(1) Context and motivation
The YouTube Disinformation Comments Corpus dataset concerns a smear-campaign maligning Haitian immigrants to Springfield, Ohio amplified by the 2024 Republican nominees for President and Vice President, Donald Trump, and J.D. Vance, respectively. In early September 2024, Springfield, Ohio resident, Erika Lee published a Facebook post in which she claimed that her neighbor’s daughter’s cat had been stolen and butchered by a family of Haitian immigrants. This claim was then reposted on Twitter on September 5, 2024, where it was amplified across right-wing media ecosystems like Fox News (see Howard and Brewster, 2024 September 12). GOP vice presidential nominee, J.D. Vance then posted to X the following:
Months ago, I raised the issue of Haitian illegal immigrants draining social services and generally causing chaos all over Springfield, Ohio.
Reports now show that people have had their pets abducted and eaten by people who shouldn’t be in this country. Where is our border czar? (@JDVance, 2024 September 9)
Vance’s Tweet has garnered 11.4 million views and 27.9 thousand shares through Retweets and Quoted Tweets. Then, during the September 10, 2024, US presidential debate, Trump repeated the story, stating “They’re [Haitian immigrants in Springfield, OH] eating the dogs. They’re eating the cats. They’re eating the pets of the people that live there,” (qtd in Catalini et al., 2024). Such statements have been roundly debunked by state official and journalistic outlets (Orazco, 2024; Lozano, 2024; Olmstead, 2024; Timotija, 2024). Nevertheless, Trump, Vance, and GOP supporters have maintained the public lie, with Vance admitting to CNN’s Dana Bash that he and Trump are knowingly spreading falsehoods about Haitian immigrants to raise awareness about real issues (Maher and Boyett, 2024; Olmstead, 2024). Consequently, narratives of Haitians abducting and eating pets in Springfield, OH can be considered political disinformation—deliberate deceit intended to agitate anti-immigrant animus for political gain.
This data was collected as part of an ongoing research project in the spread of political disinformation on YouTube.
(1.1) Limitations and Caveats
This corpus is not meant to be a representative sample of mainstream news coverage of the Haitian disinformation campaign. Videos with less than 500 comments were ignored. No consideration was made for the type of news format contained in the video (for instance, on the scene reporting versus interview), national versus local reporting agency, or replication in the corpus. By extension, the results of the Term Frequency Inverse Document Frequency (TFIDF) analysis used in this study should only be taken as relevant to the comments in the corpus, not the mainstream media or comment landscape on YouTube as a whole.
(2) Dataset description
Repository location
Repository name
Zenodo
Object name
YouTube Disinformation Comments Corpus
Format names and versions
CSV
Creation dates
2024-09-13–2024-11-14
Dataset creators
Ryan M. Omizo (Temple University)
Language
English
License
Creative Commons Attribution 4.0 International
Publication date
2024-11-14
The YouTube Disinformation Comments Corpus captures 71,025 comments on and metadata from 23 YouTube videos posted by the official YouTube accounts of ABC News, CBS News, CNN, Fox News, MSNBC, New York Post, and Forbes news outlets that reported on the debunking of J.D. Vance’s claim that Haitian immigrants were eating neighborhood pets in Springfield, Ohio soon after the first US presidential debate between Kamala Harris and Donald Trump. The corpus contains three files. First, the youtube_haitian_disinformation_videos_meta.csv file includes comments and YouTube video metadata. Data are organized around per video information. The columnar values are:
video_id
date (video publication date)
title
description
channel_title
transcript
transcript_str (Video transcript without timestamps)
views
likes
comments (all comments per video)
Second, the youtube_haitian_disinformation_comment_reply_metadata.csv file includes comments, replies, and comment metadata. Each comment occupies its own row in the spreadsheet. The columnar fields are:
video_id
comment
comment_date
comment_like_count
author
comment_id
in_reply_to
neg, neu, pos, compound (VADER polarity scores)
named_entities (spaCy named entity tags with tokens)
emoji (spaCy Emojis and token spans)
Comments and associated metadata are represented in individual rows.
The third file, youtube_disinfo_comments_tfidf_results_per_video.csv, contains the results of the TFIDF analysis described herein. The TFIDF analysis features the top 5000 terms weights for the comments to each video in a .csv file. For space considerations, I only display the top-10 results of the TFIDF analysis in this paper (Table 1) and direct readers to the complete tabulations on the dataset.
