Granted Patent B2

US 11,521,087 · App. 16/983,051 · Granted Dec 6, 2022

Inventors: Jiacheng Ni (Shanghai, CN); Zijia Wang (Shanghai, CN); Min Gong (Shanghai, CN); Pengfei Wu (Shanghai, CN); Zhen Jia (Shanghai, CN)

Assignee: EMC IP Holding Company LLC

G06N5/04G06F16/2379G06F40/279G06N20/00

Abstract

Claims (82)

1. A method for processing information, comprising:

acquiring a service request record set, each service request record in the service request record set relating to a problem encountered by a user when the user is provided with a service and a solution to the problem;

constructing a language model based on a first subset in the service request record set and an initial model, the initial model being trained using a predetermined corpus and configured to determine vector representations of words and sentences in the corpus; and

constructing a classification model based on a second subset in the service request record set and the language model, the classification model being capable of determining a solution to a pending problem, and the first subset being different from the second subset.

2. The method of claim 1 , wherein each service request record in the service request record set comprises:

an abstract of the problem comprising a plurality of words,

a detailed description of the problem comprising a plurality of words, and

an identification of the solution to the problem.

3. The method of claim 1 , wherein constructing the language model comprises:

determining the first subset from the service request record set, an identification of a solution in each service request record in the first subset being an invalid identification;

dividing, based on a generation time of each service request record in the first subset, the first subset into a first group of service request records configured to construct the language model and a second group of service request records configured to evaluate the language model; and

constructing the language model based on the initial model by using the first group of service request records.

4. The method of claim 3 , wherein constructing the language model based on the initial model comprises:

replacing at least one word in each service request record in the first group of service request records with at least one predetermined word to generate a first group of replaced service request records; and

constructing the language model by applying the first group of replaced service request records to the initial model.

5. The method of claim 3 , further comprising:

replacing at least one word in each service request record in the second group of service request records with at least one predetermined word to generate a second group of replaced service request records;

applying the second group of replaced service request records to the language model to determine at least one prediction result of the at least one word;

determining a probability that the at least one prediction result matches the at least one word; and

evaluating the language model based on the probability.

6. The method of claim 1 , wherein constructing the classification model comprises:

determining the second subset from the service request record set, an identification of a solution in each service request record in the second subset being a valid identification;

dividing, based on a generation time of each service request record in the second subset, the second subset into a third group of service request records configured to construct the classification model and a fourth group of service request records configured to evaluate the classification model; and

constructing the classification model by using the third group of service request records.

7. The method of claim 6 , further comprising:

applying the fourth group of service request records to the classification model to obtain a predicted solution;

determining a probability that the predicted solution matches the solution indicated by the identification; and

evaluating the classification model based on the probability.

8. An electronic device, comprising:

at least one processing unit; and

at least one memory coupled to the at least one processing unit and storing instructions configured to be executed by the at least one processing unit, wherein when executed by the at least one processing unit, the instructions cause the device to perform actions comprising:

acquiring a service request record set, each service request record in the service request record set relating to a problem encountered by a user when the user is provided with a service and a solution to the problem;

constructing a language model based on a first subset in the service request record set and an initial model, the initial model being trained using a predetermined corpus and configured to determine vector representations of words and sentences in the corpus; and

constructing a classification model based on a second subset in the service request record set and the language model, the classification model being capable of determining a solution to a pending problem, and the first subset being different from the second subset.

9. The device of claim 8 , wherein each service request record in the service request record set comprises:

an abstract of the problem comprising a plurality of words,

a detailed description of the problem comprising a plurality of words, and

an identification of the solution to the problem.

10. The device of claim 8 , wherein constructing the language model comprises:

determining the first subset from the service request record set, an identification of a solution in each service request record in the first subset being an invalid identification;

dividing, based on a generation time in each service request record in the first subset, the first subset into a first group of service request records configured to construct the language model and a second group of service request records configured to evaluate the language model; and

constructing the language model based on the initial model by using the first group of service request records.

11. The device of claim 10 , wherein constructing the language model based on the initial model comprises:

replacing at least one word in each service request record in the first group of service request records with at least one predetermined word to generate a first group of replaced service request records; and

constructing the language model by applying the first group of replaced service request records to the initial model.

12. The device of claim 10 , wherein the actions further comprise:

replacing at least one word in each service request record in the second group of service request records with at least one predetermined word to generate a second group of replaced service request records;

applying the second group of replaced service request records to the language model to determine at least one prediction result of the at least one word;

determining a probability that the at least one prediction result matches the at least one word; and

evaluating the language model based on the probability.

13. The device of claim 8 , wherein constructing the classification model comprises:

determining the second subset from the service request record set, an identification of a solution in each service request record in the second subset being a valid identification;

dividing, based on a generation time of each service request record in the second subset, the second subset into a third group of service request records configured to construct the classification model and a fourth group of service request records configured to evaluate the classification model; and

constructing the classification model by using the third group of service request records.

14. The device of claim 13 , wherein the actions further comprise:

applying the fourth group of service request records to the classification model to obtain a predicted solution;

determining a probability that the predicted solution matches the solution indicated by the identification; and

evaluating the classification model based on the probability.

15. A computer program product tangibly stored in a non-transitory computer-readable medium and comprising machine-executable instructions, wherein when executed, the machine-executable instructions cause a machine to perform steps of a method for processing information, the method comprising:

acquiring a service request record set, each service request record in the service request record set relating to a problem encountered by a user when the user is provided with a service and a solution to the problem;

constructing a language model based on a first subset in the service request record set and an initial model, the initial model being trained using a predetermined corpus and configured to determine vector representations of words and sentences in the corpus; and

constructing a classification model based on a second subset in the service request record set and the language model, the classification model being capable of determining a solution to a pending problem, and the first subset being different from the second subset.

16. The computer program product of claim 15 , wherein each service request record in the service request record set comprises:

an abstract of the problem comprising a plurality of words,

a detailed description of the problem comprising a plurality of words, and

an identification of the solution to the problem.

17. The computer program product of claim 15 , wherein constructing the language model comprises:

determining the first subset from the service request record set, an identification of a solution in each service request record in the first subset being an invalid identification;

dividing, based on a generation time of each service request record in the first subset, the first subset into a first group of service request records configured to construct the language model and a second group of service request records configured to evaluate the language model; and

constructing the language model based on the initial model by using the first group of service request records.

18. The computer program product of claim 17 , wherein constructing the language model based on the initial model comprises:

replacing at least one word in each service request record in the first group of service request records with at least one predetermined word to generate a first group of replaced service request records; and

constructing the language model by applying the first group of replaced service request records to the initial model.

19. The computer program product of claim 17 , further comprising:

replacing at least one word in each service request record in the second group of service request records with at least one predetermined word to generate a second group of replaced service request records;

applying the second group of replaced service request records to the language model to determine at least one prediction result of the at least one word;

determining a probability that the at least one prediction result matches the at least one word; and

evaluating the language model based on the probability.

20. The computer program product of claim 15 , wherein constructing the classification model comprises:

determining the second subset from the service request record set, an identification of a solution in each service request record in the second subset being a valid identification;

dividing, based on a generation time of each service request record in the second subset, the second subset into a third group of service request records configured to construct the classification model and a fourth group of service request records configured to evaluate the classification model; and

constructing the classification model by using the third group of service request records.

Assignments (9)

Priority Claims (1)

Continuity (1)