Knowledge fact retrieval through natural language processing
A method for fact retrieval includes receiving an input from a user, said input including a string of text that represents a user intent. The method includes performing natural language processing on the input to generate an embedding that corresponds to a semantic representation of the string of text, and based on the generated embedding, identifying an action that is associated with the user intent. The method includes executing the identified action, such that the executed action returns a result associated with the user intent, and providing the result to the user.
1 . A method, comprising:
receiving, by one or more processors, an input from an interface, the input comprising a string of text that represents a user intent;
generating, by the one or more processors, using a natural language processing technique on the input, an embedding that corresponds to a vector representation of the string of text, wherein the embedding encodes a semantic structure of the string of text within a multi-dimensional hyperspace;
executing, by the one or more processors, using a neural network, a vector search in the multi-dimensional hyperspace that determines distances between one or more vector representations of a plurality of pre-determined embeddings and the vector representation of the embedding that corresponds to the string of text, wherein each of the plurality of pre-determined embeddings maps to a respective script configured to retrieve data from at least one of a local data source, cloud service, or remote storage service;
identifying, by the one or more processors, based on the vector search using the embedding that corresponds to the string of text, at least one pre-determined embedding from the plurality of pre-determined embeddings, the at least one pre-determined embedding associated with the user intent;
retrieving, by the one or more processors, at least one script that maps to the at least one pre-determined embedding, the at least one script comprising computer-executable instructions configured to perform operations based at least in part on a semantic structure of the at least one pre-determined embedding;
executing, by the one or more processors, the computer-executable instructions from the at least one script to produce a result associated with the user intent, the result comprising a response to the input from at least one of the local data source, cloud service, or remote storage service; and
providing, by the one or more processors, the response to the input using the interface.
2 . The method of claim 1 , wherein identifying the at least one pre-determined embedding comprises determining a similarity between the embedding that corresponds to the string of text and each of the plurality of pre-determined embeddings, and based on the determined similarity, selecting the at least one pre-determined embedding from the plurality of pre-determined embeddings.
3 . The method of claim 2 , wherein selecting the at least one pre-determined embedding comprises determining that the at least one pre-determined embedding is most similar, out of all of the plurality of pre-determined embeddings, to the embedding that corresponds to the string of text.
4 . The method of claim 1 , wherein the computer-executable instructions cause an identification of a storage from a plurality of storages, executing the computer-executable instructions comprises performing a query to the identified storage, the interface corresponds to a user, and the result is responsive to the query from the identified storage.
5 . The method of claim 4 , wherein the input further comprises a first-person identifier associated with the user, and the query to the identified storage comprises the first-person identifier.
6 . The method of claim 4 , wherein the user is a first user, and the string of text comprises an ambiguous reference to a second user, the method further comprising:
identifying, by the one or more processors, a plurality of user identifiers that at least partially match the ambiguous reference to the second user; and
selecting, by the one or more processors, a user identifier from the plurality of user identifiers,
wherein the query to the identified storage comprises the selected user identifier.
7 . The method of claim 6 , wherein the user identifier is selected by providing the identified plurality of user identifiers to the first user, and receiving a selection of the user identifier from the first user.
8 . The method of claim 7 , wherein the input further comprises a role attribute associated with the first user, wherein the user identifier is selected based on the role attribute of the first user.
9 . The method of claim 8 , wherein executing the computer-executable instructions comprises determining, based on the role attribute of the first user, that the first user is authorized to receive the result, wherein the user is not authorized to receive results from at least one other script.
10 . The method of claim 4 , wherein the string of text comprises an ambiguous reference to an identifier, the method further comprising:
identifying, by the one or more processors, a plurality of identifiers that at least partially match the ambiguous reference to the identifier;
providing, by the one or more processors, the plurality of identifiers to the user; and
receiving, by the one or more processors, a selection of the identifier from the user,
wherein the query to the identified storage comprises the selected identifier.
