Parsing techniques
AssigneeWiz, Inc.
InventorsBarak BERCOVITZ, Bernie PINKENZON-HOWARD, Eshel YARON
Life of this patent
Estimating the term…
Abstract
A system and method for parsing. A method includes querying a language model based on a code sample in order to obtain a second file, wherein outputs of the language model are refined based on a comparison between a set of expected results output by the language model for at least one set of example code and at least one first parser output obtained by inputting at least one first file to a parser; providing the second file to the parser in order to obtain a second parser output; and identifying at least one endpoint based on the second parser output.
Claims25
A method for endpoint identification:
- obtaining outputs, at each iteration of a one or more iterations, from a language model based on a set of example code for a respective iteration in the one or more iterations, where the outputs include a set of expected results for the respective iteration and a first file for the respective iteration;
- obtaining, at each iteration of the one or more iterations, a parser output for the respective iteration based on the first file for the respective iteration;
- comparing, at each iteration of the one or more iterations, the set of the expected results for the respective iteration with the parser output of the respective iteration;
- determining the outputs of the language model are refined based on the comparison;
- querying the language model based on a code sample in order to obtain a second file;
- obtaining a second parser output, via the parser, based on the second file;
- and
- identifying at least one endpoint based on the second parser output.
(canceled)
(canceled)
(canceled)
A non-transitory computer readable medium having stored thereon instructions for causing a processing circuitry to execute a process, the process comprising:
- obtaining outputs, at each iteration of a one or more iterations, from a language model based on a set of example code for a respective iteration in the one or more iterations, where the outputs include a set of expected results for the respective iteration and a first file for the respective iteration;
- obtaining, at each iteration of the one or more iterations, a parser output for the respective iteration based on the first file for the respective iteration;
- comparing, at each iteration of the one or more iterations, the set of the expected results for the respective iteration with the parser output of the respective iteration;
- determining the outputs of the language model are refined based on the comparison;
- querying the language model based on a code sample in order to obtain a second file;
- obtaining a second parser output, via the parser, based on the second file;
- and
- identifying at least one endpoint based on the second parser output.
A system for endpoint identification, comprising:
- a processing circuitry;
- and
- a memory, the memory containing instructions that, when executed by the processing circuitry, configure the system to:
- obtain outputs, at each iteration of a one or more iterations, from a language model based on a set of example code for a respective iteration in the one or more iterations, where the outputs include a set of expected results for the respective iteration and a first file for the respective iteration;
- obtain, at each iteration of the one or more iterations, a parser output for the respective iteration based on the first file for the respective iteration;
- compare, at each iteration of the one or more iterations, the set of the expected results for the respective iteration with the parser output of the respective iteration;
- determine the outputs of the language model are refined based on the comparison;
- query the language model based on a code sample in order to obtain a second file;
- obtain a second parser output, via the parser, based on the second file;
- and
- identify at least one endpoint based on the second parser output.
(canceled)
(canceled)
(canceled)
Dependencies are inferred from the claim text.