Abstract: Embodiments of the present specification disclose a method, an apparatus, a system, a device, and a storage medium for answering user questions, including: obtaining a user question; encoding the user question and a schema level of pre-constructed structured data to obtain a first feature vector, wherein the structured data further comprises a data level, wherein the data level comprises knowledge for answering questions structured according to the schema level; retrieving one or more candidate sub-graphs related to the user question from the structured data; encoding the one or more candidate sub-graphs to obtain a second feature vector; performing multi-task classification for the user question based on the first feature vector and the second feature vector; and obtaining answer content for the user question based on a result of the multi-task classification.

Abstract: Embodiments of the present specification provide methods for performing marketing cost control by using a deep reinforcement learning system. One method includes the following: determining a cost of a marketing activity; determining a reward score of reinforcement learning that is negatively correlated with the cost; and returning the reward score to a smart agent of a deep reinforcement learning system, for the smart agent to update a marketing strategy, wherein the smart agent is configured to determine a marketing activity based on the marketing strategy and status of an execution environment of the deep reinforcement learning system.

Abstract: Embodiments of the present specification disclose a method, an apparatus, a system, a device, and a storage medium for answering user questions, including: obtaining a user question; encoding the user question and a schema level of pre-constructed structured data to obtain a first feature vector, wherein the structured data further comprises a data level, wherein the data level comprises knowledge for answering questions structured according to the schema level; retrieving one or more candidate sub-graphs related to the user question from the structured data; encoding the one or more candidate sub-graphs to obtain a second feature vector; performing multi-task classification for the user question based on the first feature vector and the second feature vector; and obtaining answer content for the user question based on a result of the multi-task classification.

Abstract: Embodiments of the present specification provide deep reinforcement learning methods and apparatuses for referral marketing. One method includes the following: obtaining state information of an execution environment of a deep reinforcement learning system, wherein the state information comprises user information of a current user of the deep reinforcement learning system; determining a marketing activity corresponding to the state information based on a marketing strategy, wherein the marketing activity comprises a combination of a marketing channel, marketing content, and a marketing time period; obtaining a reward score of the execution environment for the marketing activity; and updating the marketing strategy based on the reward score.

Abstract: Embodiments of the present specification provide marketing methods based on a deep reinforcement learning system. One method includes the following: obtaining, from an execution environment of a deep reinforcement learning system, a plurality of execution results generated by a user in response to marketing activities, wherein the plurality of execution results correspond to a plurality of targeted effects on a marketing effect chain; determining a reward score of reinforcement learning based on the plurality of execution results; and returning the reward score to a smart agent of the deep reinforcement learning system, for the smart agent to update a marketing strategy, wherein the smart agent is configured to determine the marketing activities based on the marketing strategy and status of the execution environment.

Abstract: Embodiments of the present specification provide methods for performing marketing cost control by using a deep reinforcement learning system. One method includes the following: determining a cost of a marketing activity; determining a reward score of reinforcement learning that is negatively correlated with the cost; and returning the reward score to a smart agent of a deep reinforcement learning system, for the smart agent to update a marketing strategy, wherein the smart agent is configured to determine a marketing activity based on the marketing strategy and status of an execution environment of the deep reinforcement learning system.

Abstract: Embodiments of the present specification provide deep reinforcement learning methods and apparatuses for referral marketing. One method includes the following: obtaining state information of an execution environment of a deep reinforcement learning system, wherein the state information comprises user information of a current user of the deep reinforcement learning system; determining a marketing activity corresponding to the state information based on a marketing strategy, wherein the marketing activity comprises a combination of a marketing channel, marketing content, and a marketing time period; obtaining a reward score of the execution environment for the marketing activity; and updating the marketing strategy based on the reward score.

Abstract: Embodiments of the present specification provide marketing methods based on a deep reinforcement learning system. One method includes the following: obtaining, from an execution environment of a deep reinforcement learning system, a plurality of execution results generated by a user in response to marketing activities, wherein the plurality of execution results correspond to a plurality of targeted effects on a marketing effect chain; determining a reward score of reinforcement learning based on the plurality of execution results; and returning the reward score to a smart agent of the deep reinforcement learning system, for the smart agent to update a marketing strategy, wherein the smart agent is configured to determine the marketing activities based on the marketing strategy and status of the execution environment.