Claims:WE CLAIM:
1. A system for optimizing memory requirement for training an artificial neural network (ANN) model employed for natural language processing (NLP), the system comprising:
a memory optimization device comprising at least one processor and a computer-readable medium storing instructions that, when executed by the at least one processor, cause the at least one processor to perform operations comprising:
receiving a plurality of training parameters and a plurality of model parameters, wherein the plurality of training parameters are derived from training data corpus for training the ANN model, and wherein the plurality of model parameters are associated with the ANN model and are derived based on the plurality of training parameters and input from user with respect to the ANN model;
selecting a set of model parameters from among the plurality of model parameters for training the ANN model based on a characteristic and an architecture of the ANN model;
masking the set of model parameters in one or more layers of the ANN model based on a set of pre-defined rules to generate a set of masked model parameters;
determining an amount of memory required for training the ANN model based on the set of masked model parameters; and
providing the set of masked model parameters for training the ANN model when the amount of memory required is less than a determined threshold.

2. The system of claim 1, wherein the operations further comprise:
receiving the training data corpus and the input from the user;
deriving the plurality of training parameters by processing the training data corpus;
deriving the plurality of model parameters based on the plurality of training parameters and the input from the user;
temporarily storing the plurality of training parameters; and
deploying the plurality of model parameters.

3. The system of claim 1, wherein determining the amount of memory required comprises performing multi-linear regression with known values of the set of model parameters.

4. The system of claim 3, wherein determining the amount of memory required further comprises performing logistic regression when one or more of then known values are binary values.

5. The system of claim 1, wherein the plurality of model parameters comprise at least one of tokens, intents, named entities, word vectors, part of speech (PoS) tags, input features, input neurons, and output neurons.

6. The system of claim 5, wherein masking the set of model parameters using the set of pre-defined rules comprises at least one of stemming tokens, lemmatizing tokens, de-duplicating tokens, adjusting weights of tokens, converting tokens in hard coded binary numbers, and converting token types in hard coded binary numbers.

7. The system of claim 1, wherein masking the set of model parameters comprise masking the set of model parameters in one or more hidden layers of the ANN model.

8. The system of claim 1, wherein the operations further comprise:
training the ANN model with the set of masked model parameters; and
unmasking a resultant by back-propagating using the set of pre-defined rules.

9. The system of claim 1, wherein the operations further comprise:
iteratively selecting an updated set of model parameters, masking the updated set of model parameters, and determining the amount of memory required based on an updated set of masked model parameters until the amount of memory required is less than the determined threshold.

10. A method of optimizing memory requirement for training an artificial neural network (ANN) model employed for natural language processing (NLP), the method comprising:
receiving, by a memory optimization device, a plurality of training parameters and a plurality of model parameters, wherein the plurality of training parameters are derived from training data corpus for training the ANN model, and wherein the plurality of model parameters are associated with the ANN model and are derived based on the plurality of training parameters and input from user with respect to the ANN model;
selecting, by the memory optimization device, a set of model parameters from among the plurality of model parameters for training the ANN model based on a characteristic and an architecture of the ANN model;
masking, by the memory optimization device, the set of model parameters in one or more layers of the ANN model based on a set of pre-defined rules to generate a set of masked model parameters;
determining, by the memory optimization device, an amount of memory required for training the ANN model based on the set of masked model parameters; and
providing, by the memory optimization device, the set of masked model parameters for training the ANN model when the amount of memory required is less than a determined threshold.

Dated 15th day of February, 2019

R Ramya Rao
Of K&S Partners
Agent for the Applicant
IN/PA-1607 , Description:TECHNICAL FIELD
This disclosure relates generally to natural language processing (NLP), and more particularly to method and system for optimizing memory requirement for training an artificial neural network (ANN) model employed for NLP.