Abstract: This specification is generally directed to techniques for generating interfacing source code between computing components based on natural language input. In various implementations, a natural language input that requests generation of interfacing source code to logically couple a first computing component with a second computing component may be processed to generate an interface request semantic embedding. The interface request semantic embedding may be processed based on one or more domain models associated with the first and second computing components to generate a pool(s) of candidate code snippets for logically coupling with first and second computing components. A plurality of candidate instances of interfacing source code may be generated between the first and second computing components. Each candidate software interface may include a different permutation of candidate code snippets from the pool(s) of candidate code snippets.

Abstract: Disclosed implementations relate to adding “bottleneck” models to machine learning pipelines that already apply domain models to translate and/or transfer representations of high-level semantic concepts between domains. In various implementations, an initial representation in a first domain of a transition from an initial state of an environment to a goal state of the environment may be processed based on a pre-trained first domain encoder to generate a first embedding that semantically represents the transition. The first embedding may be processed based on one or more bottleneck models to generate a second embedding with fewer dimensions than the first embedding. In various implementations, the second embedding may be processed in various ways to train one or more of the bottleneck model(s) based on various different auxiliary loss functions.

Abstract: As opposed to a rigid approach, implementations disclosed herein utilize a flexible approach in automatically determining an action set to utilize in attempting performance of a task that is requested by natural language input of a user. The approach is flexible at least in that embedding technique(s) and/or action model(s), that are utilized in generating action set(s) from which the action set to utilize is determined, are at least selectively varied. Put another way, implementations leverage a framework via which different embedding technique(s) and/or different action model(s) can at least selectively be utilized in generating different candidate action sets for given NL input of a user. Further, one of those action sets can be selected for actual use in attempting real-world performance of a given task reflected by the given NL input. The selection can be based on a suitability metric for the selected action set and/or other considerations.

Abstract: Training and/or utilization of a neural decompiler that can be used to generate, from a lower-level compiled representation, a target source code snippet in a target programming language. In some implementations, the lower-level compiled representation is generated by compiling a base source code snippet that is in a base programming language, thereby enabling translation of the base programming language (e.g., C++) to a target programming language (e.g., Python). In some of those implementations, output(s) from the neural decompiler indicate canonical representation(s) of variables. Technique(s) can be used to match those canonical representation(s) to variable(s) of the base source code snippet. In some implementations, multiple candidate target source code snippets are generated using the neural decompiler, and a subset (e.g., one) is selected based on evaluation(s).