The Midterm Monitor is an interactive tool designed by the Alliance for Securing Democracy (ASD) at the German Marshall Fund of the United States and the Brennan Center for Justice at NYU Law. It allows users to follow and analyze conversations about the 2022 midterm elections among key stakeholders on social media.

This page covers the methodology behind the monitor, explains its capabilities, and provides a curated list of frequently asked questions.

During the 2000 election and since, voting and election administration have become an increasing part of the national conversation, and the narratives in these discussions have had a direct impact on our democracy. The Midterm Monitor captures messages about voting and elections and allows users to analyze them and track their development.

The monitor captures social media posts by candidates for office, news outlets, and other key messengers across multiple platforms (Twitter, YouTube, Facebook, and Instagram). It defaults to showing users posts and metrics on voting and election-related content from these accounts, capturing data via election-related search terms such as “ballots,” “drop boxes,” “voting machines,” and more to help users put together a fuller picture of these actors’ election messaging and how it changes over time.

The goal of the monitor is to provide academics, election officials, journalists, researchers, and other interested individuals the opportunity to dive deeper into election narratives by these actors and to filter the narratives based on geography, source, platform, language, and more. This data can help users better understand and respond to election-related narratives in the leadup to the 2022 midterms.

Because the tool uses election-related terms to surface social media posts, the results might produce some false positives. Users should not draw conclusions from the raw number of posts that appear in each category.

The dashboard does not screen links for malware, phishing content, spam, or any other undesired content. Users should click on links at their own risk.

The Midterm Monitor tracks posts from a subset of three categories of accounts: candidates for US political offices, US media sources in English and Spanish, and foreign actors. A description of each category is below.

Click here to download the complete list of accounts monitored on the dashboard.

Candidates

The monitor tracks social media posts from candidates for the US House and US Senate nationwide, and candidates in races for governor and secretary of state.

Candidates are added to the monitor at the completion of each state’s primary elections. Only candidates who win their respective party’s nomination are included. On platforms where candidates have a personal and official campaign account, only the official candidate/campaign account is included.

Candidates’ social media accounts are identified using a list provided by Ballotpedia and are verified by ASD. ASD routinely audits accounts to identify new, altered, or deleted accounts. If an account cannot be verified as authentic by ASD, it is not included. On Facebook and Instagram, if a candidate’s profile/account is a personal profile or “gated,” it cannot be tracked due to CrowdTangle limitations. CrowdTangle is the technology platform that feeds the content from monitored Facebook and Instagram accounts into the tool.

Users can filter candidate accounts by state, party (Democrat, Republican, Independent, and other), and position (governor, secretary of state, US House, and US Senate). Users can also select individual accounts by entering the candidate’s name into the search field.

Media

The Midterm Monitor tracks posts from accounts affiliated with influential national and local media outlets, programs, and individual media personalities. It includes both English and non-English language entities.

The local media filter pulls from social media accounts associated with news outlets in 10 battleground states. The national media filter pulls from national media outlets, as well as select state and local outlets in states that are not categorized as battleground states.

Local Media Accounts

In total, the monitor collects data on more than 700 individual social media accounts across the various platforms tied to roughly 260 local media entities.

The local media filter includes only social media accounts associated with local newspapers, news websites, radio, and television broadcast news outlets in 10 battleground states (Arizona, Georgia, Florida, Michigan, Nevada, New Hampshire, North Carolina, Pennsylvania, Texas, and Wisconsin). All other local or state media accounts in the United States outside of the battleground states are classified as national accounts (more information below) and are not searchable by state.

Local accounts include media accounts with the largest number of followers on social media platforms across media type (television, radio, broadcast news, online outlets, etc.). In some cases, prominent online outlets (for example, The Texas Tribune) were included. Individual media personalities were not included in the local media list.

National Media Accounts

In total, the monitor collects data on more than 500 individual social media accounts across the various platforms tied to approximately 160 national media entities.

National media accounts were selected based on several independent rankings used to identify the most influential outlets and pundits on a range of social media platforms and other broadcast mediums. To the extent possible, we identified the top 20 most followed/subscribed to media accounts on Facebook, Instagram, Rumble, Telegram, Twitter, and YouTube. We also selected social media accounts for the top media outlets in the following categories: broadcast news networks and programs (as determined by Nielsen ratings), news websites (as determined by site visits according to the UK-based Press Gazette), news podcasts (as determined by Apple and Spotify downloads), newspapers and magazines (as determined by circulation according to the UK-based Press Gazette), and news talk radio programs (as determined by rankings from Talk Media Inc.).

