Abstract: Certain aspects of the present disclosure provide techniques for executing structured query language queries having a schema associated therewith against an application programming interface using natural language. The schema can be chunked such that embeddings of the resulting chunks are stored in a vector store. Schemas (or subschemas) generated using on or more chunks of the vector store may be provided to a large language model along with a NL query to generate a structured query language query which may be executed against the application programming interface. This allows large language models to produce structured query language queries, such as GraphQL queries even if a GraphQL schema is too large to be provided to the model as context. Aspects disclosed herein also provide techniques for client code generation and client software development kit generation.

Abstract: During a survey technique, a product survey in an original language is translated into a target language, and any errors in the translation are identified and corrected. Words in the translation that do not translate well or clearly may be reverted back to the original language based on transliteration instructions. This may allow customers to answer the translated product survey in the target language using a user interface (such as a keyboard) based on the original language. After receiving answers from the users to questions in the translated product survey, the answers may be analyzed to determine results of the product survey in the target language. Then, the results may be translated back into the original language. After translation into the original language, the results across multiple target languages can be consolidated and presented.