11 . The method of claim 10 , wherein the identifier is for one of a place, a document, a person, an access role, a job title, and a date.
12 . A non-transitory computer-readable medium storing a set of instructions that, when executed by one or more processors, cause the one or more processors to:
receive an input from an interface, the input comprising a string of text that represents a user intent;
generate, using a natural language processing technique on the input, an embedding that corresponds to a vector representation of the string of text, wherein the embedding encodes a semantic structure of the string of text within a multi-dimensional hyperspace;
executing, by the one or more processors, using a neural network, a vector search in the multi-dimensional hyperspace that determines distances between one or more vector representations of a plurality of pre-determined embeddings and the vector representation of the embedding that corresponds to the string of text, wherein each of the plurality of pre-determined embeddings maps to a respective script configured to retrieve data from at least one of a local data source, cloud service, or remote storage service;
identify, based on the vector search using the embedding that corresponds to the string of text, at least one pre-determined embedding from the plurality of pre-determined embeddings, the at least one pre-determined embedding associated with the user intent;
retrieve at least one script that maps to the at least one pre-determined embedding, the at least one script comprising computer-executable instructions configured to perform operations based at least in part on a semantic structure of the at least one pre-determined embedding;
execute the computer-executable instructions from the at least one script to produce a result associated with the user intent, the result comprising a response to the input from at least one of the local data source, cloud service, or remote storage service; and
provide the response to the input using the interface.
13 . The non-transitory computer-readable medium of claim 12 , wherein identifying the at least one pre-determined embedding comprises determining a similarity between the embedding that corresponds to the string of text and each of the plurality of pre-determined embeddings, and based on the determined similarity, selecting the at least one pre-determined embedding from the plurality of pre-determined embeddings.
14 . The non-transitory computer-readable medium of claim 13 , wherein selecting the at least one pre-determined embedding comprises determining that the at least one pre-determined embedding is most similar, of the plurality of pre-determined embeddings, to the embedding that corresponds to the string of text.
15 . The non-transitory computer-readable medium of claim 12 , wherein the computer-executable instructions cause an identification of a storage from a plurality of storages, executing the computer-executable instructions comprises performing a query to the identified storage, the interface corresponds to a user, and the result is responsive to the query from the identified storage.
16 . The non-transitory computer-readable medium of claim 15 , wherein the input further comprises a first-person identifier associated with the user, and the query to the identified storage comprises the first-person identifier.
17 . The non-transitory computer-readable medium of claim 15 , wherein the user is a first user, and the string of text comprises an ambiguous reference to a second user, and the instructions further causing the one or more processors to:
identify a plurality of user identifiers that at least partially match the ambiguous reference to the second user; and
select a user identifier from the plurality of user identifiers,
wherein the query to the identified storage comprises the selected user identifier.
18 . The non-transitory computer-readable medium of claim 17 , wherein the user identifier is selected by providing the identified plurality of user identifiers to the first user, and receiving a selection of the user identifier from the first user.
19 . The non-transitory computer-readable medium of claim 18 , wherein the input further comprises a role attribute associated with the first user, wherein the user identifier is selected based on the role attribute of the first user.
20 . The non-transitory computer-readable medium of claim 19 , wherein executing the computer-executable instructions comprises determining, based on the role attribute of the first user, that the first user is authorized to receive the result, wherein the user is not authorized to receive results from at least one other script.
21 . The non-transitory computer-readable medium of claim 15 , wherein the string of text comprises an ambiguous reference to an identifier, and the instructions further causing the one or more processors to:
identify a plurality of identifiers that at least partially match the ambiguous reference to the identifier;
provide the plurality of identifiers to the user; and
receive a selection of the identifier from the user,
wherein the query to the identified storage comprises the selected identifier.
22 . The non-transitory computer-readable medium of claim 21 , wherein the identifier is for one of a place, a document, a person, an access role, a job title, and a date.