Abstract: Devices and techniques are generally described for machine learning personalization as a service for speech processing applications. In various examples, a first request for machine learning prediction for a first speech processing skill. First skill data schema data may be received that describes content of the first speech processing skill. A first machine learning model for the first speech processing skill may be determined. A first feature definition describing a first aspect of the content may be determined. A second feature definition describing user profile data may be determined. A prediction request may be received from the first speech processing skill. First feature data may be generated according to the first feature definition and second feature data may be generated according to the second feature definition based at least in part on the prediction request.

Abstract: Techniques for performing feature computation are described. A system may gather and analyze event data to generate a timeline of the event data and corresponding feature data (e.g., statistical values representing the event data). The system can create a customized timeline that allows information to be sorted and aggregated in different timescales to improve speech processing and other functionality. For example, a feature computation system may calculate statistics and other information based on interactions with a speech processing system. These statistics provide information about previous interactions that may be leveraged to interpret future voice commands.

Abstract: Search Query Concept Based Recommendations. A method includes electronically receiving, in a computer system, a title as an input. One or more subsets of the title are generated. A list of key phrases associated with the title is obtained. Further, one or more key phrases from the list corresponding to the one or more subsets is electronically identified. Contents are selected based on the one or more key phrases identified. The contents can be ranked. The contents are then provided to the user.

Abstract: A candidate's career information is obtained. The career information comprises one or more of employment history information and education information. If the candidate's career information comprises employment history information, the employment history information is compared with stored employment history information of a plurality of individuals. If the candidate's career information comprises education information, the education information is compared with stored education information of a plurality of individuals. Stored career information meeting a similarity threshold is identified from one or more of the comparing steps. The similarity threshold relates to similarity between (1) the stored employment history information and the candidate's employment history information, and/or (2) the stored education information and the candidate's education information. The similarity threshold can also relate to a candidate's actions (e.g., apply history).