Output interpretation for a meaning representation language system
The present disclosure is related to techniques for converting a natural language utterance to a logical form query and deriving a natural language interpretation of the logical form query. The techniques include accessing a Meaning Resource Language (MRL) query and converting the MRL query into a MRL structure including logical form statements. The converting includes extracting operations and associated attributes from the MRL query and generating the logical form statements from the operations and associated attributes. The techniques further include translating each of the logical form statements into a natural language expression based on a grammar data structure that includes a set of rules for translating logical form statements into corresponding natural language expressions, combining the natural language expressions into a single natural language expression, and providing the single natural language expression as an interpretation of the natural language utterance.
1 . A computer-implemented method comprising:
accessing a Meaning Resource Language (MRL) query, wherein the MRL query is a logical form of a natural language utterance;
converting the MRL query into a MRL structure comprising one or more logical form statements, wherein the converting comprises extracting one or more operations and associated one or more attributes from the MRL query and generating the one or more logical form statements from the one or more operations and associated one or more attributes;
translating each of the one or more logical form statements into a natural language expression based on a grammar data structure that includes a set of rules for translating logical form statements into corresponding natural language expressions;
combining the one or more natural language expressions into a single natural language expression;
executing the MRL query or a system query generated from the MRL query on a system to retrieve data responsive to the natural language utterance; and
providing the single natural language expression of the natural language utterance and the data to a user.
2 . The computer-implemented method of claim 1 , further comprising:
generating an input string by concatenating the natural language utterance with a schema representation for a system;
generating, by a first encoder, one or more embeddings of the natural language utterance and the schema representation based on the input string;
encoding, by a second encoder, relations between elements in the schema representation and words in the natural language utterance based on the one or more embeddings; and
generating, by a grammar-based decoder, the MRL query based on the encoded relations and the one or more embeddings.
3 . The computer-implemented method of claim 1 , further comprising
generating the system query in a system query language based on the MRL query and an interface specification.
4 . The computer-implemented method of claim 1 , wherein the combining comprises concatenating the one or more natural language expressions together in a chain or series to form single natural language expression, and wherein an order of the one or more natural language expressions in the chain or series is specified within an operation clause used for the concatenating.
5 . The computer-implemented method of claim 1 , wherein the converting further comprises:
identifying one or more keywords and/or clauses within the one or more logical form statements; and
attaching one or more tags to the one or more logical form statements based on the identification of the one or more keywords and/or the clauses within the one or more logical form statements, wherein the one or more tags indicate specific logical form patterns requiring special translation procedures.
6 . The computer-implemented method of claim 5 , wherein the combining comprises:
determining the one or more tags are attached to the one or more logical form statements;
responsive to determining the one or more tags are attached to the one or more logical form statements, identifying a logical form pattern within a logical form statement of the one or more logical form statements based on a tag of the one or more tags; and
applying a special translation procedure of the special translation procedures to the one or more natural language expressions to obtain at least a portion of the single natural language expression, wherein the special translation procedure is associated with the logical form pattern.
7 . The computer-implemented method of claim 1 , wherein the grammar data structure defines a specialized context free grammar comprising the set of rules including non-terminal and terminal terms or clauses, wherein non-terminal terms or clauses include syntax that have different expressions for different logic clauses, including corresponding natural language terms or expressions to be used for the non-terminal terms or clauses, and the terminal terms or clauses are components of sentences generated using a grammar and specify corresponding natural language terms or expressions to be used for the terminal terms or clauses.
8 . A system comprising:
one or more processors; and
one or more computer-readable media storing instructions which, when executed by the one or more processors, cause the system to perform operations comprising:
accessing a Meaning Resource Language (MRL) query, wherein the MRL query is a logical form of a natural language utterance;
converting the MRL query into a MRL structure comprising one or more logical form statements, wherein the converting comprises extracting one or more operations and associated one or more attributes from the MRL query and generating the one or more logical form statements from the one or more operations and associated one or more attributes;
translating each of the one or more logical form statements into a natural language expression based on a grammar data structure that includes a set of rules for translating logical form statements into corresponding natural language expressions;
combining the one or more natural language expressions into a single natural language expression;
executing the MRL query or a system query generated from the MRL query on a system to retrieve data responsive to the natural language utterance; and
providing the single natural language expression of the natural language utterance and the data to a user.
