Automatic embedding of additional content to articles

A system, comprising:

• one or more processors;
• memory in electronic communication with the one or more processors;
• and
• instructions stored in the memory, the instructions being executable by the one or more processors to:
• identify a domain for an article;
• identify an article type for the article using one or more domain-specific smart tags for the identified domain;
• convert at least a portion of the article into a document vector using a pre-trained domain-specific language model for the identified domain, wherein the pre-trained domain-specific language model converts the portion of the article into a numeric vector representation located within a vector space specific to the identified domain, and wherein the pre-trained domain-specific language model is trained on a set of input text tailored to the domain such that the pre-trained domain-specific language model learns to differentiate nuances in vocabulary words particular to the domain;
• query a datastore using the document vector, wherein the datastore stores media content having the identified domain and the identified article type, and wherein the media content stored in the datastore is associated with numeric vector representations;
• obtain media content results in response to the querying;
• rank the obtained media content results based on temporal proximity of the media content to an event described in the article and on relevance scores assigned by a pre-determined relevance model;
• generate a list of additional content, the list of additional content including the obtained media content results based on the ranking;
• embed the list of additional content with the article;
• store the article with the list of additional content embedded with the article;
• cause the article including the embedded list of additional content to be presented;
• and
• update the list of additional content embedded with the article at a predetermined interval, the updated list based on temporal proximity to the event described in the article and relevance scores assigned by the relevance model.

A method for identifying additional content about an event in an article, comprising:

• receiving an identification of a domain for the article;
• converting at least a portion of the article into an article document vector using a pre-trained domain-specific language model for the domain, wherein the pre-trained domain-specific language model is trained on a set of input text specific to the domain, and wherein the pre-trained domain-specific language model converts the portion of the article into a numeric vector representation located within a vector space specific to the identified domain, and wherein the pre-trained domain-specific language model is trained on a set of input text tailored to the domain such that the pre-trained domain-specific language model learns to differentiate nuances in vocabulary words particular to the domain;
• converting at least a portion of content stored in a datastore into a content document vector using the pre-trained domain-specific language model for the domain, wherein the pre-trained domain-specific language model converts the portion of the content into an additional numeric vector representation located within the vector space specific to the identified domain;
• generating a list of predicted matches using a relevance score for the article document vector and the content document vector, wherein the relevance score indicates a similarity between the article and the content and is determined by a relevance model and the list of predicted matches includes the content similar to the article;
• applying at least one filter to the list of predicted matches to filter the content based on temporal proximity to the event described in the article;
• generating a list of additional content in response to applying the at least one filter to the list of predicted matches;
• storing the article with the list of additional content in a datastore;
• causing the article and the list of additional content to be presented, the list of additional content embedded with the article;
• and
• updating the embedded list of additional content at a predetermined interval, the updated list based on temporal proximity to the event described in the article and relevance scores assigned by the relevance model.