If an outlet, program, or pundit appeared in the top 20 of any of the reviewed lists, affiliated social media accounts on all monitored platforms were added to the monitor. For example, if a cable news show was among the 20 most viewed news broadcasts, all its social media accounts were included, regardless of whether those accounts ranked among the top 20 accounts on any given platform. Or if a pundit was one of the top 20 most-followed accounts on Rumble, then the accounts affiliated with that individual on the monitored platforms—Twitter, YouTube, Facebook, and Instagram—were included in the Midterm Monitor. In most cases there was significant overlap between individuals and entities identified as influential by independent metrics and the most followed individuals and entities on social media channels.

In assessing the most-followed media on social media platforms, outlets, shows/programs, and individual pundits were assessed collectively. This means the most-followed media accounts on each platform include a mix of accounts associated with individuals and accounts associated with media entities. In cases where there was both an account for a prominent show host and an account for the show itself, only the account affiliated with the show was included. For example, the account for The Lead with Jake Tapper is monitored, but Jake Tapper’s personal account is not.

The national media category also captures posts from social media accounts associated with the top major newspaper in every state, with the exception of outlets from the 10 monitored battleground states, which are included under the local media category as explained above.

Spanish and Other Non-English Language Media Accounts

Non-English language accounts are included in both the national and local media lists. Accounts were selected using the same criteria as those used to identify English-language national and local media accounts.

In total, the monitor collects data on more than 50 non-English language entities and more than 150 social media accounts.

NOTE: Language classification is determined by the typical language used by an account—not the language used in an individual post. That means that if a Spanish-language outlet posts in English, that post will still be classified as Spanish on the dashboard.

Foreign Accounts

The monitor collects data from social media accounts affiliated with Chinese, Iranian, and Russian government officials and state-funded media. Account lists were sourced from ASD’s Hamilton 2.0 dashboard.

Government Accounts

This category contains accounts of key individuals and institutions that represent the governments of China, Iran, and Russia, including:

  • All known embassy, consular, ambassadorial, and consul general accounts, and accounts of other prominent diplomats (as determined by more than 5,000 Twitter followers or a label applied by Twitter or Facebook designating the account as a government official)
  • Official accounts representing the Ministry of Foreign Affairs, as well as agencies and organizations that have an international or foreign policy focus (e.g., the Russian Mission to NATO)
  • Key individuals within the Chinese, Iranian, and Russian governments

State-Funded Media

This category contains accounts affiliated with China, Iran, and Russia’s various state-funded media outlets, including:

  • Accounts connected to China, Iran, or Russia’s state-funded international broadcasters and online news and information portals
  • Accounts connected to television programs on state-backed media outlets
  • Accounts of key management personnel and prominent journalists, show hosts, and pundits (as determined by more than 5,000 Twitter followers or a label applied by Twitter or Facebook designating the account as a state media account)

Because political conversations about the United States are not limited to government officials or state media outlets that are based in or exclusively target audiences in the United States, foreign actor accounts were included regardless of their location or primary language. As a result, there may be a significant amount of non-relevant content in this section, including posts categorized as “election-related” that reference elections in other countries.

WARNING: These accounts often engage with hashtags, people, topics, and URLs that are in no way affiliated with the Chinese, Iranian, or Russian governments. It would be incorrect to, without further analysis, label anyone or anything that appears on the monitor as being connected to state-backed propaganda.

The Midterm Monitor currently collects data from publicly available accounts on Twitter, YouTube, Facebook, and Instagram. ASD and the Brennan Center expect to add more platforms in the future.

Twitter and YouTube data is collected via public application programming interfaces (API). On YouTube, data is extracted from video titles and descriptions on monitored channels, not from transcripts of the videos themselves. Engagement metrics (retweets/likes/views/etc.) are updated for three days after the date of the original post, meaning displayed metrics on older posts may be slightly lower than the current numbers. For the most up-to-date engagement metrics, users should refer to the original post on the platform on which it was posted.

