Training example generation to create new intents for chatbots
A topic for building a new intent on which to train a chatbot can be received. A database of chatbot training data can be searched for a candidate intent having meta-knowledge similar to the received topic. Utterances associated with the candidate intent can be extracted. The received topic and the extracted utterances can be input to a trained machine learning model. The trained machine learning model generates example utterances for the new intent. The new intent with the generated example utterances can be used as training data for training the chatbot.
1 . A computer-implemented method comprising:
receiving a topic for building a new intent on which to train a chatbot;
mapping the received topic to meta-knowledge based on similarity of the topic and the meta-knowledge;
searching a database of chatbot training data for a candidate intent having the meta-knowledge;
extracting utterances associated with the candidate intent; and
inputting the received topic and the extracted utterances to a trained machine learning model, the trained machine learning model generating example utterances for the new intent, the trained machine learning model having been trained to convert a first intent to a second intent based on: a plurality of pairs of input samples, a pair of input sample in the plurality of pairs of input samples comprising a training utterance and a training utterance's meta-knowledge associated with the first intent, and using as ground truth a plurality of second intent's examples and associated meta-knowledge;
discarding an example utterance from the generated example utterances responsive to determining that the example utterance is similar, based on a threshold level of similarity, to an extracted utterance of the extracted utterances used as input in training the machine learning model;
presenting via a graphical user interface the at least the example utterances for validation of the example utterances;
training the chatbot using at least the example utterances without the discarded example utterance, generated by the trained machine learning model; and
automatically updating knowledge and retraining of the chatbot using a set of intents that includes the first intent and the converted second intent thereby causing the chatbot to interact with a user by carrying on a conversation associated with the new intent that is different from a previous intent.
2 . The method of claim 1 , wherein the extracted utterances include a question associated with the candidate intent, and wherein the example utterances generated by the trained machine learning model include a question associated with the topic,
the method further including generating an answer to the question by at least searching an external network for the answer for responding to the question associated with the topic, wherein the generated answer with the question is used in training the chatbot.
3 . The method of claim 2 , further including validating the answer.
4 . The method of claim 1 , wherein the machine learning model includes a neural network.
5 . A system comprising:
a hardware processor; and
a memory device coupled with the hardware processor,
the hardware processor configured to at least:
receive a topic for building a new intent on which to train a chatbot;
map the received topic to meta-knowledge based on similarity of the topic and the meta-knowledge;
search a database of chatbot training data for a candidate intent having the meta-knowledge;
extract utterances associated with the candidate intent; and
input the received topic and the extracted utterances to a trained machine learning model, the trained machine learning model generating example utterances for the new intent, the trained machine learning model having been trained to convert a first intent to a second intent based on: a plurality of pairs of input samples, a pair of input sample in the plurality of pairs of input samples comprising a training utterance and a training utterance's meta-knowledge associated with the first intent, and using as ground truth a plurality of second intent's examples and associated meta-knowledge;
discard an example utterance from the generated example utterances responsive to determining that the example utterance is similar, based on a threshold level of similarity, to an extracted utterance of the extracted utterances used as input in training the machine learning model;
present via a graphical user interface the at least the example utterances for validation of the example utterances;
train the chatbot using at least the example utterances without the discarded example utterance, generated by the trained machine learning model; and
automatically update knowledge and retrain the chatbot using a set of intents that includes the first intent and the converted second intent thereby cause the chatbot to interact with a user by carrying on a conversation associated with the new intent that is different from a previous intent.
6 . The system of claim 5 , wherein the extracted utterances include a question associated with the candidate intent, and wherein the example utterances generated by the trained machine learning model include a question associated with the topic, wherein the hardware processor is further configured to generate an answer to the question by at least searching an external network for the answer for responding to the question associated with the topic, wherein the generated answer with the question is used in training the chatbot.
7 . The system of claim 6 , wherein the hardware processor is further configured to validate the answer.
8 . The system of claim 5 , wherein the trained machine learning model includes a neural network.
9 . The system of claim 5 , wherein the trained machine learning model is trained based on input data including intents and associated utterances, wherein the trained machine learning model is trained to, given a new intent topic, generate example utterances associated with the new intent topic.
10 . A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions readable by a device to cause the device to:
receive a topic for building a new intent on which to train a chatbot;
map the received topic to meta-knowledge based on similarity of the topic and the meta-knowledge;
search a database of chatbot training data for a candidate intent having meta-knowledge similar to the received topic;
extract utterances associated with the candidate intent; and
input the received topic and the extracted utterances to a trained machine learning model, the trained machine learning model generating example utterances for the new intent, the trained machine learning model having been trained to convert a first intent to a second intent based on:
a plurality of pairs of input samples, a pair of input sample in the plurality of pairs of input samples comprising a training utterance and a training utterance's meta-knowledge associated with the first intent, and using as ground truth a plurality of second intent's examples and associated meta-knowledge;
discard an example utterance from the generated example utterances responsive to determining that the example utterance is similar, based on a threshold level of similarity, to an extracted utterance of the extracted utterances used as input in training the machine learning model;
present via a graphical user interface the at least the example utterances for validation of the example utterances;
train the chatbot using at least the example utterances without the discarded example utterance, generated by the trained machine learning model; and
automatically update knowledge and retrain the chatbot using a set of intents that includes the first intent and the converted second intent thereby cause the chatbot to interact with a user by carrying on a conversation associated with the new intent that is different from a previous intent.
11 . The computer program product of claim 10 , wherein the extracted utterances include a question associated with the candidate intent, and wherein the example utterances generated by the trained machine learning model include a question associated with the topic, wherein the device is further configured to generate an answer to the question by at least searching an external network for the answer for responding to the question associated with the topic, wherein the generated answer with the question is used in training the chatbot.
12 . The computer program product of claim 11 , wherein the device is further caused to validate the answer.
13 . The computer program product of claim 10 , wherein the trained machine learning model includes a neural network.
14 . The computer program product of claim 10 , wherein the trained machine learning model is trained based on input data including intents and associated utterances, wherein the trained machine learning model is trained to, given a new intent topic, generate example utterances associated with the new intent topic.