Systems and methods for using link graphs to demote links to low-quality webpages

A content-feed system comprising:

• an identifying module, stored in memory, that receives user-provided content items from users of the content-feed system for presentation to other users of the content-feed system;
• a sampling module, stored in memory, that samples user-provided links from the user-provided content items;
• a receiving module, stored in memory, that receives, from at least one human labeler for each of the user-provided links, at least one label indicating whether the human labeler considers a landing page of the user-provided link to be a low-quality webpage;
• a generating module, stored in memory, that generates a link graph from a crawl of the user-provided links, wherein the landing page of each of the user-provided links is represented in the link graph by a vertex;
• a deriving module, stored in memory, that uses the link graph to derive at least one graph-based feature for each of the user-provided links;
• a training module, stored in memory, that uses the label and the graph-based features of each of the user-provided links to train a model to predict likelihoods that user-provided links area to low-quality webpages, wherein the identifying module further receives additional user-provided content items that are candidates for presentation via a content feed of an additional user of the content-feed system;
• an applying module, stored in memory, that applies the model to a link of each of the additional user-provided content items to determine a likelihood that a link of the additional user-provided content item is to a low-quality webpage;
• a displaying module, stored in memory, that displays the content feed to the additional user by:
• presenting, to the additional user, at least one of the additional user-provided content items whose likelihood is below a predetermined threshold;
• and
• refraining from presenting, to the additional user, at least one of the additional user-provided content items whose likelihood is above the predetermined threshold;
• and
• at least one physical processor configured to execute the sampling module, the receiving module, the generating module, the deriving module, the training module, the identifying module, the applying module, and the displaying module.

A computer-readable medium comprising computer-readable instructions that, when executed by at least one processor of at least one computing device of a content-feed system, cause the computing device to:

• receive user-provided content items from users of the content-feed system for presentation to other users of the content-feed system;
• sample user-provided links from the user-provided content items;
• receive, from at least one human labeler for each of the user-provided links, at least one label indicating whether the human labeler considers a landing page of the user-provided link to be a low-quality webpage;
• generate a link graph from a crawl of the user-provided links, wherein the landing page of each of the user-provided links is represented in the link graph by a vertex;
• use the link graph to derive at least one graph-based feature for each of the user-provided links;
• use the label and the graph-based feature of each of the user-provided links to train a model to predict likelihoods that user-provided links are to low-quality webpages;
• receive additional user-provided content items that are candidates for presentation via a content feed of an additional user of the content-feed system;
• apply the model to a link of each of the additional user-provided content items to determine a likelihood that a link of the additional user-provided content item is to a low-quality webpage;
• and
• display the content feed to the user by:
• presenting, to the additional user, at least one of the additional user-provided content items whose likelihood is below a predetermined threshold;
• and
• refraining from presenting, to the additional user, at least one of the additional user-provided content items whose likelihood is above the predetermined threshold.