Methods and apparatus for generating intelligible data segments with large language models

1 . A non-transitory, processor-readable medium storing instructions that, when executed by a processor, cause the processor to:

receive, from a user compute device, a signal representative of a selection of at least one of a first data set or a plurality of second data sets

input, without human intervention, the first data set to a machine learning model to generate a plurality of clusters in the first data set, each cluster from the plurality of clusters associated with a consumer type;

determine, without human intervention, a plurality of statistically significant variables in a cluster from the plurality of clusters based on statistical significance tests including variables in at least one of the first data set or the plurality of second data sets;

generate, without human intervention, a set of pseudosentences associated with the cluster based on an in-group value and an out-group value for each statistically significant variable from the plurality of statistically significant variables;

send, without human intervention, a signal to cause the set of pseudosentences to be inserted into a first no-code prompt for a large language model to predict a segment description for the consumer type associated with the cluster, the first no-code prompt having a code-style to reduce a variance of the segment description relative to an output from the large language model based on a second no-code prompt without the code-style, the segment description being a representation of relationships among each statistically significant variable from the plurality of statistically significant variables based on the set of pseudosentences; and

send, to the user compute device, data representing the segment description.

2 . The non-transitory, processor-readable medium of claim 1 , wherein:

the first data set is any one of an attitudinal data set or a behavioral data set, and

the plurality of second data sets (1) includes at least one of the attitudinal data set or the behavioral data set, and (2) does not include the first data set.

3 . The non-transitory, processor-readable medium of claim 1 , wherein the segment description is a first segment description, the set of pseudosentences is a first set of pseudosentences, the plurality of statistically significant variables is a first plurality of statistically significant variables, the non-transitory, processor-readable medium storing further instructions that cause the processor further to:

receive, from the user compute device, a signal indicative of a request associated with an evaluation of the segment description;

generate, without human intervention, a second set of pseudosentences associated with the cluster based on an in-group value and an out-group value for each statistically significant variable from a second plurality of statistically significant variables;

send, without human intervention, a signal to cause the second set of pseudosentences to be inserted into a third no-code prompt for the large language model to predict a second segment description for the consumer type associated with the cluster; and

send, to the user compute device, data representing the second segment description.

4 . The non-transitory, processor-readable medium of claim 1 , wherein the segment description is a first segment description, the non-transitory, processor-readable medium storing further instructions that cause the processor further to:

update, in response to receiving input text from the user compute device, the first segment description with the input text to produce a second segment description; and

send, to the user compute device, data representing the second segment description.

5 . The non-transitory, processor-readable medium of claim 1 , wherein:

the in-group value is an in-group mean value, the out-group value is an out-group mean value, and

the set of pseudosentences are defined by a difference of the in-group mean value and the out-group mean value meeting a threshold value and a direction represented by a sign of the difference.

6 . The non-transitory, processor-readable medium of claim 1 , wherein:

each statistically significant variable from the plurality of statistically significant variables in the first data set is a compressed representation of at least one variable from a third data set,

the third data set being input to a compression model to produce the first data set.

7 . The non-transitory, processor-readable medium of claim 1 , wherein the machine learning model is a first machine learning model, the non-transitory, processor-readable medium storing further instructions that cause the processor further to:

produce a training data set based on the plurality of statistically significant variables, each statistically significant variable having a label associated with the segment description;

train a second machine learning model based on the training data set; and

predict a plurality of labels for a plurality of variables in a third data set based on inputting a third data set into the second machine learning model.

8 . A method, comprising:

determining, using a processor of a first compute device and without human intervention, a plurality of statistically significant variables in a cluster associated with a first data set based on a statistical significance test including variables in at least one of the first data set or a plurality of second data sets, the cluster being associated with a consumer type;

generating, using the processor and without human intervention, a set of pseudosentences associated with the cluster based on a high value and a low value for each statistically significant variable from the plurality of statistically significant variables;

sending, without human intervention, a signal to cause the set of pseudosentences to be inserted into a first no-code prompt for a large language model to predict a segment description for the consumer type associated with the cluster, the first no-code prompt having a code-style to reduce a variance of the segment description relative to output from the large language model based on a second no-code prompt without the code-style, the segment description being a representation of relationships among each statistically significant variable from the plurality of statistically significant variables based on the set of pseudosentences; and

sending, from the first compute device and to a second compute device, data representing the segment description.

9 . The method of claim 8 , wherein:

the high value for each statistically significant variable is a first mean value in at least one of the first data set or the plurality of second data sets,

the low value for each statistically significant variable is a second mean value in at least one of the first data set or the plurality of second data sets,

the set of pseudosentences are defined by a difference of the high value and the low value meeting a threshold value and a direction represented by a sign of the difference.

10 . The method of claim 8 , wherein:

the first data set is an attitudinal data set,

the plurality of statistically significant variables includes at least one of one-hot encoded variables or scaling variables, and

the plurality of second data sets includes a behavioral data set.

