Quantized dialog language model for dialog systems

A computer usable program product comprising one or more computer-readable storage devices, and program instructions stored on at least one of the one or more storage devices, the stored program instructions comprising:

• program instructions to receive, at runtime, an utterance associated with a user-initiated dialog between a client device and a dialog system, wherein each utterance represents a turn in the dialog;
• program instructions to preprocess the utterance by replacing a particular phrase of the utterance with an indication of a type of the particular phrase;
• program instructions to map the preprocessed utterance to a vector representation of the utterance;
• program instructions to identify, by use of the vector representation, at least one cluster to which the preprocessed utterance belongs from among a plurality of clusters,
• wherein each of the plurality of possible clusters comprises clustered utterance vectors;
• program instructions to predict a next cluster based upon a conditional probability of the next cluster following a set of a predetermined number of previous clusters using a language model,
• wherein the language model is an n-gram language model trained to predict a next cluster from a previous sequence of n clusters, where n≥1 and where the clusters are sequenced to represent turns of dialog;
• program instructions to predict a next utterance from among a plurality of possible utterances within the predicted next cluster;
• and
• program instructions to output to the client device the predicted next utterance as a next system utterance.

A computer system comprising one or more processors, one or more computer-readable memories, and one or more computer-readable storage devices, and program instructions stored on at least one of the one or more storage devices for execution by at least one of the one or more processors via at least one of the one or more memories, the stored program instructions comprising:

• program instructions to receive, at runtime, an utterance associated with a user-initiated dialog between a client device and a dialog system, wherein each utterance represents a turn in the dialog;
• program instructions to preprocess the utterance by replacing a particular phrase of the utterance with an indication of a type of the particular phrase;
• program instructions to map the preprocessed utterance to a vector representation of the utterance;
• program instructions to identify, by use of the vector representation, at least one cluster to which the preprocessed utterance belongs from among a plurality of clusters,
• wherein each of the plurality of possible clusters comprises clustered utterance vectors;
• program instructions to predict a next cluster based upon a conditional probability of the next cluster following a set of a predetermined number of previous clusters using a language model,
• wherein the language model is an n-gram language model trained to predict a next cluster from a previous sequence of n clusters, where n≥1 and where the clusters are sequenced to represent turns of dialog;
• program instructions to predict a next utterance from among a plurality of possible utterances within the predicted next cluster;
• and
• program instructions to output to the client device the predicted next utterance as a next system utterance.