Systems and methods for navigating nodes in channel based chatbots using natural language understanding
Systems and methods for using agents are provided. A first message is received from a user to engage with a corresponding session of an automated content logic. The automated content logic is configured to facilitate electronic communication between with the user and a plurality of agents associated with the automated content logic. The first message is applied as an input to a first set of agents to receive a corresponding output for each respective agent in the first set of agents. Each corresponding output identifies an action performed by the automated content logic responsive to the first message that best matches with the first message. The corresponding output for each respective agent in the first set of agents is used to identify a first action. A process is performed in accordance with the first action to generate a second message to engage with the corresponding session.
1 . A method comprising:
A) receiving, in electronic form, a first message from a user to engage with a corresponding session of an automated content logic configured, at least in part, to facilitate electronic communication between with the user and a plurality of agents associated with the automated content logic;
B) applying the first message as an input to a first set of agents, in the plurality of agents, to receive a corresponding output for each respective agent in the first set of agents, wherein each corresponding output identifies an action, in a plurality of actions, performed, at least in part, by the automated content logic, wherein the plurality of actions is defined in a registered set of actions, each action in the registered set of actions comprising a typed schema specifying a plurality of parameters, one or more authentication types, and one or more return types;
C) using the corresponding output for each respective agent in the first set of agents to identify a first action in the plurality of actions; and
D) performing a process in accordance with the first action to generate a second message to engage with the corresponding session.
2 . The method of claim 1 , wherein the receiving A) comprises detecting a first protocol, in a plurality of protocols, associated with the corresponding session, and selecting the first set of agents from the plurality of agents in accordance with the first protocol.
3 . The method of claim 2 , wherein the plurality of protocols comprises
(i) one or more security protocols,
(ii) one or more privacy protocols,
(iii) one or more communication protocols,
(iv) one or more display protocols,
(v) one or more session protocols,
(vi) one or more data protocols,
(vii) one or more regional protocols,
(viii) one or more resource protocols,
(ix) one or more translation protocols,
(x) one or more latency protocols,
(xi) one or more transformation protocols,
(xii) one or more model context protocols, or
(xiii) a combination thereof.
4 . The method of claim 1 , wherein the method further comprises, prior to the applying B), selecting the first set of agents from the plurality of agents in accordance with a plurality of rules of the automated content logic comprising
a first set of rules associated with one or more learned routings of two or more agents in the plurality of agents; or
a second set of rules in the plurality of rules, associated with one or more contexts of a respective message received during a respective session.
5 . The method of claim 1 , wherein
the automated content logic is based, at least in part, on a plurality of rules, and
at least one respective rule in the plurality of rules is associated with a corresponding action in the plurality of actions, wherein
the automated content logic is configured receive a request, prior to the applying B), to configure a first rule in the plurality of rules comprising a change of at least an output dependency or an input dependency of the corresponding action associated with the first rule.
6 . The method of claim 1 , wherein each respective agent in the first set of agents is configured to independently identify a classification in a plurality of classifications associated with the action.
7 . The method of claim 1 , wherein each respective agent in the first set of agents is independently selected from the group consisting of: a linear regression model, a logistic regression model, a decision tree model, a classification model, a regression tree model, a Naïve Bayes model, a nearest neighbor clustering model, an a Apriori model, a means clustering model, a principal component analysis model, a random forest model, an adaptive boosting model, a generative model, a Gaussian model, a multinomial distribution model, a hidden Markov model, a low density separation model, a node graph model, a minimum cut model, a harmonic function model, a manifold regularization model, a heuristic model, a support vector machine model, a survival analysis model, a cox proportional hazard model, a ranking model, a product limit estimation model, and a neural network model.
8 . The method of claim 1 , wherein each respective agent in the first set of agents is a component module of a linear regression model, a logistic regression model, a decision tree model, a classification model, a regression tree model, a Naïve Bayes model, a nearest neighbor clustering model, an Apriori model, a means clustering model, a principal component analysis model, a random forest model, an adaptive boosting model, a generative model, a Gaussian model, a multinomial distribution model, a hidden Markov model, a low density separation model, a node graph model, a minimum cut model, a harmonic function model, a manifold regularization model, a heuristic model, a support vector machine model, a neural network model, or a combination thereof.
9 . The method of claim 1 , wherein a first agent in the first set of agents comprises a plurality of layers, wherein
a first layer in the plurality of layers is associated with one or more persistent natural language rules, and
a second layer in the plurality of layers is associated with one or more context natural language processing rules.