Facebook and Instagram posts and metrics are sourced from publicly available pages on CrowdTangle.

Non-English language posts are displayed in English using Microsoft’s Translator API.

New posts are added to the dashboard daily. Account-level metrics (followers, subscribers, etc.) are also updated daily. Engagement metrics for posts (likes, shares, retweets, and comments) are updated for at least three days after the original post. Historical data is accessible from the date data collection began on an individual account.

Data has been collected on all monitored accounts on Twitter, YouTube, Facebook, and Instagram since at least June 1, 2022, except for accounts affiliated with candidates whose primaries occurred after August 1. The Midterm Monitor also has available historical Facebook and Instagram posts and metrics for all monitored pages and accounts starting in July 2021. Historical posts and metrics on YouTube videos are available from as early as 2020. The start date for data collection on individual Twitter accounts is dependent on the aforementioned criteria, as well as on the ability to backfill historical data, which is based on Twitter’s API rate limits and the volume of posts made by a specific account. As a result, available data predates when the account was added to the monitor. Please contact ASD at info@securingdemocracy.org for more information or questions about specific accounts.

The election-related posts toggle switch is an automated filter that helps identify posts from monitored accounts that contain at least one word or phrase related to voting or elections, with a particular focus on election administration and election security.

Election-related posts are searchable by sub-categories of narratives that are relevant to election security and election administration. The “general election” filter is the broadest, including terms like “voter,” “election,” and “midterms.” As a result, like all automated functions, this category may include posts about unrelated or tangentially related topics—including posts about elections outside the United States.

The six other categories that users can filter results by are listed below and focus on specific elements of election administration. Each category captures posts with key terms (examples also below) deemed to be relevant to the category. The term list was created by researchers based on extensive background research. ASD and Brennan Center researchers looked to previous studies on this topic and sought feedback from partners on the ground and election administration experts.

Click here to download the complete list of election-related key terms.

  • Ballots (ballot; ballots)
  • Election security (software glitch; rigged count; hacked machine)
  • Registration (non-citizens voting; dead registered)
  • Vote by mail (drop boxes; absentee voting)
  • Vote counting (recount; votes flipped; decertified)
  • Voting in person (lines polls; voter ID; poll workers)

If a post includes key terms from multiple categories, the post will appear under each related categorical header. Given the interconnected nature of many of the narrative categories, posts that appear in one category may in fact be more pertinent to one of the other categories. Because certain terms are more likely to produce true or false positives than others, users should not draw conclusions from the raw number of posts that appear in each category.

The monitor’s default setting is to display posts from all the election-related narrative categories over the selected date range. When this filter is applied, the toggle switch will be in the green “on” position. Users can select one or more of the narrative categories using the dropdown feature. To see all posts from monitored accounts (posts that may or may not contain an election keyword), flip the election-related posts toggle switch to the gray “off” position.

The Brennan Center created the following list of false or misleading election-related claims to assist analysts with narrative coding:

