Computerized system and method for automatic moderation of online content
The disclosed systems and methods provide a framework for a proactive prediction of the toxic propensity of an article. Prior to the publication and/or reception of comments to online content, the disclosed framework determines the toxic propensity of the content's context and/or specific words, sentences, sentiments, tone or other messages receivable from consumption of the content. Thus, disclosed framework performs proactive forecasting of the content's toxicity propensity”, which quantifies how likely the content is prone to incur or attract toxic comments. The framework can function and/or be configured to operate in a manner that can perform specifically adherent moderation actions that correspond to the content and control how the content can be interacted with, based on the toxic propensity determination, prior to the content's publication in an effort to thwart, prevent or stop toxic environments surrounding or stemming from the content from coming into existence.
1. A method comprising:
training, by a device, using training data, a regression classifier to determine a toxic propensity score based on input;
identifying, by the device, an article, the article comprising a plurality of words comprised of text;
analyzing, by the device, the article, and identifying, as the input to the regression classifier, article information related to the plurality of word;
executing, by the device, the regression classifier on the article information as the input to the regression classifier, execution of the regression classifier on the article information determining, for the article, a toxic propensity score indicating how likely the article is to attract toxic comments;
determining, by the device, a type of moderation to perform on the article based on the determined toxic propensity score indicating that the article is likely to attract toxic comments; and
moderating, by the device, the article based on the type of determined moderation, the moderation comprising providing instructions for the article comprising the plurality of words comprised of text to be published with commenting disabled.
2. The method of claim 1 , further comprising:
publishing the moderated article at a network location with commenting disabled.
3. The method of claim 1 , providing instructions further comprising:
augmenting the article with the instructions for the article to be published with commenting disabled.
4. The method of claim 1 , further comprising:
identifying a set of words that cause the toxic propensity score to be above a toxicity threshold; and
providing a suggestion for a modification of the set of words, wherein the modifications to the portion of text correspond to the suggestion.
5. The method of claim 4 , further comprising:
identifying an alternative set of words as an alternative to the identified set of words, the alternative set of words enabling lowering of the toxic propensity score; and
providing the alternative set of words within the suggestion.
6. The method of claim 4 , further comprising:
executing, by the device, the regression classifier on a revised article responsive to the provided suggestion, and determining, for the revised article, a new toxic propensity score indicating how likely the revised article is to attract toxic comments;
determining that the new toxic propensity score is below a toxicity threshold; and
publishing the revised article at a network location without disabling commenting in connection with publication of the article.
7. The method of claim 1 , wherein training a regression classifier further comprising:
identifying, by the device, a set of training articles;
analyzing, by the device, each training article, and identifying a set of text features;
performing a Beta regression analysis on the set of text features;
executing a machine learning (ML) classifier based on the Beta regression analysis; and
compiling a Beta regression ML classifier based on the execution of the ML classifier.
8. The method of claim 7 , wherein the regression classifier executed on the article information is the compiled Beta regression ML classifier.
9. The method of claim 7 , wherein the text features correspond to feature vectors for each article, wherein identification of the set of training articles further comprises identifying each training article's feature vector.
10. The method of claim 1 , further comprising:
determining a plurality of toxic propensity scores in relation to the plurality of words; and
determining an average toxic propensity score, wherein the determined toxic propensity score for the article is based on the average toxic propensity score.
11. The method of claim 10 , wherein each of the plurality of toxic propensity scores corresponds to an individual word in the article.
12. The method of claim 10 , wherein each of the plurality of toxic propensity scores corresponds to a combination of a set of words in the article.
13. The method of claim 1 , further comprising:
receiving, from a user, a request to publish the article, wherein the identification of the article is based on the request.
14. The method of claim 1 , wherein the article information further comprises data selected from a group consisting of: the text, vector data of the article, author identity, source identity, destination identity, topic, types of words, word count, paragraph count, sentence count and other content in the article.
15. A non-transitory computer-readable storage medium tangibly encoded with computer-executable instructions, that when executed by a processor associated with a device, performs a method comprising:
training, by the device, using training data, a regression classifier to determine a toxic propensity score based on input;
identifying, by the device, an article, the article comprising a plurality of words comprised of text;
analyzing, by the device, the article, and identifying, as input to the regression classifier, article information related to the plurality of words;
executing, by the device, the regression classifier on the article information as the input to the regression classifier, execution of the regression classifier on the article information determining, for the article, a toxic propensity score indicating how likely the article is to attract toxic comments;
determining, by the device, a type of moderation to perform on the article based on the determined toxic propensity score indicating that the article is likely to attract toxic comments;
moderating, by the device, the article based on the type of determined moderation, the moderation comprising providing instructions for the article comprising the plurality of words comprised of text to be published with commenting disabled; and
publishing the moderated article at a network location.
16. The non-transitory computer-readable storage medium of claim 15 , providing instructions further comprising:
augmenting the article with instructions for the article to be published with commenting disabled.
17. The non-transitory computer-readable storage medium of claim 15 , further comprising:
identifying a set of words that cause the toxic propensity score to be above a toxicity threshold;
identifying an alternative set of words for the identified set of words, the alternative set of words enabling lowering of the toxic propensity score; and
providing a suggestion for the modification of the set of words, the suggestion comprising information related to the alternative set of words, wherein the modifications to the portion of text correspond to the suggestion.
18. A device comprising:
a processor configured to:
train, using training data, a regression classifier to determine a toxic propensity score based on input;
identify an article, the article comprising a plurality of words comprised of text;
analyze the article, and identify, as input to the regression classifier, article information related to the plurality of words;
execute the regression classifier on the article information as the input to the regression classifier, execution of the regression classifier on the article information determining, for the article, a toxic propensity score indicating how likely the article is to attract toxic comments;
determine a type of moderation to perform on the article based on the determined toxic propensity score indicating that the article is likely to attract toxic comments;
moderate the article based on the type of determined moderation, the moderation comprising providing instructions for the article comprising the plurality of words comprised of text to be published with commenting disabled; and
publish the moderated article at a network location.
19. The device of claim 18 , providing instructions further comprising:
augment the article with the instructions for the article to be published with commenting disabled.
20. The device of claim 18 , further comprising:
identify a set of words that cause the toxic propensity score to be above a toxicity threshold;
identify an alternative set of words for the identified set of words, the alternative set of words enabling lowering of the toxic propensity score; and
provide a suggestion for a modification of the set of words, the suggestion comprising information related to the alternative set of words, wherein the modifications to the portion of text correspond to the suggestion.