Adaptive training data augmentation to facilitate training named entity recognition models
Techniques are disclosed herein for adaptive training data augmentation to facilitate training named entity recognition (NER) models. Adaptive augmentation techniques are disclosed herein that take into consideration the distribution of different entity types within training data. The adaptive augmentation techniques generate adaptive numbers of augmented examples (e.g., utterances) based on the distribution of entities to make sure enough numbers of examples for minority class entities are generated during augmentation of the training data.
1 . A computer-implemented method comprising:
receiving, by a chatbot system, an utterance, wherein the utterance comprises at least one named entity:
processing the utterance using a trained named entity recognition (NER) model of the chatbot system to extract one or more named entities from the utterance, wherein the trained NER model is configured to identify named entities in utterances, and wherein the trained NER model is trained by:
accessing a machine learning model;
accessing training data comprising a plurality of utterances, wherein each utterance of the plurality of utterances comprises at least one named entity corresponding to a named entity category of a plurality of named entity categories;
calculating an augmentation rate for each named entity category of the plurality of named entity categories based on a distribution of each named entity category of the plurality of named entity categories in the training data and a function selected from among a plurality of functions in a hyperparameter tuning process of a named entity recognition (NER) model, wherein the plurality of functions comprises at least one of a first function that calculates an augmentation rate for a given named entity category based on an exponential function that receives a normalized distribution value for the given named entity category and a beta value as an input and a second function that calculates an augmentation rate for the given named entity category based on a binary logarithm function that receives a number of occurrences that the given named entity category occurs in the training data as an input;
generating a plurality of additional utterances based on the augmentation rate calculated for each named entity category of the plurality of named entity categories;
augmenting the training data to form augmented training data by combining the plurality of additional utterances with the plurality of utterances; and
training the machine learning model with the augmented training data to result in the trained NER model; and
generating, by the chatbot system, a response to the utterance based on the one or more named entities.
2 . The computer-implemented method of claim 1 , wherein the hyperparameter tuning process comprises:
dividing the training data into sets of training data, each set of training data of the sets of training data comprising a corresponding set of utterances of the plurality of utterances;
for each set of training data of the sets of training data:
selecting a candidate hyperparameter from among a set of candidate hyperparameters,
generating, based on the candidate hyperparameter, a set of augmented utterances from the corresponding set of utterances for the respective set of training data based on a function of the plurality of functions associated with the candidate hyperparameter,
training the NER model with the set of augmented utterances, and
calculating a performance metric for the NER model trained with the set of augmented utterances; and
determining a set of hyperparameters for the NER model based on the performance metric calculated for the NER model trained with the respective sets of augmented utterances.
3 . The computer-implemented method of claim 1 , wherein a number of additional utterances generated for a first named entity category of the plurality of named entity categories is different than a number of additional utterances generated for a second named entity category of the plurality of named entity categories.
4 . The computer-implemented method of claim 1 , wherein determining the augmentation rate for each named entity category of the plurality of named entity categories comprises calculating a number of occurrences of each named entity category of the plurality of named entity categories in the training data and normalizing the number of occurrences of each named entity category of the plurality of named entity categories in the training data.
5 . The computer-implemented method of claim 4 , wherein determining the augmentation rate for each named entity category of the plurality of named entity categories comprises calculating a normalized distribution value for each named entity category of the plurality of named entity categories, wherein the normalized distribution value is calculated for each named entity category of the plurality of named entity categories by dividing the number of occurrences calculated for a respective named entity category by a sum of the number of occurrences of all of the named entity categories of the plurality of named entity categories occurring in the training data.
6 . The computer-implemented method of claim 1 , further comprising:
prior to generating the response, classifying an intent associated with the utterance based on the one or more named entities.
7 . A chatbot system comprising:
one or more processors; and
one or more computer-readable media storing instructions which, when executed by the one or more processors, cause the chatbot system to perform operations comprising:
receiving an utterance, wherein the utterance comprises at least one named entity;
processing the utterance using a trained named entity recognition (NER) model to extract one or more named entities from the utterance, wherein the trained NER model is configured to identify named entities in utterances, and wherein the trained NER model is trained by:
accessing a machine learning model;
accessing training data comprising a plurality of utterances, wherein each utterance of the plurality of utterances comprises at least one named entity corresponding to a named entity category of a plurality of named entity categories;
calculating an augmentation rate for each named entity category of the plurality of named entity categories based on a distribution of each named entity category of the plurality of named entity categories in the training data and a function selected from among a plurality of functions in a hyperparameter tuning process of a named entity recognition (NER) model, wherein the plurality of functions includes at least one of a first function that calculates an augmentation rate for a given named entity category based on an exponential function that receives a normalized distribution value for the given named entity category and a beta value as an input and a second function that calculates an augmentation rate for the given named entity category based on a binary logarithm function that receives a number of occurrences that the given named entity category occurs in the training data as an input;
generating a plurality of additional utterances based on the augmentation rate calculated for each named entity category of the plurality of named entity categories;
augmenting the training data to form augmented training data by combining the plurality of additional utterances with the plurality of utterances; and
training the machine learning model with the augmented training data to result in the trained NER model; and
generating a response to the utterance based on the one or more named entities.