  1. The 2020 election was fraudulent and/or stolen.
  2. Donald Trump actually won the 2020 election.
  3. Donald Trump is so popular, he couldn’t have lost the 2020 election.
  4. Everyone knows that something fishy happened with the 2020 election. We feel it in our gut.
  5. Joe Biden winning the 2020 election was statistically impossible / improbable.
  6. There were hidden suitcases full of ballots in the 2020 election, which is evidence of widespread voter fraud.
  7. Biden supporters engaged in ballot tampering in the 2020 election.
  8. Massive ballot harvesting took place, which is evidence of voter fraud.
  9. Widespread voter fraud is a problem.
  10. There were more votes than voters.
  11. There were more mail-in ballots received than ballots that were sent out.
  12. Truckloads of ballots were sent from one state to another.
  13. There were widespread, suspicious voting irregularities in the 2020 election.
  14. There were shredded ballots found following the 2020 election.
  15. Thousands of mail in ballots were illegal in Wisconsin because they used illegal dropboxes.
  16. Late ballots were backdated so they could be counted illegally.
  17. Dominion voting system switched votes from Trump to Biden.
  18. Dominion voting systems are part of a broad conspiracy to overthrow the U.S. election.
  19. Venezuela and/or Hugo Chavez is part of a broad conspiracy to overthrow the U.S. election.
  20. Noncitizens  (“illegals” or “illegal aliens”) were voting in Arizona or elsewhere in the 2020 election.
  21. Some ballots were flown in from China to tip the scales in favor of Biden. These ballots could be identified because they had traces of bamboo.
  22. President Biden invited TikTok to be part of the Federal Voting Assistance Program to allow China to influence U.S. elections.
  23. The votes of dead people were counted in the 2020 election.
  24. The 2020 election can be decertified.
  25. The midterm elections are likely to be fraudulent.
  26. Election officials are handing out the wrong type of pen to void voter ballots.
  27. Voters should steal the felt tip pens provided at voting sites, and replace them with ballpoint pens instead.
  28. Voting officials are handing out the wrong pens intentionally to invalidate your vote. Voters should bring their own (ballpoint or other) pens instead.
  29. People will get Covid-19 if they vote.
  30. Covid-19 is a hoax designed to swing the election in favor of Democrats.
  31. Monkeypox is a hoax designed to swing the election in favor of Democrats.
  32. USPS will destroy ballots to interfere in the election.
  33. You can skip the lines and vote via text or online platform from home.
  34. Nursing home residents are having their votes stolen by the staff.
  35. Poll workers provided specific writing instruments, such as Sharpies, to specific voters to cause their ballots to be miscounted.
  36. There were “mules” caught stuffing ballot boxes on camera, which is evidence of widespread voter fraud.
  37. The use of any of the following hashtags: #stopthesteal #2000Mules #TTV #TrueTheVote

The homepage provides a snapshot overview of available data on all monitored platforms.

Candidate and Media Influence by State

This visual displays voting- and election-related posts made by accounts affiliated with individual states, for example, local media accounts where applicable (in 10 battleground states) and candidates affiliated with all states. Hovering over a state shows the number of posts made by relevant accounts on monitored platforms over the user-selected date range. Clicking on a state provides a more detailed view of posts based on account type and platform, and updates data in the table. Scroll down lower on the page to see the top social media posts and accounts from a selected state.

NOTE: Ten states were selected as battleground states (Arizona, Georgia, Florida, Michigan, Nevada, New Hampshire, North Carolina, Pennsylvania, Texas, and Wisconsin) and are shaded purple. The monitor includes additional local media tracking in those states. Total posts from battleground states will therefore be higher; the data about media outlets’ social media posts is not comparable from state to state.

Posts by Platform and Account Type

This visual displays the number of election-related posts made by account types over a user-defined period (the default is the last seven days). Clicking an account type (for example, Republican Senate candidates) will filter the map to display the states with the most relevant posts made by affiliated accounts. Scroll down lower on the page to see the top social media posts and accounts by account type.

NOTE: If users want to apply multiple filters when engaging with the “candidate and media influence by state” map or the “posts by platform and account type” table, the user must hold down the CTRL function on a PC or the cmd function on a Mac while making their selections. For example, if a user wants to see posts from local media accounts in Georgia, the user needs to click “local” and then press CTRL or cmd before clicking Georgia (or vice versa).

Top Social Media Posts

These charts show the top three election-related posts by platform (as determined by retweets, reactions, views, and likes). User-selected filters on other parts of the page (for example, selecting an individual state from the map or a specific election narrative category) will update the relevant posts displayed in this chart.

Top Social Media Accounts

These charts show the top three accounts by platform (as determined by the number of retweets, reactions, views, and likes made on election-related posts). User-selected filters on other parts of the page (for example, selecting an individual state from the map or a specific election narrative category) will update the relevant posts displayed in this chart.

Overview

The Twitter page displays key metrics for tweets made by monitored public accounts. The page defaults to show metrics for election-related tweets made by monitored candidates over the past seven days. Users can change the date range, select specific election narratives (or flip the election-related posts toggle switch off to view all posts), or select a different category of accounts (foreign actors or media).

Top Tweets by Retweets and Top Tweets by Likes

These charts display the most retweeted and liked tweets made by monitored accounts during a user-specified date range and with user-selected filters. Engagement metrics are updated for three days after the date of the original post. Clicking on a specific tweet will take the user to Twitter to view the original post. We urge users to check the original post on Twitter for the most up-to-date engagement metrics.