Table 1
Top-10 TFIDF Bigrams for each YouTube Video in the YouTube Disinformation Corpus.
| id ID | TITLE | SOURCE/DATE | EXAMPLE COMMENT | TOP-10 TFIDF TERMS |
|---|---|---|---|---|
| 0VSV99PXbbw | Why Trump keeps pushing debunked claim about Haitian immigrants | CNN, 2024b, Sep 14 | Hunter Biden Laptop 2.0 Russia Collusion 2.0 Its all pre-election spin to manipulate voters, and then sometime in January the NYT will put out an article saying that the illegals are in fact eating local wildlife and pets. (@Zio-e5t, 2024, Sep 16) | (‘fake news’, 0.47), (‘elon musk’, 0.09), (‘laura loomer’, 0.08), (‘cnn joke’, 0.08), (‘russia russia’, 0.08),(‘proven true’, 0.07), (‘neo nazi’, 0.06), (‘white supremacist’, 0.05), (‘comment section’, 0.05), (‘blood libel’, 0.04), (‘clown news’, 0.04), (‘cnn communist’, 0.04), (‘news channel’, 0.04), (‘bon appetit’, 0.04), (‘freedom speech’, 0.03) |
| 39voRIZciVI | Springfield mayor responds to JD Vance’s claims about Haitian immigrants | CNN, 2024e, Sep 15 | Another corrupt worthless mayor that is lying out his asssss (@gregc2069, 2024, Oct 17) | (‘dana bash’, 0.33), (‘fake news’, 0.21), (‘rob rue’, 0.13), (‘mayor democrat’, 0.09), (‘bright light’, 0.08), (‘god bless’, 0.06), (‘liar liar’, 0.06), (‘lie lie’, 0.05), (‘free speech’, 0.05), (‘bon appetit’, 0.05) |
| 4IGd0BrrXoI | Trump, Harris face off at the ABC News Presidential Debate | Fox News, 2024a, Sep 10 | Blame Game the Great Narcissist has spoken! The Great Twist & Shout has FAILED ! Our eyes have been opened across the country to see how much America  needs President Trump ! Jesus Christ has preserved Trump for a time such as this! (@sandraschwenn414, 2024, Oct 17) | (‘abc news’, 0.17), (‘taylor swift’, 0.14), (‘david muir’, 0.12), (‘god bless’, 0.1), (‘blah blah’, 0.1), (‘fake news’, 0.09), (‘word salad’, 0.08), (‘lie lie’, 0.08), (‘old man’, 0.08), (‘good job’, 0.07) |
| 6 W0 mzgcDNv | JD Vance claims on Ohio’s Haitian immigrants eating pets debunked | CBS News, 2024a Sep 10 | Ah yes the same people who said they wouldn’t fact check during the debate and then proceeded to do just that, and lie. Why should Americans trust you? (@boodledemic6430, 2024, Oct 16) | (‘fake news’, 0.36), (‘american woman’, 0.34), (‘drug addict’, 0.08), (‘matter fact’, 0.06), (‘russia russia’, 0.06), (‘true watch’, 0.05), (‘front lawn’, 0.05), (‘lie saw’, 0.05), (‘true though’, 0.05), (‘show proof’, 0.04) |
| AuItvjZ1Tro | Missing cat at center of Springfield conspiracy theory found hiding in basement | CNN, 2024f, Sep 20 | What about the other cats that have gone missing. What about the cell phone footage that showed the man grabbing the cat (@JamesBrian-r2b, 2024, Sep 28) | (‘van lathan’, 0.18), (‘fear fear’, 0.15), (‘one cat’, 0.12), (‘neo nazi’, 0.1), (‘fake news’, 0.08), (‘hate crime’, 0.08), (‘white woman’, 0.06), (‘lost cat’, 0.06), (‘shame shame’, 0.06), (‘bottom line’, 0.06) |
| bPbVUxQ7btE | Residents in Springfield, Ohio react to accusations of immigrants abducting and eating pets | WBNS 10TV, 2024, Sep 12 | But ppl still think it’s fake (jessica-db4is, 2024, Sep 18) | (‘go home’, 0.33), (‘back haiti’, 0.18), (‘blah blah’, 0.1), (‘wrong wrong’, 0.07), (‘voo doo’, 0.06), (‘teach children’, 0.05), (‘might eat’, 0.05), (‘born america’, 0.05), (‘learn drive’, 0.05), (‘back came’, 0.04) |