8 . The chatbot system of claim 7 , wherein the hyperparameter tuning process comprises:
dividing the training data into sets of training data, each set of training data of the sets of training data comprising a corresponding set of utterances of the plurality of utterances;
for each set of training data of the sets of training data:
selecting a candidate hyperparameter from among a set of candidate hyperparameters,
generating, based on the candidate hyperparameter, a set of augmented utterances from the corresponding set of utterances for the respective set of training data based on a function of the plurality of functions associated with the candidate hyperparameter,
training the NER model with the set of augmented utterances, and
calculating a performance metric for the NER model trained with the set of augmented utterances; and
determining a set of hyperparameters for the NER model based on the performance metric calculated for the NER model trained with the respective sets of augmented utterances.
9 . The chatbot system of claim 7 , wherein a number of additional utterances generated for a first named entity category of the plurality of named entity categories is different than a number of additional utterances generated for a second named entity category of the plurality of named entity categories.
10 . The chatbot system of claim 7 , wherein determining the augmentation rate for each named entity category of the plurality of named entity categories comprises calculating a number of occurrences of each named entity category of the plurality of named entity categories in the training data and normalizing the number of occurrences of each named entity category of the plurality of named entity categories in the training data.
11 . The chatbot system of claim 10 , wherein determining the augmentation rate for each named entity category of the plurality of named entity categories comprises calculating a normalized distribution value for each named entity category of the plurality of named entity categories, wherein the normalized distribution value is calculated for each named entity category of the plurality of named entity categories by dividing the number of occurrences calculated for a respective named entity category by a sum of the number of occurrences of all of the named entity categories of the plurality of named entity categories occurring in the training data.
12 . The system of claim 7 , the operations further comprising:
prior to generating the response, classifying an intent associated with the utterance based on the one or more named entities.
13 . One or more non-transitory computer-readable media storing instructions which, when executed by one or more processors, cause a chatbot system to perform operations comprising:
receiving an utterance, wherein the utterance comprises at least one named entity;
processing the utterance using a trained named entity recognition (NER) model to extract one or more named entities from the utterance, wherein the trained NER model is configured to identify named entities in utterances, and wherein the trained NER model is trained by:
accessing a machine learning model;
accessing training data comprising a plurality of utterances, wherein each utterance of the plurality of utterances comprises at least one named entity corresponding to a named entity category of a plurality of named entity categories;
calculating an augmentation rate for each named entity category of the plurality of named entity categories based on a distribution of each named entity category of the plurality of named entity categories in the training data and a function selected from among a plurality of functions in a hyperparameter tuning process of a named entity recognition (NER) model, wherein the plurality of functions includes at least one of a first function that calculates an augmentation rate for a given named entity category based on an exponential function that receives a normalized distribution value for the given named entity category and a beta value as an input and a second function that calculates an augmentation rate for the given named entity category based on a binary logarithm function that receives a number of occurrences that the given named entity category occurs in the training data as an input;
generating a plurality of additional utterances based on the augmentation rate calculated for each named entity category of the plurality of named entity categories;
augmenting the training data to form augmented training data by combining the plurality of additional utterances with the plurality of utterances; and
training the machine learning model with the augmented training data to result in the trained NER model; and
generating a response to the utterance based on the one or more named entities.
14 . The one or more non-transitory computer-readable media of claim 13 , wherein the hyperparameter tuning process comprises:
dividing the training data into sets of training data, each set of training data of the sets of training data comprising a corresponding set of utterances of the plurality of utterances;
for each set of training data of the sets of training data:
selecting a candidate hyperparameter from among a set of candidate hyperparameters,
generating, based on the candidate hyperparameter, a set of augmented utterances from the corresponding set of utterances for the respective set of training data based on a function of the plurality of functions associated with the candidate hyperparameter,
training the NER model with the set of augmented utterances, and
calculating a performance metric for the NER model trained with the set of augmented utterances; and
determining a set of hyperparameters for the NER model based on the performance metric calculated for the NER model trained with the respective sets of augmented utterances.
15 . The one or more non-transitory computer-readable media of claim 13 , wherein a number of additional utterances generated for a first named entity category of the plurality of named entity categories is different than a number of additional utterances generated for a second named entity category of the plurality of named entity categories.
16 . The one or more non-transitory computer-readable media of claim 13 , wherein determining the augmentation rate for each named entity category of the plurality of named entity categories comprises calculating a number of occurrences of each named entity category of the plurality of named entity categories in the training data and normalizing the number of occurrences of each named entity category of the plurality of named entity categories in the training data.
17 . The one or more non-transitory computer-readable media of claim 13 , the operations further comprising:
prior to generating the response, classifying an intent associated with the utterance based on the one or more named entities.