TASK DETECTION IN COMMUNICATIONS USING DOMAIN ADAPTATION

• US 20200134442A1
• Filed: 10/29/2018
• Published: 04/30/2020
• Est. Priority Date: 10/29/2018
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Sim, Robert A., White, Ryen W., Azarbonyad, Hosein
• Granted Patent: US 11,704,552 B2
• Time in Patent Office: Days
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06Q 10/06311 Scheduling, planning or task assignment for a person or group
• CPC Class Codes: H04L 51/224 providing notification on incoming messages, e.g. pushed notifications of received messages

Abstract

Generally discussed herein are devices, systems, and methods for task classification. A method can include modifying a representation of a source sentence of a source sample from a source corpus to more closely resemble a representation of target sentences of target samples from a target corpus, operating, using a machine learning model trained using the modified representation of the source sentence, with the target sample to generate a task label, the task label indicating whether the target sample includes a task. and causing a personal information manager (PIM) to generate a reminder, based on whether the target sample includes the task.

Claims

1. A system comprising:

a memory;

processing circuitry coupled to the memory to implement operations for task classification on a target sample from a target corpus, the task classification including;

modifying a representation of a source sentence of a source sample from a source corpus to more closely resemble target sentences of target samples from a target corpus;

operating, using a machine learning model trained using the modified representation of the source sentence, with the target sample to generate a task label, the task label indicating whether the target sample includes a task; and

causing a personal information manager (PIM) to generate a reminder, based on whether the target sample includes the task.

2. The system of claim 1, wherein modifying the representation of the source sentence of the source sample includes performing feature adaptation to change features of the source samples to one or more specified values.

3. The system of claim 2, wherein the feature adaptation includes:

training a corpus classifier machine learning model to discriminate between the source samples and the target samples;

selecting top K features from the corpus classifier machine teaming model; and

replacing each of the top K features in the target samples with the specified value.

4. The system of claim 2, wherein modifying the representation of e source sentence of the source sample includes:

identifying a mapping between features of the source sample and features of the target samples based on a distance between words of the source sample and respective mapped target words of the target samples; and

replacing a vector representation of a respective word of the source sample with a mapped vector of the respective mapped target word.

5. The system of claim 1, wherein modifying the representation of the source sentence of the source sample includes performing importance sampling to change weights of the source samples to importance weights.

6. The system of claim 5, wherein the importance weights are determined based on a probability of a marginal distribution of a sample classified as being from the source corpus and a probability of a marginal distribution of the sample classified as being from the target corpus.

7. The system of claim 1, wherein the machine learning model includes a neural network (NN) autoencoder and wherein the operations further comprise modifying a representation of a source sentence of a source sample from a source corpus to include a representation of target sentences of the target samples from the target corpus includes adversarial training the NN to increase a corpus classification loss while reducing a task classification loss and a sample reconstruction loss.

8. The system of claim 7, wherein the NN autoencoder includes a sample representation NN that:

implements a sequence-to-sequence encoder that reads an embedded representation of the target sample in a first direction, creates a first sequence of hidden states, reads the embedded representation of the target sample in a second direction, opposite the first direction, and creates a second sequence of hidden states.

9. The system of claim 8, wherein the sample representation NN provides a concatenation of a last hidden state of reading the embedded representation in the first direction and a first hidden state from reading the embedded representation in the second direction.

10. The system of claim 1, wherein the task classification further includes training the machine learning model using the modified representation of the source sentence.

11. The system of claim 10, wherein the task classification further includes receiving feedback data from the PIM and further training the model based on the feedback data, wherein the feedback data indicates whether the target sample includes the task based on a user input to an interface of the PIM.

12. The system of claim 1, wherein the reminder includes (i) an apppointment, (ii) a meeting, (iii) a notification, (iv) a text message, (v) an email, or (vi) an instant message.

13. At least one machine-readable storage medium including instructions for execution by processing circuitry to perform operations for task detection, the operations comprising:

modifying a representation of a source sentence of a source sample from a source corpus to more closely resemble a representation of target sentences of target samples from a target corpus;

operating, using a machine teaming model trained using the modified representation of the source sentence, with the target sample to generate a task label, the task label indicating whether the target sample includes a task; and

causing a personal information manager (PIM) to generate a reminder, based on whether the target sample includes the task.

14. The at least one machine-readable storage medium of claim 13, wherein modifying the representation of the source sentence of the source sample includes performing feature adaptation to change features of the source samples to one or more specified values, importance sampling to change weights of the source samples to importance weights, or using a neural network (NN) autoencoder adversarial trained to increase a corpus classification loss while reducing a task classification loss and a sample reconstruction loss.

15. The at least one machine-readable storage medium of claim 14, wherein the feature adaptation includes:

training a corpus classifier machine learning model to discriminate between the source samples and the target samples;

selecting top K features from the corpus classifier machine learning model; and

replacing each of the top K features in the target samples with the specified value.

16. The at least one machine-readable storage medium of claim 14, wherein modifying the representation of the source sentence of the source sample includes:

identifying a mapping between features of the source sample and features of the target samples based on a distance between words of the source sample and respective mapped vectors; and

replacing the representation of the word of the source sample with a mapped vector of the respective mapped vectors.

17. The at least one machine-readable storage medium of claim 14, wherein the importance weights are determined based on a probability of a marginal distribution of a sample classified as being from the source corpus and a probability of a marginal distribution of the sample classified as being from the target corpus.

18. A method, performed by at least one processor of a computing system, the method comprising:

modifying a representation of a source sentence of a source sample from a source corpus to more closely resemble a representation of target sentences of target samples from a target corpus;

operating, using a machine learning model trained using the modified representation of the source sentence, with the target sample to generate a task label, the task label indicating whether the target sample includes a task; and

causing a personal information manager (PIM) to generate a reminder, based on whether the target sample includes the task.

19. The method of claim 18, wherein the task classification further includes training the machine learning model using the modified representation of the source sentence.

20. The method of claim 19, further comprising receiving feedback data from the PIM and further training the model based on the feedback data, wherein the feedback data indicates whether the target sample includes the task based on a user input to an interface of the PIM.