Top Mentions

This chart displays the Twitter handles most frequently included in tweets made by monitored accounts during a user-specified date range and with user-selected filters. Mentions include when a monitored account tags a non-monitored account by including its handle in a tweet or when a monitored account replies to a specific tweet from a non-monitored account.

Click on a mentioned account and scroll down to the “matching tweets” chart to see the Tweets from monitored accounts that contain the mentioned account.

Top Countries Referenced

This chart displays the countries that are referenced most often in Tweets made by monitored accounts during a user-specified date range and with user-selected filters. Special territories and regions (for example, Hong Kong and the EU) are included as individual entities in the country data. The chart lists all references of a country, including alternative country names, under that country’s official name (for example, mentions of US, USA, U.S., and America are all included in tabulations for the United States of America).

Click on a country and scroll down to the “matching tweets” chart to see the Tweets that mention a country.

NOTE: Because of false positives due to mentions of the state of Georgia, Congressman Jim Jordan, and the state of New Jersey, the countries of Georgia and Jordan and the self-governing island of Jersey are removed from results.

Top URLs

This chart displays the hyperlinks most shared by monitored accounts in tweets during a user-specified date range and with user-selected filters. Shortened links have been expanded to display the full URL. Links are organized by common domains, but specific URLs can be identified by expanding those domains. The dashboard does not currently filter out self-references (for example, the Fox News account linking to Fox News URLs), so results are heavily influenced by accounts that promote links to their own content.

Click on a URL and scroll down to the “matching tweets” chart to see the tweets that contain the URL.

Again, the dashboard does not screen links for malware, phishing content, spam, or any other undesired content. Users should click on the links at their own risk.

Top Key Phrases

This chart displays the most-used phrases (including proper names, terms, etc.) that appear in tweets made by monitored accounts during a user-specified date range and with user-selected filters.

Click on a key phrase and scroll down to the “matching tweets” chart to see the tweets that contain the key phrase.

NOTE: When election-related posts are selected, the “key phrases” chart will likely show selected terms used to identify election-related posts. The chart has more value when looking at all posts, when looking at individual accounts, or when other filters are applied.

See below to learn how we extract key phrases from tweets.

How are key phrases extracted from Tweets?

We use Microsoft’s Cognitive Services to automatically translate non-English tweets and to extract key phrases. Non-relevant or non-specific key phrases (for example, common nouns, filler words like “subscribe,” etc.) are filtered out to remove noise from the results. We extract key phrases instead of key words to link together related words (for example, “INF Treaty” or “Barack Obama”). It is possible, however, for two or more related nouns to appear as multiple entries in search results. For example, “Trump,” “Donald Trump,” and “President Trump” may appear as three unique phrases.

Top 10 Most Influential Accounts

This chart displays the most retweeted monitored accounts during a user-specified date range and with user-selected filters. Users can also view the most-liked and most-followed accounts.

Click on an account and scroll up on the page to see relevant metrics and top tweets (for example, most retweeted tweets) and down to see additional information (for example, the accounts most-retweeted by a monitored account)

NOTE: The most-followed accounts view does not change when a user selects different election-related filters.

Top 10 Change in Followers

This chart displays the monitored accounts that have gained or lost the most followers during a user-specified date range.

Click on an account and scroll up on the page to see relevant metrics (for example, top tweets) and down to see additional information (for example, the accounts most-retweeted by a monitored account)

NOTE: Follower data is updated daily. The most-followed accounts view does not change when a user selects different election-related filters.

Most Retweeted Accounts

This chart displays accounts that have been retweeted most often by a selected category of monitored accounts (for example, candidates) that are NOT part of that category of accounts. For example, if the user selects the candidate category, the chart will display nonmonitored accounts and monitored media or foreign actor accounts that have been retweeted most often by candidates, but it will not display monitored candidate accounts. This is to minimize self-references—that is, accounts that retweet their own content.

Click on an account and scroll up on the page to find the specific tweets retweeted by monitored accounts.

Top Retweeted Tweets

This chart displays the most-retweeted tweets that were retweeted by one or more monitored accounts. This chart is sorted by total retweets, not just retweets made by monitored accounts. As with the most retweeted accounts section, this chart displays tweets made by nonmonitored accounts and by monitored accounts that are not part of the user-selected category.

