Provided is a system and method for training and validating models in a machine learning pipeline for failure mode analytics. The machine learning pipeline may include an unsupervised training phase, a validation phase and a supervised training and scoring phase. In one example, the method may include receiving an identification of a machine learning model, executing a machine learning pipeline comprising a plurality of services which train the machine learning model via at least one of an unsupervised learning process and a supervised learning process, the machine learning pipeline being controlled by an orchestration module that triggers ordered execution of the services, and storing the trained machine learning model output from the machine learning pipeline in a database associated with the machine learning pipeline.

Devices and techniques are generally described for cohort determination in natural language processing. In various examples, a first natural language input to a natural language processing system may be determined. The first natural language input may be associated with a first account identifier. A first machine learning model may determine first data representing one or more words of the first natural language input. A second machine learning model may determine second data representing one or more acoustic characteristics of the first natural language input. Third data may be determined, the third data including a predicted performance for processing the first natural language input by the natural language processing system. The third data may be determined based on the first data representation and the second data representation.

An illustrative embodiment of a computer-implemented method for predictive analytic queries includes creating a user-defined predictive analytics query using a set of syntactic grammar that defines a correct syntax of the user-defined predictive analytics query including a created set of predictive analytics by-example vocabularies and a set of subject-specific by-example vocabularies forming a set of by-example vocabularies, wherein the set of syntactic grammar defines semantics of each syntactically correct predictive analytics query using the by-example vocabularies such that predictive analytics queries can be expressed with semantic precision using this constrained Natural Language Processing (cNLP) approach. The computer-implemented method further generates a predictive analytic model and runtime query, using the user-defined predictive analytics query, executes the runtime query using the predictive analytic model to create a result, and returns the result to the user.

Provided is a system and method for training and validating models in a machine learning pipeline for failure mode analytics. The machine learning pipeline may include an unsupervised training phase, a validation phase and a supervised training and scoring phase. In one example, the method may include receiving a request to create a machine learning model for failure mode detection associated with an asset, retrieving historical notification data of the asset, generating an unsupervised machine learning model via unsupervised learning on the historical notification data, wherein the unsupervised learning comprises identifying failure topics from text included in the historical notification data and mapping the identified failure topics to a plurality of predefined failure modes for the asset, and storing the generated unsupervised machine learning model via a storage device.