| DPCXxNXraKw | ‘NO DOGS OR CATS’ Haitian immigrant pleads with Trump and Springfield, Ohio residents | LiveNow from FOX, 2024, Sep 18 | Of coarse they deny it. (@saulovernow, 2024, Oct 12)’ | (‘david muir’, 0.44), (‘know haitian’, 0.22), (‘fake news’, 0.22), (‘last night’, 0.12), (‘yeah right’, 0.07), (‘abc news’, 0.06), (‘bird flu’, 0.06), (‘said fake’, 0.06), (‘game warden’, 0.06), (‘make fake’, 0.05) |
| eytlX2OTKPs | Trump falsely claims Haitian migrants are eating pets in Ohio | CBS News, 2024b, Sep 10 | Haitians commonly eat cats in Haiti and can even buy them at the store URL (@Lar-invxib, 2024, Oct 9) | (‘deport deport’, 0.31), (‘send back’, 0.14), (‘fake news’, 0.13), (‘god bless’, 0.1), (‘vote red’, 0.09), (‘follow money’, 0.07), (‘wake america’, 0.06), (‘close border’, 0.06), (‘fight fight’, 0.05), (‘martha vineyard’, 0.05) |
| flRF5OQrURI | Trump amplifies false claims about Haitian immigrants eating pets | ABC News, 2024b, Sep 19 | Isnt it more possible that african immigrants are eating the pets since they cant stop eating bush meat and have been flooding the boarder too? (@bannedintheusa5714, 2024, Sep 25) | (‘fake news’, 0.11), (‘abc news’, 0.1), (‘lie lie’, 0.09), (‘last woman’, 0.04), (‘blah blah’, 0.04), (‘go home’, 0.04), (‘true right’, 0.04), (‘free speech’, 0.04), (‘take video’, 0.03), (‘david muir’, 0.03) |
| GTmdOGut4Gw | Springfield, Ohio, man reported Haitian migrants snatching geese out of park just 2 weeks ago | New York Post, 2024, Sep 11 | how the guy know they are Haitians? they can be local residents too… at least they are not dogs and cats (@AnythingJW, 2024, Sep 26) | (‘fake news’, 0.13), (‘springfield sue’, 0.1), (‘still support’, 0.09), (‘russia russia’, 0.09), (‘hate crime’, 0.07), (‘big fat’, 0.06), (‘dumb dumber’, 0.06), (‘could vote’, 0.06), (‘shame shame’, 0.06), (‘send bill’, 0.06) |
| i0IvIekwz3c | Criminal charges filed by Haitian nonprofit against Trump and JD Vance | WSYX ABC 6, 2024a, Sep 11 | These idiots are going to wind up proving that they in fact did eat geese and cats  but of course by that time the election will be over (@uarbor70, 2024, Sep 29) | (‘send bill’, 0.15), (‘hate crime’, 0.14), (‘fake news’, 0.12), (‘god bless’, 0.09), (‘make pay’, 0.09), (‘springfield sue’, 0.06), (‘roy cohn’, 0.06), (‘drink bleach’, 0.06), (‘full stop’, 0.06), (‘old man’, 0.06) |
| L5Gq1Tc63UE | Springfield resident describes impact of Haitian migrants on community | Fox News, 2024d, Sep 12 | Unbelievable!! I went to Haiti when I was younger and I saw unimaginable things such as toileting in the streets and eating animals! I saw it! It is true! (@laurabellew9040, 2024, Sep 27) | (‘fake news’, 0.44), (‘cnn stop’, 0.07), (‘hey cnn’, 0.06), (‘cnn still’, 0.06), (‘maga moron’, 0.05), (‘lol cnn’, 0.05), (‘usa usa’, 0.05), (‘see cnn’, 0.05), (‘word salad’, 0.05), (‘cnn propaganda’, 0.04) |
| L9FiWz07t2o | Springfield neighbors describe immigrant situation; say they believe pet-snatching story | WSYX ABC News, 2024a, Sep 11 | ‘Dont fuck with cats and dogs. People will find you just like “ Dont fuck with cats” on netflix lol (@arvinstz, 2024, Oct 14)’ | (‘get first’, 0.08), (‘mean right’, 0.08), (‘spoke english’, 0.08), (‘dear springfield’, 0.08), (‘still thing’, 0.08), (‘bring home’, 0.08), (‘dei hire’, 0.08), (‘bad move’, 0.08), (‘logansport indiana’, 0.08), (‘vote evil’, 0.08) |