Click on a specific tweet and scroll up on the page to the “top tweets by retweets” chart to find the monitored accounts that retweeted that tweet.

A retweet does not necessarily imply that the monitored account doing the retweeting endorses the account they are retweeting or the content of a specific post.

Matching Tweets

This chart shows relevant tweets based on user-selected filters applied on other charts on the page. Click the link button in this chart to view the relevant tweet on Twitter.

Analysis

Trend for Top 5 Hashtags

This chart shows the trendlines for the five most-used hashtags over a user-selected date range and with user-selected filters. Click on a hashtag and scroll down to see related metrics.

Key Phrase Trends

These charts allow users to compare the prevalence of related terms during a user-specified date range and with user-selected filters. Users can search for a term (for example, Democrats) to see the top related key phrases. Users can use the two charts to examine trends related to different terms (for example, Democrats and Republicans).

Top 10 Key Phrases

This chart displays the most-used key phrases (including proper names, terms, etc.) that appear in tweets made by monitored accounts during a user-specified date range and with user-selected filters.

NOTE: When the toggle switch for election-related posts is in the default “on” position, the key phrases chart will likely show selected terms used to identify election-related posts. The chart has more value when looking at all posts, when looking at individual accounts, or when other filters are applied.

Click on a key phrase and scroll up to see related comparative metrics.

Top 10 Hashtags

This chart displays the hashtags used most often in tweets made by monitored accounts during a user-specified date range and with user-selected filters.

Click on a hashtag and scroll up to see related comparative metrics.

The Facebook page displays key metrics for posts made by monitored public pages on Facebook. The page defaults to show metrics for election-related posts made by monitored candidates over the past seven days. Users can change the date range, select specific election narratives (or flip the election-related posts toggle switch off to view all posts), or select a different category of accounts (foreign actors or media).

Top Posts

This chart displays the posts with the most reactions made by monitored pages during a user-specified date range and with user-selected filters. Click on “Post” to view the post on Facebook.

Users can also sort by comments and shares. For the most accurate results, select ALL posts (not the top 10 or top 100) before sorting by comments and shares.

Click on the post and scroll down to see related account information.

Top Pages

This chart displays the pages with the most reactions during a user-specified date range and with user-selected filters. Click on the account “name” to view the page on Facebook.

Users can also sort by followers, comments, shares, and followers percentage changed. For the most accurate results, select ALL posts (not the top 10 or top 100) before sorting by additional metrics.

Click on any page and scroll up to see related posts.

The Instagram page displays key metrics for posts made to monitored public accounts on Instagram. The page defaults to show metrics for election-related posts made by monitored candidates over the past seven days. Users can change the date range, select specific election narratives (or flip the election-related posts toggle switch off to view all posts), or select a different category of accounts (foreign actors or media).

Top Posts

This chart displays the posts with the most likes made by monitored accounts during a user-specified date range and with user-selected filters. Click on the post to view the post on Instagram.

Users can also sort by the number of comments. For the most accurate results, select ALL posts (not the top 10 or top 100) before sorting by comments.

Click on the post and scroll down on the page to see related account information.

NOTE: Messages are truncated after the first 300 characters, but key phrases are extracted from the entire post.

Accounts Summary

This chart displays the monitored accounts that have received the most comments during a user-specified date range and with user-selected filters. Click on the name to view the account on Instagram.

Users can also sort by likes, followers, and number of posts. For the most accurate results, select ALL posts (not the top 10 or top 100) before sorting by additional metrics.

Click on any account and scroll up on the page to see related posts.

The YouTube page displays key metrics for videos posted by monitored YouTube channels. Data is extracted from the video titles and descriptions of videos posted by monitored channels, not from the content or transcriptions of the videos themselves. The page defaults to show metrics for election-related videos made by monitored candidates over the past seven days. Users can change the date range, select specific election narratives (or flip the election-related posts toggle switch off to view all posts), or select a different category of accounts (foreign actors or media).

Top YouTube Videos

This chart displays the most-viewed videos made by monitored channels during a user-specified date range and with user-selected filters. Select any video title to view the video on YouTube.

Users can also sort by likes and comments. For the most accurate results, select ALL videos (not the top 10 or top 100) before sorting by likes or comments.

