Systems and methods to implement neural networks that have transformed to computable functions to enhance coordinated natural language processing by server-client systems

1 . A system configured to implement neural networks that have been transformed to computable functions to enhance coordinated natural language processing by server/client systems, the system comprising:

electronic storage that stores a trained neural network model and a set of output values that correspond to outcomes, wherein the trained neural network model generates an output value based on inputs,

wherein the trained neural network model includes individual sets of nodes, wherein the individual sets of the nodes are configured to sequentially perform different analyses to produce separate node outputs, wherein the node outputs produced by the sets of the nodes are input to the sets of the nodes that follow;

wherein the nodes include values for node parameters and function parameters, wherein the node parameters include at least weights and a bias, wherein the function parameters include operations related to the weights and the bias, and/or one or more coefficients;

a first set of one or more processors configured by machine-readable instructions to:

analyze the trained neural network model to determine node values for the node parameters applied to the inputs at individual ones of the nodes in the sets;

analyze the trained neural network model to determine function values for the function parameters applied to the node values at the individual nodes;

generate, based on the node values and the function values, individual intermediate representations of the individual nodes, wherein the intermediate representations impartially represent the nodes and the particular analyses performed;

generate, based on the individual intermediate representations, sets of instructions in different programming languages for the individual nodes such that the sets of instructions are executable on a client computing platform, wherein the sets of instructions include a first set of instructions in a first programming language;

the client computing platform including:

local electronic storage that stores the first set of instructions and the set of output values that correspond to the outcomes; and

a second set of one or more processors configured by machine-readable instructions to:

obtain audio information that represents sounds conveying utterances by a user;

perform analysis on the audio information to determine whether the audio information satisfies criteria for the second set of one or more processors to execute the first set of instructions on the audio information;

execute the first set of instructions on the audio information to generate an instruction output responsive to determination that the audio information satisfies the criteria;

determine whether the instruction output is equal to one of the output values stored in the local electronic storage; and

effectuate the corresponding outcome responsive to determination that the instruction output equals one of the output values.

2 . The system of claim 1 , wherein the electronic storage stores trained neural network models that generate output values based on different inputs, wherein the trained neural network models include the trained neural network model.

3 . The system of claim 2 , wherein the different inputs include an image, a video, and/or entity classifications.

4 . The system of claim 1 , wherein the outcomes include command intake and individual commands.

5 . The system of claim 1 , wherein the instruction output includes a number, an array, and/or a multidimensional array.

6 . The system of claim 2 , wherein the electronic storage stores the sets of instructions in the programming languages for the individual nodes of the trained neural network model as a first collection, wherein the electronic storage stores multiple collections for the nodes of different ones of the trained neural network models.

7 . The system of claim 1 , wherein the electronic storage stores a list of wake words and a set of commands, wherein the criteria include utterance of one or more of the wake words and/or utterance of one or more of the commands.

8 . The system of claim 1 , wherein the second set of the one or more processors are further configured by the machine-readable instructions to:

responsive to determination that the instruction output does not equal one of the output values, transmit the audio information to a server for execution by the trained neural network model.

9 . A system configured to implement neural networks that have been transformed to computable functions to enhance coordinated natural language processing by server/client systems, the system comprising:

a client computing platform including:

local electronic storage that stores a first set of instructions and a set of output values that correspond to outcomes, wherein the first set of instructions are a computable form of a trained neural network model,

wherein the trained neural network model includes individual sets of nodes, wherein the individual sets of the nodes are configured to sequentially perform different analyses to produce separate node outputs, wherein the node outputs produced by the sets of the nodes are input to the sets of the nodes that follow;

wherein the nodes include values for node parameters and function parameters, wherein the node parameters include at least weights and a bias, wherein the function parameters include operations related to the weights and the bias, and/or one or more coefficients;

wherein the first set of instructions are generated based on individual intermediate representations of the individual nodes, wherein the intermediate representations are generated based on the values for node parameters and function parameters applied to the given input at the individual nodes so that the intermediate representations impartially represent the nodes and the particular analyses performed; and

a second set of one or more processors configured by machine-readable instructions to:

obtain audio information that represents sounds conveying utterances by a user;

perform analysis on the audio information to determine whether the audio information satisfies criteria for the second set of one or more processors to execute the first set of instructions on the audio information;

execute the first set of instructions on the audio information to generate an instruction output responsive to determination that the audio information satisfies the criteria;

determine whether the instruction output is equal to one of the output values stored in the local electronic storage; and

effectuate the corresponding outcome responsive to determination that the instruction output equals one of the output values.

10 . A method to implement neural networks that have been transformed to computable functions to enhanced coordinated natural language processing by server/client systems, the method comprising:

analyzing, by a first set of one or more processors, a trained neural network model to determine node values for node parameters applied to given inputs at individual nodes in sets of nodes, wherein electronic storage that stores the trained neural network model and a set of output values that correspond to outcomes, wherein the trained neural network model generates an output value based on the inputs,

wherein the trained neural network model includes individual sets of the nodes, wherein the individual sets of the nodes are configured to sequentially perform different analyses to produce separate node outputs, wherein the node outputs produced by the sets of the nodes are input to the sets of the nodes that follow,

wherein the nodes include values for node parameters and function parameters, wherein the node parameters include at least weights and a bias, wherein the function parameters include operations related to the weights and the bias, and/or one or more coefficients;

analyzing, by the first set of one or more processors, the trained neural network model to determine function values for the function parameters applied to the node values at the individual nodes;

generating, by the first set of one or more processors, based on the node values and the function values, individual intermediate representations of the individual nodes, wherein the intermediate representations impartially represent the nodes and the particular analyses performed;

generating, by the first set of one or more processors, sets of instructions in different programming languages for the individual nodes based on the individual intermediate representations such that the sets of instructions are executable on a client computing platform, wherein the sets of instructions include a first set of instructions in a first programming language;

obtaining, by a second set of one or more processors included in the client computing platform, audio information that represents sounds conveying utterances by a user;

performing, by the second set of one or more processors, analysis on the audio information to determine whether the audio information satisfies criteria for the second set of one or more processors to execute the first set of instructions on the audio information;

executing, by the second set of one or more processors, the first set of instructions on the audio information to generate an instruction output responsive to determination that the audio information satisfies the criteria, wherein local electronic storage included in the client computing platform stores the first set of instructions and the set of output values that correspond to the outcomes;

determining, by the second set of one or more processors, whether the instruction output is equal to one of the output values stored in the local electronic storage; and

effectuating, by the second set of one or more processors, the corresponding outcome responsive to determination that the instruction output equals one of the output values.

11 . The method of claim 10 , wherein the electronic storage stores trained neural network models that generate output values based on different inputs, wherein the trained neural network models include the trained neural network model.

12 . The method of claim 11 , wherein the different inputs include an image, a video, and/or entity classifications.

13 . The method of claim 10 , wherein the outcomes include command intake and individual commands.

14 . The method of claim 10 , wherein the instruction output includes a number, an array, and/or a multidimensional array.

15 . The method of claim 11 , wherein the electronic storage stores the sets of instructions in the programming languages for the individual nodes of the trained neural network model as a first collection, wherein the electronic storage stores multiple collections for the nodes of different ones of the trained neural network models.

16 . The method of claim 10 , wherein the electronic storage stores a list of wake words and a set of commands, wherein the criteria include utterance of one or more of the wake words and/or utterance of one or more of the commands.

17 . The method of claim 10 , responsive to determination that the instruction output does not equal one of the output values, transmitting, by the second set of the one or more processors, the audio information to a server for execution by the trained neural network model.