| LKauI2Gtj4g | JD Vance exposed for knowingly lying about ‘pet-eating’ rumors in Springfield, OH | MSNBC, 2024a, Sep 18 | Vance is a puppet. (@samlott99, 2024, Sep 23) | (‘mlk fought’, 0.1), (‘mlk fight’, 0.09), (‘fake news’, 0.08), (‘go home’, 0.07), (‘random guy’, 0.04), (‘enough enough’, 0.04), (‘send back’, 0.04), (‘back haiti’, 0.03), (‘much easier’, 0.03), (‘got paid’, 0.03) |
| M547oAXe3O4 | ‘You think this is the last cup of B.S.?’: Steele slams Trump campaign for lying about migrants | MSNBC, 2024b, Sep 18 | MAGA REPUBLICAN GRIFTER AND DONALD TRUMP, JD VANCE ARE BOTH ARE DICTATOR RULE UNITED STATES OF TRUMPIZE (@stevenseabrook8003, 2024, Sep 25) | (‘fake news’, 0.47), (‘ohio true’, 0.08), (‘told truth’, 0.07), (‘comment section’, 0.06), (‘facebook hoax’, 0.05), (‘trust bro’, 0.05), (‘abc news’, 0.04), (‘video true’, 0.04), (‘video fact’, 0.04), (‘covid came’, 0.04) |
| mr5Bv-evY9A | Reporter Presses JD Vance About ‘Debunked False Claims’ About Haitians In Springfield, Ohio | Forbes Breaking News, 2024, Sep 23 | Hes a joke! (@sonyahutchins6197, 2024, Oct 13) | (‘fake news’, 0.44), (‘abc news’, 0.34), (‘news abc’, 0.13), (‘freedom speech’, 0.08), (‘fake fake’, 0.07), (‘take note’, 0.06), (‘anti propaganda’, 0.05), (‘listen abc’, 0.05), (‘money money’, 0.05), (‘maga maga’, 0.05) |
| oBCSCrhnfu0 | JD Vance defends amplifying false claims about immigrants | Face the Nation, 2024, Sep 15 | Looks like Face The Nation is a Democrat owned Network Defending KAREN ! VANCE tries to get his point across by the Democrat Narrator keeps interrupting Vance ! (@davidblanchard5252, 2024, Sep 30) | (‘non profit’, 0.43), (‘good luck’, 0.3), (‘go home’, 0.2), (‘free speech’, 0.18), (‘freedom speech’, 0.12) (‘deport deport’, 0.12), (‘back haiti’, 0.12), (‘great news’, 0.11), (‘sue biden’, 0.1), (‘fake news’, 0.1) |
| oITtLIAYln4 | Ohio residents rage at officials over Haitian migrant surge: ‘No vacancy’ | Fox News, 2024b, Sep 12 | Turns out, it was a 64 year old white guy, ha ha ha ah !!!!!!!! (@davidheaslip4413, 2024, Oct 13) | (‘fake news’, 0.12), (‘great nation’, 0.11), (‘dont want’, 0.1), (‘truth social’, 0.08), (‘proven true’, 0.06),(‘great job’, 0.05), (‘media media’, 0.05), (‘tampon tim’, 0.05), (‘blah blah’, 0.04), (‘good job’, 0.04) |
| ra2PLZlP1o0 | Dana Bash and JD Vance clash over baseless claims about Haitian immigrants (Part 1/2) | CNN, 2024c, Sep 15 | This is a garbage interview. You are a garbage reporter. (@amcurella, 2024, Oct 16) | (‘face nation’, 0.34), (‘fake news’, 0.21), (‘made china’, 0.13), (‘good job’, 0.1), (‘great job’, 0.09) (‘minimum wage’, 0.07), (‘bon appetit’, 0.06), (‘fake media’, 0.06), (‘tim walz’, 0.06), (‘silicon valley’, 0.05) |
| Ru34 gTyHKlw | Trump spreads more lies about Springfield, Ohio | CNN, 2024a, Sep 14 | Not legal and not wanted. They are a burden to American society. We want them deported! (danielkim7841, 2024, Sep 15) | (‘deport deport’, 0.29), (‘send back’, 0.27), (‘vote red’, 0.08), (‘god bless’, 0.08), (‘smirk face’, 0.07), (‘back haiti’, 0.07), (‘save act’, 0.06), (‘close border’, 0.06), (‘fake news’, 0.06), (‘back came’, 0.06) |
| si1Tbz1kbzw | JD Vance reacts to Springfield father saying he politicized his son’s death (Part 2/2) | CNN, 2024d, Sep 15 | This has brought to you by Pfizer (@brettschultz5150, 2024, Oct 9) | (‘dana bash’, 0.56), (‘sperm count’, 0.07), (‘dana got’, 0.07), (‘fake news’, 0.07), (‘jim crow’, 0.05), (‘white supremacist’, 0.05), (‘put place’, 0.05), (‘great job’, 0.05), (‘past eight’, 0.05), (‘good job’, 0.04) |