NOTE: Descriptions are truncated after the first 300 characters. Key phrases are only extracted from the truncated descriptions to minimize the chance that words unrelated to the purpose of the dashboard—like channel descriptions, subscriber links, and more—get included in the dashboard results.

Top Key Phrases

This chart displays the most-used phrases (including proper names, terms, etc.) that appear in video headlines and descriptions made by monitored accounts during a user-specified date range and with user-selected filters.

Click on a key phrase and scroll up to see relevant videos.

NOTE: When the toggle switch for election-related posts is on, the key phrases chart will likely show terms used to identify election-related posts. The chart has more value when looking at all posts, when looking at individual accounts, or when other filters are applied.

How are key phrases extracted from videos?

We use Microsoft’s Cognitive Services to translate non-English videos and to extract key phrases. Non-relevant or non-specific key phrases (for example, verbs, common nouns, etc.) are filtered out to remove noise from the results. We extract key phrases instead of key words to link together related words (for example, “INF Treaty” or “Barack Obama”). It is possible, however, for two or more related nouns to appear as multiple entries in search results. For example, “Trump,” “Donald Trump,” and “President Trump” may appear as three unique phrases.

Top Countries Referenced

This chart displays the countries that are referenced most often in videos made by monitored accounts during a user-specified date range and with user-selected filters. Special territories and regions (for example, Hong Kong and the EU) are included as individual entities in the country data. The chart lists all references of a country, including alternative country names, under that country’s official name (for example, mentions of US, USA, U.S., and America are all included in tabulations for the United States of America).

Click on a country and scroll up to see relevant videos.

NOTE: Because of false positives due to mentions of the state of Georgia, Congressman Jim Jordan, and the state of New Jersey, the countries of Georgia and Jordan and the self-governing island of Jersey are removed from results.

Most Viewed Videos

This chart shows the dates the five most-viewed videos were uploaded to monitored accounts during a user-specified date range and with user-selected filters.

Video Engagement Rate

This chart displays the videos with the highest engagement rate (as determined by total comments per video view) posted by monitored accounts during a user-specified date range and with user-selected filters.

Click on any video and scroll up to see relevant metrics associated with the selected video.

The Midterm Monitor allows users to search for any key term that appears in a tweet made by a monitored account by clicking the Social Data Search button at the top of the site.

Social Data Search is a searchable repository of all tweets (additional platforms may be added in the future) made by monitored Twitter accounts. This tool allows the user to view all tweets that contain a key word, term, or compound phrase of interest.

Click on the arrow on the right side of any tweet to see the original tweet on Twitter.

Users can search for any query within quotes (“”) to search for a specific phrase. For example, a search for “Georgia election” will only show results with the words “Georgia” and “election” back-to-back.

Use an * as a wildcard to search for similar words. For example, a search for elect* will show results that start with “elect,” such as elected, elector, and election. It can be used as a suffix (alpha* matches “alphanumeric” or “alphabetical”) or a prefix (*elect matches “select” and “elect”).

Use a + sign as an AND. For example, a search for voter + turnout will show results with both “voter” and “turnout” in them—in any part of the text.

Use | as an OR. For example, a search for voter | turnout will show results with either “voter” or “turnout” or both in them.

Use – as a NOT. For example, a search for vote –Emmy will show results that include “vote” but do not also include “Emmy.”

Use ~ (and a space after) for a fuzzy search. For example, a search for “blue~” will show results with “blue”, “blues”, and “glue.”

Use the Toggle Additional Filters & Legend button to view related metrics (for example, most mentioned hashtags in searches for “fraud”) and also to add additional filters to a search.

NOTE: The Midterm Monitor and Social Data Search may yield slightly different results for the same search phrase due to (1) how search terms are processed by the search algorithm; (2) Social Data Search includes usernames, hashtags, countries, and other text fields in its searchable text; and (3) different data update schedules. Generally, these differences should be minor. External links may not work when either a tweet or an account has been removed. Tweets have been translated from their original language to English; the original text may be viewed by clicking the relevant tweet to show additional information.

We do not allow users to export data directly from the Midterm Monitor or Social Data Search. Researchers can submit requests for data to the Alliance for Securing Democracy at info@securingdemocracy.org.