×

Sequence to sequence to classification model for generating recommended messages

  • US 10,721,190 B2
  • Filed: 07/31/2018
  • Issued: 07/21/2020
  • Est. Priority Date: 07/31/2018
  • Status: Active Grant
First Claim
Patent Images

1. A method comprising:

  • receiving a first message being transmitted as part of communication session, the first message including a plurality of sequenced words;

    generating, using a sequence to sequence encoder included in a sequence to sequence to classification model, a first embedding vector representing the plurality of sequenced words included in the first message the sequence to sequence to classification model including the sequence to sequence encoder and a text classification model, the sequence encoder having been trained based on historical message data that includes messages transmitted between users of a messaging system and the text classification model having been trained based on embedding vectors generated by the sequence to sequence encoder from the historical message data;

    generating a feature vector based on the first embedding vector and at least a first feature that is not included in the first embedding vector;

    generating a set of candidate responses for replying to the first message by using the feature vector as input into the text classification model, the set of candidate responses being a subset of a set of available responses;

    selecting, from the set of candidate responses, a set of recommended responses to the first message, the set of recommended responses being a subset of the set of candidate responses; and

    causing the set of recommended responses to be presented by a client device of a recipient user of the first message.

View all claims
  • 1 Assignment
Timeline View
Assignment View
    ×
    ×