Microsoft

Content Moderator

Custom Scenarios – Have a content moderation scenario (content type or usage) that you would like to suggest?

Documentation – Any ideas or suggestions for the API Reference or Documentation.

Product – Any ideas or feedback pertaining to features or enhancements to Moderator APIs or the review tool.

Samples & SDK Request – Let us know if you would like to see a Code sample or SDK provided.

Sample Content – Have some content you’ve tested and not getting the results you are seeking? Upload and describe your expectations.

  1. Better document what it means when offensive scoring results in generic high scores

    I'm reviewing a test output of 100,000 items with the Text Moderator. A larger number of the OffensiveScore outputs appear as one of two high numbers:
    0.98949163 or 0.987999976

    After those it drops off quite dramatically to much lower numbers that are more differentiated and incremental.

    What do these two numbers entail? Why are they so generic/identical for thousands of items?

    I'm guessing it may have something to do with the Review threshold... but it's unclear.

    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Documentation  ·  Flag idea as inappropriate…  ·  Admin →
  2. Small images error our: Image size is not in allowed range

    SVGs often error out when using IsImageAdultClassified because of their small size. For example, I upload a 4kb image and I get the following error:
    {"Message":"Image Size Error","TrackingId":"-redacted-","Errors":[{"Title":"Image Size Error","Message":"Image size is not in allowed range."}]}

    At the very least, update the documentation to reflect that there is a minimum size. I can't find this info anywhere.

    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  3. Screen API - Offensive Emojis

    Emojis can be offensive.
    The middle finger emoji for example is not screened as offensive by the system.

    Here is an example response I received:

    {
    "OriginalText": "🖕",
    "NormalizedText": "🖕",
    "Misrepresentation": null,
    "Classification": {

    "ReviewRecommended": false,
    
    "Category1": {
    "Score": 0.00019999999494757503
    },
    "Category2": {
    "Score": 0.2296631783246994
    },
    "Category3": {
    "Score": 0.21779759228229523
    }

    },
    "Language": "eng",
    "Terms": null,
    "Status": {

    "Code": 3000,
    
    "Description": "OK",
    "Exception": null

    },
    "TrackingId": "REDACTED"
    }

    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Product  ·  Flag idea as inappropriate…  ·  Admin →
  4. PII is generally pretty broken, particularly for phone number

    PII filter is not functioning correctly when ONLY PII is present in the body, e.g. phone number.

    Example: 202-555-0156 does not return a PII object, whereas "Test phone number 202-555-0156" does. This makes it pretty unusable when people are, i don't know, exchanging phone numbers when they shouldn't be?

    Likewise, something like "aaaa aaaa 203-456-5928" doesn't return it either. Can you either fix this or eliminate any charges for PII calls? This is broken to the point of being unusable and Microsoft shouldn't be charging for this.

    2 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  5. Screen API - Even though the 'listId' query param is empty or excluded from the request, response uses one of my created lists

    Screen API - Even though the 'listId' query param is empty or excluded from the request, response uses one of my created lists

    Here is the curl:
    curl -X POST \ 'https://westus.api.cognitive.microsoft.com/contentmoderator/moderate/v1.0/ProcessText/Screen?autocorrect=false&PII=false&classify=True' \ -H 'Accept: /' \ -H 'Accept-Encoding: gzip, deflate' \ -H 'Cache-Control: no-cache' \ -H 'Connection: keep-alive' \ -H 'Content-Length: 18' \ -H 'Content-Type: text/plain' \ -H 'Host: westus.api.cognitive.microsoft.com' \ -H 'Postman-Token: a477b3d6-f0e9-4964-bf90-b6b861154d7c,eb527ed8-0d1a-47c1-9c80-dba784dbf0b9' \ -H 'User-Agent: PostmanRuntime/7.19.0' \ -H 'cache-control: no-cache' \ -d '1 800 1801 stupid '

    Here is the response:
    { "OriginalText": "1 800 1801 stupid ", "NormalizedText": "1 800 1801 stupid ",…

    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Product  ·  Flag idea as inappropriate…  ·  Admin →
  6. 1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    1 comment  ·  Product  ·  Flag idea as inappropriate…  ·  Admin →
  7. 1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Documentation  ·  Flag idea as inappropriate…  ·  Admin →
  8. Detect UK Bank Accounts and Credit Cards

    Spot leaking financial information.

    Bank accounts are typically combined with a sort code so 999999 99999999 or perhaps 99-99-99 99999999.