| TQm_EAmLhJw | Springfield, Ohio resident: This is a dystopian nightmare | Fox News, 2024c, Sep 12 | TRUMP 2024  (@WiseUpRiseUp42, 2024, Oct 15) | (‘dana bash’, 0.29), (‘ever thought’, 0.11), (‘half time’, 0.11), (‘job journalist’, 0.11), (‘love fact’, 0.11), (‘stir hate’, 0.11), (‘total bullshit’, 0.11), (‘asylum claim’, 0.11), (‘dana got’, 0.1), (‘mop floor’, 0.1) |
| xc0IkVmiN8o | Haitian community in Springfield, Ohio, under threat after debunked comments | ABC News, 2024a, Sep 16 | ‘ABC is lLYING!! I get real true news from the local people-NOT the dishonest lying bought and paid for fake news media. (@RonaldMacDonald-hd8 gh, 2024, Sep 23) | (‘fake news’, 0.34), (‘abc news’, 0.3), (‘shame abc’, 0.1), (‘back haiti’, 0.09), (‘still job’, 0.09), (‘abc joke’, 0.07), (‘governor said’, 0.06), (‘abc lie’, 0.06), (‘far left’, 0.06), (‘news channel’, 0.06) |
(3) Method
Video metadata and comments were downloaded using YouTube’s Academic API between September 10, 2024 and October 2, 2024. Videos were selected if they were uploaded by a major news outlet (such as ABC or CNN) and featured comment counts greater than 500. Metadata for the corpus features the YouTube video id, publication date, video description, transcripts with and without timestamps, view count, like counts, comment counts, and YouTube tags. Comments, replies to comments, and comment metadata were also collected using the YouTube Academic API. Comment and reply metadata include the date of publication, like count, comment author name, comment id, and the account name of the comment receiving replies.
To enrich the data, I extracted named entities and emojis from comments using spaCy (Honnibal et al., 2020). I also computed comment sentiment polarity using VADER (Hutto and Gilbert, 2014). I include negative (neg), neutral (neu), positive (pos), and compound scores for each sample.
To describe the comments in the corpus, I employ TFIDF analysis to surface the most definitive words in comments to each video (see Gallagher et al, 2023a; Gallagher et al., 2023b). TFIDF is a well-established method in information retrieval, which represents documents as a series of term weights. Weights for terms are a function of a term’s frequency in a document compared to the frequency of the term across the entire corpus. Terms common across the corpus are downweighted. Terms that are specific to a document are elevated (Manning and Schutze, 1999). For this study, I vectorize the comments and label each comment by its associated video_id. I use the video_ids as class labels and calculate the most informative features for each class using the chi-square test of significance. Key terms such as “harris,” “trump,” “donald,” “vance,” “j.d.,” “kamala,” and “president” were removed because they refer to content already manifest in the general description of videos.
(4) Results and Discussion
I present the top-10 ranked terms for each YouTube video in Table 1 below as an overview of comments. To contextualize the top bigrams/trigrams, I present example comments for select video in which the bigram appears. Readers are encouraged to peruse the data repository to view additional comments, video data citations and more extensive TIFDF information.
