Automatically suggesting resources for responding to a request
A user may respond to a request of another user by entering text, such as a customer service representative responding to a customer. Suggestions of resources may be provided to the responding user to assist the responding user in providing a response. For example, a resource may provide information to the responding user or allow the responding user to perform an action. Previous messages between the two users and other information may be used to select a resource. A conversation feature vector may be determined from previous messages, and feature vectors may be determined from the resources. The conversation feature vector and the feature vectors determined from the resource may be used to select a resource to suggest to the responding user.
1. A computer-implemented method for suggesting a resource to a second user in responding to a first user, the method comprising:
obtaining a plurality of resource feature vectors for a plurality of resources, wherein each resource feature vector is computed from text of a corresponding resource;
receiving first text of a first message between the first user and the second user;
generating a first conversation feature vector by processing the first text of the first message with a neural network, wherein the neural network is configured to process text of a conversation to compute a conversation feature vector that is close to resource feature vectors of resources that are relevant to the conversation;
computing a plurality of scores, wherein in each score is computed by comparing the first conversation feature vector to a resource feature vector of the plurality of resource feature vectors, and wherein the plurality of scores comprises a first score computed for a first resource by comparing the conversation feature vector to a first resource feature vector for the first resource;
selecting the first resource from the plurality of resources using the plurality of scores; and
causing the first resource to be presented to the second user.
2. The computer-implemented method of claim 1 , wherein the first resource feature vector comprises a topic vector, wherein each element of the topic vector comprises a score corresponding to a topic of a plurality of topics.
3. The computer-implemented method of claim 1 , wherein the neural network comprises a recurrent neural network with long short-term memory units.
4. The computer-implemented method of claim 1 , generating the first score for the first resource comprises processing the conversation feature vector and the first resource feature vector with a classifier.
5. The computer-implemented method of claim 4 , wherein the classifier comprises a multi-layer perceptron neural network.
6. The computer-implemented method of claim 1 , wherein the first user is a customer of a company and the second user is a customer service representative of the company.
7. The computer-implemented method of claim 1 , wherein generating the conversation feature vector comprises using an identity of the first user, an identity of the second user, or information about the first user.
8. The computer-implemented method of claim 1 , wherein the neural network was trained by processing a training corpus comprising conversations and resources relevant to the conversations.
9. A system for suggesting a resource to a second user in responding to a first user, the system comprising:
at least one server computer comprising at least one processor and at least one memory, the at least one server computer configured to:
obtain a plurality of resource feature vectors for a plurality of resources, wherein each resource feature vector is computed from text of a corresponding resource;
receive first text of a first message between the first user and the second user;
generate a first conversation feature vector by processing the first text of the first message with a neural network, wherein the neural network is configured to process text of a conversation to compute a conversation feature vector that is close to resource feature vectors of resources that are relevant to the conversation;
compute a plurality of scores, wherein in each score is computed by comparing the first conversation feature vector to a resource feature vector of the plurality of resource feature vectors, and wherein the plurality of scores comprises a first score computed for a first resource by comparing the conversation feature vector to a first resource feature vector for the first resource;
select the first resource from the plurality of resources using the plurality of scores; and
cause the first resource to be presented to the second user.
10. The system of claim 9 , wherein the at least one server computer is configured to generate the conversation feature vector using text of a plurality of messages between the first user and the second user.
11. The system of claim 9 , wherein the at least one server computer is configured to generate the first score for the first resource by computing a distance using the conversation feature vector and the first resource feature vector.
12. The system of claim 9 , wherein the at least one server computer is configured to select the first resource by selecting a resource with a highest score.
13. The system of claim 9 , wherein the at least one server computer is configured to cause the first resource to be presented to the second user by:
generating HTML using information about the first resource; and
inserting the HTML into a web page.
14. The system of claim 13 , wherein the presentation of the first resource includes a selectable item, and the at least one server computer is configured to:
receive a selection of the selectable item by the second user; and
cause a message to be sent to the first user in response to the selection of the selectable item.
15. The system of claim 9 , wherein the at least one server computer is configured to generate the first score for the first resource using an identity of the first user, an identity of the second user, or information about the first user.
16. The system of claim 9 , wherein the neural network comprises a recurrent neural network.
17. One or more non-transitory computer-readable media comprising computer executable instructions that, when executed, cause at least one processor to perform actions comprising:
obtaining a plurality of resource feature vectors for a plurality of resources, wherein each resource feature vector is computed from text of a corresponding resource;
receiving first text of a first message between a first user and a second user;
generating a first conversation feature vector by processing the first text of the first message with a neural network, wherein the neural network is configured to process text of a conversation to compute a conversation feature vector that is close to resource feature vectors of resources that are relevant to the conversation;
computing a plurality of scores, wherein in each score is computed by comparing the first conversation feature vector to a resource feature vector of the plurality of resource feature vectors, and wherein the plurality of scores comprises a first score computed for a first resource by comparing the conversation feature vector to a first resource feature vector for the first resource;
selecting the first resource from the plurality of resources using the plurality of scores; and
causing the first resource to be presented to the second user.
18. The one or more non-transitory computer-readable media of claim 17 , wherein the plurality of resource feature vectors are computed before receiving the text of the message between the first user and the second user.
19. The one or more non-transitory computer-readable media of claim 17 , wherein the first resource feature vector is computed using an autoencoder.
20. The one or more non-transitory computer-readable media of claim 17 , wherein computing the first score for the first resource comprises processing the conversation feature vector and the first resource feature vector with a classifier.
21. The one or more non-transitory computer-readable media of claim 20 , wherein the classifier comprises a multi-layer perceptron neural network.