    Credit cards will be 9999999999999999 or 9999 9999 9999 9999 format.

    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Custom Scenarios  ·  Flag idea as inappropriate…  ·  Admin →
  9. Have a way to define regex patterns to flag PII such as employee ids

    I don't want to have to create custom lists for text patterns like employee ids so this PII can be flagged with built-in capabilities

    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Product  ·  Flag idea as inappropriate…  ·  Admin →
  10. Have names be flagged as PII (this is supported in other systems such as Google)

    I don't want to have to create custom lists for first and last names but rather have a name option so this PII can be flagged with built in capabilities

    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Product  ·  Flag idea as inappropriate…  ·  Admin →
  11. Review demo is not working

    The demos for text, image and video, described in the tutorial are not working.

    2 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    Resolved  ·  1 comment  ·  Documentation  ·  Flag idea as inappropriate…  ·  Admin →
  12. demora en cargar la informacion

    debe ser necesario una mejora

    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Custom Scenarios  ·  Flag idea as inappropriate…  ·  Admin →
  13. demora en cargar la informacion

    debe ser necesario una mejora

    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Custom Scenarios  ·  Flag idea as inappropriate…  ·  Admin →
  14. Text Moderation in large files

    Currently since the character limit of the content moderator REST API is 1024 characters, the character stream of the file needs to be broken down into multiple parts which affect the relation between words and further more the text moderation.

    Kindly add a method to break large character strings into smaller logical parts.

    4 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    Under Review  ·  0 comments  ·  Sample & SDK Requests  ·  Flag idea as inappropriate…  ·  Admin →
  15. Ability to define patterns to recognize PII like phone number of a particular region

    Would be nice to have ability to define patterns to recognize PII like phone number of a particular region

    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    Under Review  ·  0 comments  ·  Product  ·  Flag idea as inappropriate…  ·  Admin →
  16. shahid opu

    feedback

    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Custom Scenarios  ·  Flag idea as inappropriate…  ·  Admin →
  17. Security Issue - SSL Cookie Without Secure Flag Set & Cross Site Scripting (XSS)

    There are a few concerns for which we need justification or if applicable a fix to resolve issue. We contacted the support team and they suggested to mention the issue in this form as well for PG tracking. Here is the response I got from Support team:

    "Issue 1:
    The cookie we see below is an affinity cookie being dropped by Azure Websites infra for more efficient routing. Having said that, the Review API itself is stateless and does not have any user sessions. We will plan to get this rectified in our next release.

    Issue 2:
    We will create…

    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    Under Review  ·  0 comments  ·  Product  ·  Flag idea as inappropriate…  ·  Admin →
  18. Remove text from images

    We need the ability to not only detect text in images but also to remove it or at least get the coordinates in the OCR data so we can remove it ourselves.

    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Product  ·  Flag idea as inappropriate…  ·  Admin →
  19. Need better way to distinguish offensive content

    Consider the following two tweets. The first is benign. The second is offensive. How can I know that progamatically? They have the same ReviewRecommended, the same Category3 score, low Cat1 and Cat2 scores, and neither have any Terms flagged. This is useless.

    {

    &quot;OriginalText&quot;: &quot;Wanna win tix to our @RoughTradeNYC show NEXT MON 5/14?  @Thrillcall&#39;s got you covered. Enter to win here &amp;gt;… <a rel="nofollow noreferrer" href="https://t.co/s0KT4JDCB1&quot;">https://t.co/s0KT4JDCB1&quot;</a>,
    
    &quot;NormalizedText&quot;: &quot;Wanna win tix to our @ RoughTradeNYC show NEXT MON 5/ 14? @ Thrillcall&#39; s got you covered. Enter to win here &amp; gt; … <a rel="nofollow noreferrer" href="https://t.co/s0KT4JDCB1&quot;">https://t.co/s0KT4JDCB1&quot;</a>,
    &quot;Misrepresentation&quot;: null,
    &quot;Classification&quot;:
    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    Resolved  ·  2 comments  ·  Sample Content  ·  Flag idea as inappropriate…  ·  Admin →
  20. Multilanguage Recognition

    Add the ability to pick up and moderate text from different languages in a single piece of content. If I had to use a profane word in my home language in an English piece of content it will not be picked up.

    2 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    Under Review  ·  0 comments  ·  Custom Scenarios  ·  Flag idea as inappropriate…  ·  Admin →
← Previous 1
  • Don't see your idea?

Feedback and Knowledge Base