9 . The system of claim 8 , wherein the operations further comprise:
generating an input string by concatenating the natural language utterance with a schema representation for a system;
generating, by a first encoder, one or more embeddings of the natural language utterance and the schema representation based on the input string;
encoding, by a second encoder, relations between elements in the schema representation and words in the natural language utterance based on the one or more embeddings; and
generating, by a grammar-based decoder, the MRL query based on the encoded relations and the one or more embeddings.
10 . The system of claim 8 , wherein the operations further comprise:
generating the system query in a system query language based on the MRL query and an interface specification.
11 . The system of claim 8 , wherein the combining comprises concatenating the one or more natural language expressions together in a chain or series to form single natural language expression, and wherein an order of the one or more natural language expressions in the chain or series is specified within an operation clause used for the concatenating.
12 . The system of claim 8 , wherein the converting further comprises:
identifying one or more keywords and/or clauses within the one or more logical form statements; and
attaching one or more tags to the one or more logical form statements based on the identification of the one or more keywords and/or the clauses within the one or more logical form statements, wherein the one or more tags indicate specific logical form patterns requiring special translation procedures.
13 . The system of claim 12 , wherein the combining comprises:
determining the one or more tags are attached to the one or more logical form statements;
responsive to determining the one or more tags are attached to the one or more logical form statements, identifying a logical form pattern within a logical form statement of the one or more logical form statements based on a tag of the one or more tags; and
applying a special translation procedure of the special translation procedures to the one or more natural language expressions to obtain at least a portion of the single natural language expression, wherein the special translation procedure is associated with the logical form pattern.
14 . The system of claim 8 , wherein the grammar data structure defines a specialized context free grammar comprising the set of rules including non-terminal and terminal terms or clauses, wherein non-terminal terms or clauses include syntax that have different expressions for different logic clauses, including corresponding natural language terms or expressions to be used for the non-terminal terms or clauses, and the terminal terms or clauses are components of statements generated using a grammar and specify corresponding natural language terms or expressions to be used for the terminal terms or clauses.
15 . One or more non-transitory computer-readable media storing instructions which, when executed by one or more processors, cause a system to perform operations comprising:
accessing a Meaning Resource Language (MRL) query, wherein the MRL query is a logical form of a natural language utterance;
converting the MRL query into a MRL structure comprising one or more logical form statements, wherein the converting comprises extracting one or more operations and associated one or more attributes from the MRL query and generating the one or more logical form statements from the one or more operations and associated one or more attributes;
translating each of the one or more logical form statements into a natural language expression based on a grammar data structure that includes a set of rules for translating logical form statements into corresponding natural language expressions;
combining the one or more natural language expressions into a single natural language expression;
executing the MRL query or a system query generated from the MRL query on a system to retrieve data responsive to the natural language utterance; and
providing the single natural language expression of the natural language utterance and the data to a user.
16 . The one or more non-transitory computer-readable media of claim 15 , wherein the operations further comprise:
generating an input string by concatenating the natural language utterance with a schema representation for a system;
generating, by a first encoder, one or more embeddings of the natural language utterance and the schema representation based on the input string;
encoding, by a second encoder, relations between elements in the schema representation and words in the natural language utterance based on the one or more embeddings; and
generating, by a grammar-based decoder, the MRL query based on the encoded relations and the one or more embeddings.
17 . The one or more non-transitory computer-readable media of claim 15 , wherein the operations further comprise;
generating the system query in a system query language based on the MRL query and an interface specification.
18 . The one or more non-transitory computer-readable media of claim 15 , wherein the combining comprises concatenating the one or more natural language expressions together in a chain or series to form single natural language expression, and wherein an order of the one or more natural language expressions in the chain or series is specified within an operation clause used for the concatenating.
19 . The one or more non-transitory computer-readable media of claim 15 , wherein the converting further comprises:
identifying one or more keywords and/or clauses within the one or more logical form statements; and
attaching one or more tags to the one or more logical form statements based on the identification of the one or more keywords and/or the clauses within the one or more logical form statements, wherein the one or more tags indicate specific logical form patterns requiring special translation procedures.
20 . The one or more non-transitory computer-readable media of claim 19 , wherein the combining comprises:
determining the one or more tags are attached to the one or more logical form statements;
responsive to determining the one or more tags are attached to the one or more logical form statements, identifying a logical form pattern within a logical form statement of the one or more logical form statements based on a tag of the one or more tags; and
applying a special translation procedure of the special translation procedures to the one or more natural language expressions to obtain at least a portion of the single natural language expression, wherein the special translation procedure is associated with the logical form pattern.