Artificial intelligence based-document processing

A method of processing an interactive document comprising:

• enabling presentation of an interactive visualization to a user, the interactive visualization including knowledge nuggets regarding one or more primary entities that are extracted from an information request;
• receiving logs of one or more recorded user interactions with the interactive visualization;
• identifying the one or more recorded user interactions with the interactive visualization via natural language processing of the logs;
• identifying an information gap corresponding to one or more of the primary entities within the interactive visualization presented to the user based on absence of predetermined conclusive user actions from the one or more recorded user interactions;
• extracting one or more auxiliary entities from the logs of the one or more recorded user interactions using named entity recognition techniques;
• obtaining one or more search terms related to the auxiliary entities from at least a knowledge base and internet by employing trained information extraction (IE) models;
• identifying relevant documents from one or more of the internet and the knowledge base using one or more of the search terms, the primary entity and the auxiliary entities;
• extracting datasets responsive to the information gap from the relevant documents using the trained IE models and rules from the knowledge base;
• and
• automatically updating the interactive visualization using the extracted datasets.

A non-transitory computer-readable storage medium comprising machine-readable instructions that cause a processor to:

• receive a request for information, the request for information including one or more primary entities;
• extract the primary entities from the request using at least named entity recognition (NER);
• obtain one or more search terms related to the primary entities from at least a knowledge base and internet by employing trained machine learning (ML) models in conjunction with N-gram analysis;
• identify relevant documents from one or more of the internet and the knowledge base using one or more of the search terms and the primary entities;
• extract data responsive to the request as one or more knowledge nuggets from the relevant documents using trained information extraction (IE) models and rules from the knowledge base;
• automatically select at least one visualization template for generating an interactive visualization of the extracted data, the visualization template being selected from a collection of visualization templates;
• generate the interactive visualization by incorporating at least the knowledge nuggets within the selected visualization template;
• receive one or more recorded user actions executed via the interactive visualization;
• determine an information gap associated with the interactive visualization based on a mapping of the recorded user actions to particular intents;
• and
• update the interactive visualization in response to the determination of the information gap and in accordance with the mapping, wherein updating the interactive visualization includes:
• obtaining additional search terms based on the information gap;
• and
• based on the additional search terms,
• repeating the steps of identifying the relevant documents, extracting the data, automatically selecting the at least one visualization template, generating the interactive visualization and receiving the recorded user actions.