(5) Implications/Applications
The YouTube video and comment data provide several opportunities for reuse in the study of (1) natural language processing, (2) network text analysis, (3) political disinformation on social media, (4) disinformation/misinformation on YouTube, (5) news framing effects, and (6) public discussions of immigration.
(5.1) Natural Language Processing
As suggested by the TFIDF analysis above, the corpus is suitable for natural language processing. The comments can inform topic modelling research or demonstrations. SpaCy named entity information can inform fine-grained linguistic analysis of named agents in the corpus. Emoji tags can be used to assess what and how often YouTube commenters employ emojis in their arguments (Gray & Holmes, 2020). VADER scores can be used to assess the sentiment of comments. Engagement data provided by like and comments counts, and date of publication information can also scaffold time series analyses and novel explorations into social media metrics.
(5.2) Network Text Analysis
Reply data captured in the in_reply_to and authorship fields of youtube_haitian_disinformation_comment_reply_metadata.csv can be used for network modelling, uncovering relationships between comments and between commenters. Such modelling can reveal what types of comments facilitate the most conversation by measuring the in-degree and out-degree of replies.
(5.3) Political Disinformation on Social Media
For researchers interested in quantitative and qualitative approaches to disinformation on social media, transcript, video description, video title, and comment data are amenable to manual, qualitative coding and automated approaches such as sentiment analysis (Han, 2022) or topic modelling (Latorre & Amores, 2021; Shekhar, et al., 2021).
(5.4) Disinformation/Misinformation on YouTube
The corpus can underpin research into YouTube-specific forms of political disinformation, political polarization (Munger & Phillips, 2022), and racial/xenophobic agitation akin to the work of Inwood and Zappavigna (2023a; 2023b), who view YouTube as a platform conducive to the propagation of conspiracy theories and describe YouTube comments as critical for examining multiple discursive activities (see also Dynel, 2014; Heydari et al., 2019; Cinelli et al., 2021).
(5.5) News Framing Effects
Relationships between textual data categories can be mined. For example, textual similarity between YouTube transcript data and attendant comments could be calculated to determine how video content is being recontextualized by YouTube audiences (Boyd, 2014). From another perspective, the YouTube video metadata included with the comments can be studied for their “news framing effects” (Gamson & Modigliani, 1989; Entman, 1993). Theories of news framing propose that news outlets guide audience reception of journalistic content by emphasizing, de-emphasizing, and/or contracting elements of the reporting (de Vreese, 2005, p. 52; Lecheler et al., 2015, p. 814; Lecheler & de Vreese, 2012, p. 186). This orchestration of priorities is achieved through textual, visual, or audio “framing devices” (Gamson & Modigliani, 1989; see also de Vreese, 2005, p. 54). Such devices include news headlines, charts, pull quotes, and corporate branding (de Vreese, 2005, p. 54). Within the corpus, tags, transcripts, video descriptions, and title information function as comparable devices and can inform traditional news framing analysis. The effects of these news framing devices can then be assessed with the comment data through qualitative or quantitative methods. For research into the recognized news frame of “immigration” (Eberl & Meltzer, 2021; Quinsaat, 2014; Eberl et al., 2018; Kim & Want, 2018; Yang, 2020; Guo et al., 2023; Reedy et al., 2023), the corpus provides opportunities to trace the use of immigration framing in media in the context of a historic national election.
(5.6) Public Discussions of Immigration
The corpus can inform research into public discussions of immigration in the era of social media and Trumpism by building upon existing studies of video content as Lybecker et al. (2015) and Laaksonen et al. (2020) do. Following Lybecker et al. (2015), an analysis on the construction of Haitian immigrants through video content could be undertaken using video transcript data. Incorporating comment data would represent an extension of Lybecker et al.’s (2015) analysis of depictions of the U.S.-Mexican Border, which uses video content as its data. Similarly, the qualitative discursive analysis of video content by Laaksonen et al. (2020) to reveal major patterns of movement-building argumentation in Finnish anti-immigration discussions could be applied to the corpus. Qualitative and/or quantitative analysis of comments could then be used to inspect direct reactions to the anti- or pro-immigrant stances.
Funding Information
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
Competing Interests
The author has no competing interests to declare.
Author Contributions
Conceptualization
Data curation
Formal Analysis
Investigation
Methodology
Project administration
Resources
Software
Visualization
Writing – original draft
Writing – review & editing