'Similar' claims are variations of the same claim or story. By definitions, those variations can be summarized or fact-checked using the same report or analysis.
This article explains:
Enabling, disabling, and limiting similarity matching
Navigate to the ‘Similarity’ tab (1), and click the slider (2).
You can also limit similarity matching to recently submitted content (3) by indicating how old an item must be in order to be considered for similarity matching (4). Items older than that setting will only be suggested for matching.
Save your similarity settings by clicking ‘Update’ (5).
On your Check workspace, similar items:
Inherit the same status as main items they are matched with, and
Inherit the same Tipline report as that sent by publishing the main item
From your workspace dashboard, using the similarity slider (1), you can easily spot main items which have blue icons (2) and secondary matched items indicated with grey icons (3).
If the similarity switch is ON, for matched items, actions such as deleting or moving to different folders is not possible from this view, as the checkbox is disabled (1). This is different from main items where the checkbox is enabled (2), and it is possible to move or delete the item.
The similarity switch is only visible if the similarity matching is enabled in the workspace.
Check’s algorithm evaluates content according to:
Media - including titles, descriptions, URL link content, image, video, and audio fingerprints, and
Published and unpublished reports - titles, descriptions, visual card titles and content.
Similarity matching only works within media types:
Text items are matched to text items*, link items, and images with optical character recognition (OCR) extracted text
Image items are only matched with similar images,
Link items are suggested to similar link and text items.
Video items are only matched with video items**.
Audio items are only matched to similar audio files.
* Text items with less than eight words are always suggested and not automatically matched.
** Video matching only uses the visual content and does not consider audio.
Note: Imported reports are not automatically matched as secondary to any item, including other imported reports.
Automatic (machine) matching
When a new item is added to your workspace, Check will compare its media to all other existing item's media of the same type and the best match found. If this match is above the Check’s similarity matching threshold, then the new item will be added to the same cluster as the existing message.
Automatic matching works in the following ways:
If the new media's type is a Text item, it will be compared to all other text media types in your workspace by measuring the similarity in meaning of sentences. Amongst Check’s text similarity models is one specialized for Indian languages. Workspaces can also be configured to different similarity models according to language used.
If the new media's type is an Image, it will be compared to all other image media types in your workspace.
It is possible for automatic matching to be disabled for certain media types. Please get in touch if you want to know which media types have automatic matching in your workspace or want to make a change.
Text and Image algorithms will evaluate how similar the new media is to other existing ones and if similarities are found, it will classify the similarity as one of the following:
If the algorithm identifies that items are similar, but it does not reach its confidence threshold, then they are added as “suggested media”.
Clicking the ‘SUGGESTED MEDIA’ link displays the item(s) that the algorithm assumes closely matches the main item
To reject the suggested item, click ‘X’
If the algorithm is confident that items match or surpass the threshold of similarity, they are automatically added as “similar media”.
Important! Make sure you add the column 'Similar' to 'Displayed columns' in your list settings in order to be able to sort items and find the ones that have the most similar media. Those may be the ones you want to fact-check first!
Within an item, you can see all its confirmed media by clicking '#'* Confirmed similar media'
* '#' denotes a number
The main item will be displayed at the top of the list (1), with all confirmed similar media items below it (2).
If there are more matched items on the page than can be shown on your screen, use the scroll bar to view the matched items (3). The top bar showing the similar media count will remain frozen at the top. You will also be able to see who matched the items (4).
Adding similar media manually to the 'Confirmed' list
If Check doesn't automatically group similar media together, you can do so manually:
STEP #1: Click the 'Add similar' button on the item page
STEP #2: Choose whether you want to add media to the current item (Import similar media to this item), if you want to send the media in the current item to another item (Export all media to another item) or if you want to add the item to an imported report from your non-Check database.
Matched items will be moved to the same folder as the main item.
Importing similar media
To import similar media, after clicking on the ‘ADD SIMILAR’ button, click ‘Import similar media to this item’. This opens a search window, where inputting the type of similar items, will display item cards that match the search.
Clicking the gear opens up extra settings to refine your search (1). Use the scroll to view more search options (2).
Search options are filtered according to the media title and content, tags, metadata, and tasks. Click ‘UPDATE’ when you select the search settings to refresh the search result media cards.
You can select multiple item cards, then click ‘IMPORT ALL MEDIA FROM ONE ITEM’.
You can also export all media to another item by clicking the ‘ADD SIMILAR’ button.
Further search filters are available according to the media title and content, tags, metadata and tasks. Click ‘UPDATE’ to refresh the search result media cards, then ‘EXPORT ALL MEDIA’.
Add to Imported Report
If you import reports from your workspace to Check, this option allows you to add attach your externally generated reports to the main item on Check.
Input the report topic in the search bar and select the item where you want the open item to appear as similar.
The 'Main media' at the top of the screen will by default be the one that has been added first to a workspace. The item will use this media to provide the thumbnail overview as well as the analysis Title and Content for the item.
To choose any other media in the list as 'Main', click the option 'Pin as Main'
Detaching matched items
Click ‘Detach’ to remove the item from the similar media list and select the destination list
Viewing machine (automatic) vs. human attached media
From your workspace page, you will be able to view similar items that have either been machine or human-human matched.
To review your attached similar items, on the main item, click the ‘SIMILAR MEDIA’ link
The attached items will then be displayed, showing you how they were matched to the main item, either by human or machine action
Once you have selected all similar items to your main item and verified that they match, you can send reports to all requesters whose items are matched to the main item.