouTube	Lee and Oli Barrett
@BeehiveChina	Barrie Jones (Best China Info)
@ChinaTeacher1	Fernando Munoz Bernal (FerMuBe)
@DanielDumbrill	Daniel Dumbrill
@JaYoeLife	Matthew Galat
@Jingjing_Li	Li Jingjing
@LivingChina	Jason Lightfoot (Living in China)
@Noel_Calibre	Noel Lee
@thecyrusjanssen	Cyrus Janssen

Influencers identified by ASPI: https://www.aspi.org.au/report/borrowing-mouths-speak-xinjiang

Figures

Figure 1: Top mentionees (in-degree) and mentioners (out-degree)

Figure 2: Total mentions to handles in network from all Twitter users

Figure 3: Total mentions between handles in network

Figure 4: Mentionees in network (in-degree)

Figure 5: Mentioners in network (out-degree)

Figure 6 (diplomacy) open here

Figure 7 (media) open here

Figure 8a

Figure 8b

Figure 8c

Figure 8d

Extract data

The code used for creating dataframes from the JSON files can be found in the extract_data folder.

Topic model

Latent Dirichlet Allocation topic modelling using gensim package in Python (See documentation: https://radimrehurek.com/gensim_3.8.3/models/ldamodel.html). LDA is a hierarchical Bayesian model with three levels, in which each item of a collection, in this case tweets, is modeled as a finite mixture over an underlying set of topics. In turn, each topic is modeled as an infinite mixture over an underlying set of topic probabilities. An explicit representation of each tweet is provided by the topic probabilities.

A total of 180 models were trained for both diplomat and media tweets with a variation of the following three hyperparameters:

• Number of Topics (K)
  • The topic model was trained requesting 10, 15, 20, 25, 30 and 35 latent topics
• Dirichlet hyperparameter alpha: A-priori document-topic density
  • The topic model was trained using 6 different a-priori beliefs about the document-topic density, including 0.01, 0.31, 0.61, 0.91, symmetric and asymmetric
• Dirichlet hyperparameter beta: A-priori word-topic density
  • The topic model was trained using 5 different a-priori beliefs about the word-topic density, including 0.01, 0.31, 0.61, 0.91 and symmetric

The model with the best coherence score is chosen for analysis.

Usage:

1. Navigate to the topic model folder

cd topic_model

2. Install requirements

• Pip install

pip install -r requirements.txt

• Download en_core_web_sm

python -m spacy download en_core_web_sm

3. Lemmatization and cleaning of tweets

python preprocess/prep_text.py

4. Generate multiple models with a variety of hyperparameters

python preprocess/gen_model.py

5. Evaluate the topic models created above, and determining which is the best one

python preprocess/eval_model.py

When the model has been generated using above commands, run the code in the topic_model.ipynb to visualize the results. Furthermore, visualisations of how prevalent each topic was over time (averaged topic weight) can be found in the topics_over_time.ipynb.

Semantic kernel

There are three parameters which can be adjusted:

• First order associations: Indicates the number associations wanted from each of the seeds (Written in uppercase letters along side the seeds in the graph)
• Second order associations: Indicates the number of associations wanted from each of the first order associations. (Written in lowercase in the graph)
• Pruning: Can be set to none, soft and hard. According to the pruning settings, words that are not linked closely enough to the rest of the graph across the hierarchical levels are removed.

Input given to the model is lists of seeds. Each list should be written as a seperate txt file in the res folder.

Usage:

1. Navigate to the semantic kernel folder

cd semantic_kernel

2. Prepare data for semantic kernel (the data used is what was preprocessed for the topic model)

python prep_semantic/create_subsets.py

python prep_semantic/csv2txt.py -i data/text_diplomat.csv -o data/data_semantic/text_diplomat
python prep_semantic/csv2txt.py -i data/text_diplomat_orig.csv -o data/data_semantic/text_diplomat_orig

3. Train model and generate graphs

First time running make sure to set train to True

cd semantic_kernel/semantic-kernel

run_diplomats.sh
run_diplomats_orig.sh

4. Tweaking of parameters

• Pruning: modify the txt file in prun folder
• First and second order associations: modify the txt file in assoc folder

Network Analysis

Network analysis performed using the networkx package in python (https://networkx.org/) and the network visualizations are generated from the file network_main.py (see usage below). Nodes in the networks are Twitter handles, and edges (connections) are weighted by the number of mentions between the Twitter handles that are displayed. The network visualizations only plot Twitter handles that are either flagged as (i) Chinese diplomats or (ii) Chinese media outlets. The edgewidth (strength of connections) is determined by the number of mentions between Twitter handles of Chinese diplomats and media outlets (see below). The nodesize (size of handle) is determined by various attributes, such as:

• total mentions (Figure 2): number of total mentions to the Twitter handle in question from all users (also non-diplomats and non-media that are not shown as nodes in the plot). This shows how "popular" the Chinese diplomats and media outlets are on Twitter broadly, rather than just their popularity/activity within the diplomat/media sub-network.
• weighted degree (Figure 3): node-size scaled by number of total number of connections between Twitter handle in question and other Chinese diplomats and media outlets (both directions counted, and each mention counted). The weighted degree plot corresponds to in-degree + out-degree (i.e. we count both directions).
• in-degree (Figure 4): number of mentions from other Chinese diplomats and media outlets to the Twitter handle in question (only one direction counted).
• out-degree (Figure 5): number of mentions from the Twitter handle in question to other Chinese diplomats and media outlets (only one direction counted).

In addition to the network visualizations, we also show the top 10 handles (based on weighted degree) in Figure 1. The plot is generated in summary_stats_focus.py (see usage below). Clearly, some handles are primarily mentionees and have high in-degree (e.g. CHNews) while others are primarily mentioners and have high out-degree (e.g. zlj517) within the diplomat/media sub-network.

Usage:

1. Activate environment

source cnenv/bin/activate

2. Navigate to the network code folder

cd networks/src

3. Run bash script

bash main.sh

in main.sh set:
PRE=true
NET=true
SUM=true

This ensures that the bash script calls (runs)

1. preprocessing (concat_files.py)
2. network visualizations (network_main.py)
3. summary data analysis (summary_stats_focus.py)

Bot Detection

We train a logistic classifier on the cresci-2017 (Cresci et al., 2017) data set (available: https://botometer.osome.iu.edu/bot-repository/datasets.html) to classify Twitter handles as genuine or spam/bot/fake. We use the widely used fofo metric (e.g. Yang et al., 2013; Tavazoee et al., 2020) which is (following/followers) of an account. We use (following+1/followers+1) to avoid division with zero, and when an account appears more than once in a data set we use only the last appearance (i.e. the number of following and followers for the handle at that time). The intuition behind the metric is that bot-accounts tend to follow many other accounts (following) but they tend to have few followers. This means that they will generally have a high fofo-ratio (i.e. high following, low followers). Using the trained model, we estimate the fraction of genuine accounts vs. spam/bot/fake accounts in our own data set, as well as in a baseline data set consisting of vaccine-related tweets from 2020-2021 (https://www.kaggle.com/datasets/gpreda/all-covid19-vaccines-tweets). We estimate 27.22% of the accounts in the baseline (vaccine) data set to be non-genuine accounts and 46.44% of accounts in our data set of Chinese state media and diplomats to be non-genuine accounts. There is considerable uncertainty around this estimate since (1) our data set might differ in other respects than the amount of bot-activity from the baseline data set and (2) while the fofo-metric is widely used (Yang et al., 2013) it is not universally found to be accurate in detecting bots.