Methods for reinforcement document transformer for multimodal conversations and devices thereof

A system comprising:

• a memory;
• and
• a processor comprising a reinforcement document transformer configured to:
• create one or more corpus embeddings for one or more sentences in a corpus;
• store the one or more corpus embeddings in the memory;
• perform a semantic comparison of query embeddings and the one or more corpus embeddings;
• determine a closest corpus embedding by identifying relevant documents from the corpus for one or more query inputs;
• retrieve one or more corpus responses based on the closest corpus embedding using a transformer model;
• assign custom weights to key terms;
• and
• compute a total weighted score for the one or more corpus responses to identify most appropriate corpus response.

A system comprising:

• a memory;
• and
• a processor comprising a reinforcement document transformer configured to:
• generate a tree structure for each corpus document based on extracted structural information of the corpus document;
• create one or more corpus embeddings for one or more sentences in a corpus;
• store the one or more corpus embeddings in the memory;
• perform a semantic comparison of query embeddings and the one or more corpus embeddings;
• determine a closest corpus embedding by identifying relevant documents from the corpus for one or more query inputs;
• retrieve one or more corpus responses based on the closest corpus embedding using a transformer model;
• identify key terms in the one or more corpus responses and query using the tree structure to identify most appropriate corpus response;
• compare the key terms with the corpus response;
• and
• determine an extent of a content based on the key terms matching in a level of a hierarchy.