11 . The method of claim 8 , wherein:

the first data set is a behavioral data set defined in part by a target category,

the plurality of statistically significant variables are numerical variables, and

the plurality of second data sets includes an attitudinal data set.

12 . The method of claim 8 , wherein the segment description is a first segment description, the method further comprising:

receiving, from the second compute device, a signal indicative of a request associated with an evaluation of the first segment description;

generating, without human intervention, a second set of pseudosentences associated with the cluster based on a high value and a low value for each statistically significant variable from a second plurality of statistically significant variables;

sending, without human intervention, a signal to cause the second set of pseudosentences to be inserted into a third no-code prompt for the large language model to predict a second segment description for the consumer type associated with the cluster; and

sending, from the first compute device and to the second compute device, data including the second segment description.

13 . The method of claim 8 , wherein each statistically significant variable from the plurality of statistically significant variables in the first data set is a compressed representation of at least one variable from a third data set, the third data set being input to a compression model to produce the first data set.

14 . The method of claim 8 , wherein the segment description is a first segment description, the method further comprising:

updating, in response to receiving input text from the second compute device, the first segment description with the input text to produce a second segment description; and

sending, from the first compute device and to the second compute device, data representing the second segment description.

15 . The method of claim 8 , further comprising:

sending, from the first compute device and to the second compute device before determining the plurality of statistically significant variables, a signal to cause a display of the second compute device to display a plot including the first data set;

receiving, from the second compute device, a signal indicative of a request associated with an evaluation of the plot; and

inputting, using the processor, the first data set to a machine learning model to generate a plurality of clusters that is in the first data set and that includes the cluster.

16 . A non-transitory, processor-readable medium, storing instructions that, when executed by a processor, cause the processor to:

send, to a user compute device, a signal to cause a display at the user compute device to display a plot representing a first plurality of clusters in a first data set, each cluster from the first plurality of clusters being associated with a consumer type;

receive, from the user compute device, a signal indicative of a request associated with an evaluation of the plot;

input, without human intervention, the first data set to a machine learning model to generate a second plurality of clusters in the first data set;

determine, without human intervention, a plurality of statistically significant variables in a cluster from the second plurality of clusters based on a statistical significance test for variables in at least one of the first data set or a plurality of second data sets;

generate, without human intervention, a set of pseudosentences associated with the cluster from the second plurality of clusters based on an in-group value and an out-group value for each statistically significant variable from the plurality of statistically significant variables;

send, without human intervention, a signal to cause the set of pseudosentences to be inserted into a first no-code prompt for a large language model to predict a segment description for the consumer type associated with the cluster from the second plurality of clusters, the first no-code prompt having a code-style to reduce a variance of the segment description relative to output from the large language model based on a second no-code prompt without the code-style, the segment description being a representation of relationships among each statistically significant variable from the plurality of statistically significant variables based on the set of pseudosentences; and

send, to the user compute device, data representing the segment description for the cluster from the second plurality of clusters.

17 . The non-transitory, processor-readable medium of claim 16 , wherein the first data set is any one of an attitudinal data set or a behavioral data set, and the plurality of second data sets (1) includes at least one of the behavioral data set or the attitudinal data set and (2) does not include the first data set.

18 . The non-transitory, processor-readable medium of claim 16 , wherein the machine learning model is a first machine learning model, the non-transitory, processor-readable medium storing further instructions that cause the processor further to:

produce a training data set based on the plurality of statistically significant variables, each statistically significant variable having a label associated with the segment description;

train a second machine learning model based on the training data set; and

predict a plurality of labels for a plurality of variables in a third data set based on inputting the third data set into the second machine learning model.

19 . The non-transitory, processor-readable medium of claim 16 , wherein the segment description is a first segment description, the set of pseudosentences is a first set of pseudosentences, the plurality of statistically significant variables is a first plurality of statistically significant variables, the non-transitory, processor-readable medium storing further instructions that cause the processor further to:

receive, from the user compute device, a signal indicative of a request associated with an evaluation of the first segment description;

generate, without human intervention, a second set of pseudosentences associated with the cluster based on an in-group value and an out-group value for each statistically significant variable from a second plurality of statistically significant variables different from the first plurality of statistically significant variables;

send, without human intervention, a signal to cause the second set of pseudosentences to be inserted into a third no-code prompt for the large language model to predict a second segment description for the consumer type associated with the cluster; and

send, to the user compute device, data representing the second segment description.

20 . The non-transitory, processor-readable medium of claim 16 , wherein the segment description is a first segment description, the non-transitory, processor-readable medium storing further instructions that cause the processor further to:

update, in response to receiving input text from the user compute device, the first segment description with the input text to produce a second segment description for the consumer type associated with the cluster; and

send, to the user compute device, data representing the second segment description.