Identifying The Source And Description Of A Television News COVID-19 Clip On CNN

American television news footage of COVID-19 frequently shows the pandemic's impact around the world, but doesn't always provide sufficient context for viewers to understand what a given clip depicts. Take the August 11, 2020 CNN Erin Burnett Out Front episode during which a brief clip is played of a patient in an unidentified hospital ward. CNN's broadcast doesn't provide any detail about what the clip actually depicts other than a credit to "BAZA | Telegram". The clip appears during a segment on the Russian response to COVID-19 so one might assume it was taken somewhere in Russia, but there is simply no further detail available to viewers.

Extracting a frame from the clip and removing the chyron and sidebars we arrive at the following image:

We pass this to Google's Cloud Vision Web Entities reverse image search:

curl -s -H "Content-Type: application/json" -H "Authorization: Bearer $(gcloud auth print-access-token)" https://vision.googleapis.com/v1/images:annotate -d '
{
  "requests": [
    {
      "image": {
        "source": {
          "gcsImageUri": "[IMAGEPATH]"
        }
      },
      "features": [
        {
          "maxResults": 200,
          "type": "WEB_DETECTION"
        },
      ]
    }
  ]
}' > OUT

This yields the following descriptive Web Entities:

 "webEntities": [
          {
            "entityId": "/m/0fczf",
            "score": 0.8428,
            "description": "Nurse"
          },
          {
            "entityId": "/m/03fk5c",
            "score": 0.61206055,
            "description": "Clinic"
          },
          {
            "entityId": "/g/1q2xhj1kf",
            "score": 0.4164,
            "description": "Postnikovo"
          },
          {
            "entityId": "/m/0hpnr",
            "score": 0.3084,
            "description": "Hospital"
          },
          {
            "entityId": "/m/02gnf8",
            "score": 0.2988,
            "description": "Derbent"
          },
          {
            "entityId": "/m/028hfb",
            "score": 0.2539321,
            "description": "Patient"
          },
          {
            "entityId": "/m/03y8krd",
            "score": 0.2536,
            "description": "Health fair"
          },
          {
            "entityId": "/t/2dgg60z74vt0",
            "score": 0.2155
          },
          {
            "entityId": "/m/0hzqwhh",
            "score": 0.2059,
            "description": "Moscow 24"
          },
          {
            "entityId": "/m/0kt51",
            "score": 0.2002,
            "description": "Health"
          },
          {
            "entityId": "/m/0221lk",
            "score": 0.1996592,
            "description": "Operating theater"
          },
          {
            "entityId": "/g/1pp2tj6lm",
            "score": 0.19844,
            "description": "Moroz"
          },
          {
            "entityId": "/m/03c1dkx",
            "score": 0.1885,
            "description": "Therapy"
          },
          {
            "entityId": "/m/012sdttf",
            "score": 0.1753,
            "description": "United Television Ghana"
          },
          {
            "entityId": "/m/04swd",
            "score": 0.07977,
            "description": "Moscow"
          }

Immediately clear is that this image likely has something to do with Derbent, Dagestan and potentially a nurse or medical professional. The Moscow 24 and United Television Ghana relate to Facebook pages on which the image was seen as part of video clips.

Among the exact matches Cloud Vision found on the web is a Facebook video titled "Медсестёр в Дербенте лечили в подсобке — Москва 24" which Google Translate offers as "Nurses in Derbent were treated in the back room – Moscow 24". Viewing the video on Facebook displays a description of "8 медсестёр разместили на лечение в подсобке инфекционного отделения Центральной Городской Больницы Дербента. У всех докторов предварительный диагноз ОРВИ.", which Google Translate translates as "8 nurses were placed for treatment in the back room of the infectious diseases department of the Central City Hospital of Derbent. All doctors have a preliminary diagnosis of ARVI [acute respiratory viral infection]."

None of the returned images have EXIF metadata, but a Google keyword search for the image tags returned by the Vision API of "derbent nurse hospital" yields an article in The Independent captioning the video as "nurses in a makeshift ward in Derbent, Dagestan in April" and CBC captioning it as "Social media video from Derbent, in the Russian republic of Dagestan, shows patients stacked in bunk beds to get treatment for coronavirus, with staff who aren't wearing face masks or protective gear. (MoshebabaV/YouTube)". The Guardian returns a textual match describing the clip as "a video leaked last month from a hospital in Derbent in Dagestan showed nurses being treated for coronavirus symptoms on shelves usually reserved for clean sheets."

A video on Yandex titled "Глава Дербента отреагировал на самолечение медсестер в подсобках" ("The head of Derbent reacted to the self-medication of nurses in the back rooms") offers the most detailed description of all with:

В соцсетях появилось видео, на котором медсестры Дербентской центральной городской больницы самостоятельно лечились от ОРВИ в кладовке. Причем в этом медучреждении находятся пациенты с COVID-19. На ситуацию отреагировал мэр города Хизри Абакаров. Он сообщил на своей странице в Instagram, что уже поговорил с главврачом Дербентской ЦГБ Абдулкафаром Шихмагомедовым. Тот подтвердил, что часть младшего персонала почувствовала себя плохо и организовала импровизированную палату в подсобном помещении. Медсестры работали в отделении, где находились люди с вирусными заболеваниями. Пациенты с СОVID-19 проходят лечение в другом месте. У девушек уже взяли анализ на коронавирус. Результат оказался отрицательным. Теперь сестры лечатся в другом медучреждении. «По итогам проверки заведующую инфекционным отделением отстранили от работы, выговоры получили инфекционист и заместитель главного врача по лечебной работе», — заявил глава Дербента. По последним данным, в Дагестане зарегистрировано 417 случаев заражения коронавирусом. Болезнь побороли 43 человека, еще 11 умерли. Среди жертв вируса оказалась 27-летняя беременная женщина из Хасавюрта. Ее госпитализировали на вертолете в больницу Махачкалы. Однако медики не смогли ее спасти.

A video appeared on social networks in which the nurses of the Derbent central city hospital were independently treated for ARVI in the closet. Moreover, in this medical institution there are patients with COVID-19. The mayor of the city Khizri Abakarov reacted to the situation. He said on his Instagram page that he had already talked with the head physician of the Derbent Central City Hospital Abdulkafar Shikhmagomedov. He confirmed that some of the junior staff felt unwell and organized an impromptu ward in the back room. The nurses worked in the department where there were people with viral diseases. Patients with COVID-19 are being treated elsewhere. The girls have already been tested for coronavirus. The result was negative. Now the sisters are being treated in another medical facility. “Following the results of the check, the head of the infectious diseases department was dismissed from work, the infectious disease specialist and the deputy chief physician for medical work were reprimanded,” the head of Derbent said. According to the latest data, 417 cases of coronavirus infection have been registered in Dagestan. 43 people fought the disease, 11 more died. Among the victims of the virus was a 27-year-old pregnant woman from Khasavyurt. She was hospitalized by helicopter to a hospital in Makhachkala. However, doctors could not save her.

From an unknown clip on CNN, Cloud Vision combined with a bit of human analysis led us to a description of what was actually depicted and basic provenance information, allowing us if needed to dig further into the veracity and context of the clip.

This work was made possible through the Media-Data Research Consortium (M-DRC)'s Google Cloud COVID-19 Research Grant to support “Quantifying the COVID-19 Public Health Media Narrative Through TV & Radio News Analysis.”