| CPC G06F 16/24558 (2019.01) [G06F 16/24578 (2019.01); G06N 20/00 (2019.01)] | 20 Claims |

|
1. A computer-implemented method for answer ranking for question answering comprising:
determining an index using a machine learning model trained on datasets by:
converting a text of the datasets to a sparse vector representation, wherein the text includes answer candidates;
storing the converted text in a database associated with the index; and
generating a structure associated with the index that represents a relationship between questions and the answer candidates;
determining a score between the questions and the answer candidates from the index, wherein the score includes a summation; and
providing a real-time inference by:
receiving a question from a user;
determining a rank feature of the sparse vector representation for the question and the answer candidates in a time associated with the index;
storing the rank feature in the index;
determining a final ranking score by computing a summation of a look up of the stored rank feature in the index; and
providing a natural language output to the user that includes at least a part of the answer candidates for the question based on the final ranking score.
|