Document-based response generation system

1. A system for customized document-based response generation to a user query, the system comprising:

a memory;

at least one processor communicatively coupled to the memory;

a filter component, implemented on the at least one processor, identifies a set of filtered utterances from a set of generated utterances having a same intent as an identified intent associated with a user-provided utterance;

a calculation component, implemented on the at least one processor, computes a distance between the user-provided utterance and each utterance in the set of filtered utterances;

a scoring component, implemented on the at least one processor, generates a similarity score for each utterance in the set of filtered utterances based on the distance for each utterance;

a selection component, implemented on the at least one processor, selects a set of utterances from the set of filtered utterances having the similarity score exceeding a threshold value;

a response generation component, implemented on the at least one processor, generates a weighted summarization response based on a predefined answer to each utterance in a selected set of utterances; and

a user interface device outputs the weighted summarization response to a user associated with the user-provided utterance.

2. The system of claim 1 , further comprising:

an utterance generator, implemented on the at least one processor, analyzes a text associated with a set of sentences in a set of policy documents to create the set of generated utterances.

3. The system of claim 1 , further comprising:

an input analysis component, implemented on the at least one processor, analyzes the user-provided utterance and identifies an intent from a set of pre-defined intents associated with an entity of the user-provided utterance.

4. The system of claim 1 , further comprising:

a document creation component, implemented on the at least one processor, generates a frequently asked questions (FAQ) document based on the set of generated utterances and a pre-generated response to each utterance in the set of generated utterances.

5. The system of claim 1 , further comprising:

a feedback component, implemented on the at least one processor, requests feedback from the user regarding accuracy of the weighted summarization response relative to the user-provided utterance.

6. The system of claim 1 , further comprising:

a machine learning component, implemented on the at least one processor, analyzes feedback received from at least one user with regard to accuracy of the weighted summarization response with regard to the user-provided utterance and updates a set of response weights based on the feedback.

7. The system of claim 1 , further comprising:

an assignment component, implemented on the at least one processor, assigns an intent from a set of pre-defined intents to each utterance in the set of generated utterances based on a set of keywords in each utterance.

8. The system of claim 1 , further comprising:

a training component, implemented on the at least one processor, inputs a set of training utterances and corresponding answers and a predefined set of possible intents into an utterance generation component to train the component in generating utterances based on text in at least one policy document.

9. A computer-implemented method for generating responses to user provided queries, the computer-implemented method comprising:

creating, by an utterance generator, a set of generated utterances based on text associated with a set of sentences in at least one document, each utterance in the set of generated utterances assigned an intent;

filtering, by a filter component, at least one utterance from the set of generated utterances having a same intent as a user-provided utterance to form a set of filtered utterances, the user-provided utterance received from a user device associated with the user via a network;

selecting, by a selection component, a set of utterances from the set of filtered utterances having a similarity score exceeding a threshold value, the similarity score indicates a degree of similarity between the generated utterance and the user-provided utterance;

generating, by a response generation component, a weighted summarization response based on a predefined answer to each utterance in a selected set of utterances; and

outputting, via the user device, the weighted summarization response to the user associated with the user-provided utterance.

10. The computer-implemented method of claim 9 , further comprising:

generating, by a calculation component, a distance between the user-provided utterance and each utterance in the set of filtered utterances.

11. The computer-implemented method of claim 9 , further comprising:

analyzing, by a machine learning component, feedback received from at least one user with regard to accuracy of the weighted summarization response with regard to the user-provided utterance and updates a set of response weights based on the feedback.

12. The computer-implemented method of claim 9 , further comprising:

training a machine learning component via a set of training utterances, a response corresponding to each utterance in the set of training utterance and a predefined set of possible intents.

13. The computer-implemented method of claim 9 , further comprising:

assigning, by an assignment component, a selected intent from a set of pre-defined intents to each utterance in the set of generated utterances based on a set of keywords in each utterance.

14. The computer-implemented method of claim 9 , further comprising:

generating, by a document creation component, a FAQ document based on the set of generated utterances and a pre-generated response to each utterance in the set of generated utterances.

15. The computer-implemented method of claim 9 , further comprising:

assigning, by a scoring component, the similarity score for each utterance in the set of filtered utterances based on a distance for each utterance.

16. The computer-implemented method of claim 9 , further comprising:

generating a set of variation utterances associated with a selected utterance from the set of generated utterances, wherein the set of variation utterances comprises at least one similar question that is a variation of the selected utterance; and

adding the set of variation utterances to the set of generated utterances, wherein each utterance is the set of variation utterances is mapped to a same answer for the selected utterance.

17. One or more computer storage devices, having computer-executable instructions for a dynamic response generation system for generating customized responses to user queries that, when executed by a computer cause the computer to perform operations comprising:

filtering, by a filter component, at least one utterance from a set of generated utterances having a same intent as a user-provided utterance to form a set of filtered utterances, the user-provided utterance received from a user device associated with the user via a network;

selecting, by a selection component, a set of utterances from the set of filtered utterances having a similarity score exceeding a threshold value, the similarity score indicating a degree of similarity between the generated utterance and the user-provided utterance;

generating, by a response generation component, a weighted summarization response based on a predefined answer to each utterance in a selected set of utterances and a set of weights;

outputting, via the user device, the weighted summarization response to the user associated with the user-provided utterance; and

updating the set of weights based on feedback received from at least one user rating the weighted summarization response relative to the user-provided utterance.

18. The one or more computer storage devices of claim 17 , wherein the response generation component when further executed by a computer cause the computer to perform operations comprising:

creating the set of generated utterances based on text associated with a set of sentences in at least one document, each utterance in the set of generated utterances assigned an intent.

19. The one or more computer storage devices of claim 17 , wherein the response generation component when further executed by a computer cause the computer to perform operations comprising:

generating a FAQ document based on the set of generated utterances and a pre-generated response to each utterance in the set of generated utterances.

20. The one or more computer storage devices of claim 17 , wherein the response generation component when further executed by a computer cause the computer to perform operations comprising:

a calculation component, implemented on at least one processor, computes a distance between the user-provided utterance and each utterance in the set of filtered utterances; and

a scoring component, implemented on the at least one processor, generates the similarity score for each utterance in the set of filtered utterances based on the distance for each utterance.