10 . The method of claim 1 , wherein the plurality of agents comprises
a first agent trained to identify a respective action in the plurality of actions based on one or more user demographic classifications in a plurality of classifications,
a second agent trained to identify the respective action in the plurality of actions based on one or more geographic classifications in the plurality of classifications,
a third agent trained to identify the respective action in the plurality of actions based on one or more historical classifications in the plurality of classifications,
a fourth agent trained to identify the respective action in the plurality of actions based on one or more pattern-based classifications in the plurality of classifications,
a fifth agent trained to identify the respective action in the plurality of actions based on one or more sentiment classifications in the plurality of classifications,
a sixth agent trained to identify the respective action in the plurality of actions based on one or more intent classifications in the plurality of classifications,
a seventh agent trained to identify the respective action in the plurality of actions based on one or more tone classifications in the plurality of classifications,
an eighth agent trained to identify the respective action in the plurality of actions based on one or more personalized classifications uniquely associated with the user,
a ninth agent trained to identify the respective action in the plurality of actions based on the plurality of classifications,
a tenth agent trained as a retrieval relevance scorer that queries a vector index, or
a combination thereof.
11 . The method of claim 1 , wherein each respective agent in the plurality of agents is independently associated with one or more corresponding functions and a corresponding plurality of parameters applied to the one or more corresponding functions.
12 . The method of claim 1 , wherein a first agent in the first set of agents comprises
a first component, in a plurality of components, configured to generate a first output using the first message, and
a second component, in the plurality of components, configured generate the corresponding output using the first output.
13 . The method of claim 1 , wherein
the applying B) comprises generating a first plurality of tokens associated with the first message, and
the action is generating a second plurality of tokens different from the first plurality of tokens.
14 . The method of claim 1 , wherein
the applying B) comprises generating a first vector matrix associated with the first message, and
the action is selecting a second vector matrix different from the first vector matrix.
15 . The method of claim 1 , wherein the applying B) comprises
applying (i) the first message and (ii) a first data set associated with a content of the corresponding session as the input to each agent in the first set of agents.
16 . The method of claim 1 , wherein the applying B) comprises
applying (i) the first message and (ii) a second data set as the input to each agent in the first set of agents, wherein the second data set comprises one or more prior conversations associated with the user and/or one or more attributes associated with a corresponding user profile of the user.
17 . The method of claim 1 , wherein the applying B) comprises
forming an n-dimensional hyperspace using the first message, and
applying the n-dimensional hyperspace to the first set of agents.
18 . The method of claim 1 , wherein a first agent in the first set of agents is an attention based neural network.
19 . The method of claim 1 , wherein a first agent in the first set of agents is neural network comprises a transformer architecture.
20 . The method of claim 1 , wherein the using C) comprises identifying the first action through
a majority vote among the corresponding output for each respective agent in the first set of agents,
a weighted vote among the corresponding output for each respective agent in the first set of agents, or
a proximity of actions among the corresponding output for each respective agent in the first set of agents.
21 . The method of claim 1 , wherein the using C) comprising
applying a normalization function collectively to the corresponding output for each respective agent in the first set of agents, and/or
applying an aggregation function collectively to the corresponding output for each respective agent in the first set of agents.
22 . The method of claim 21 , wherein the normalization function comprises a temperature scaling of logits.
23 . The method of claim 21 , wherein the corresponding output for each respective agent in the first set of agents comprises a respective set of values, and wherein the normalization function determines a collective distribution of the respective set of values associated with each respective agent in the first set of agents.
24 . The method of claim 1 , wherein
the using C) comprises using the corresponding output for each respective agent in the first set of agents to identify a set of actions, in the plurality of actions, comprising the first action and a second action different from the first action, and
the performing D) comprises generating the second message in accordance with a determination a first value associated with the first message exceeds a second value associated with the second message.
25 . The method of claim 1 , wherein the first action comprises
(i) obtaining, via a communication network, a third data set from a remote device in accordance with an application programing interface token associated with the remote device,
(ii) communicating, via the communication network, a fourth data set to the remote device in accordance with the application programing interface token associated with the remote device,
(iii) identifying a second set of agents in the plurality of agents different from the first set of agents to for the performing D),
(iv) identifying a first parameter in a plurality of parameters associated with the first message,
(v) generating a text string for the second message,
(vi) displaying the second message,
(vii) identifying a language of the first message,
(viii) executing one or more instructions,
(ix) evaluating a first graphical image,
(x) generating a second graphical image,
(xi) joining a human user different from the user to the corresponding session,
(xii) a future best action by the automated content logic,
(xiii) a workflow for the automated content logic,
(xiv) compiling a survey for the user,
(xv) a function call,
(xvi) a tool invocation,
(xvii) issuing a scoped capability token with per-action permissions and expiry, or
(xviii) a combination thereof.
26 . The method of claim 1 , wherein the performing D) comprises
applying (i) the first message and (ii) a second data set associated with a content of the corresponding session as an input to a first agent to receive as output the second message; and/or
causing (i) the second message and (ii) one or more labels associated with the second message to display at a remote device associated with the user.
27 . The method of claim 1 , wherein the method further comprises recording one or more score outputs, one or more selected actions, one or more execution results, or a combination thereof in a tamper-evident audit.
28 . A non-transitory computer readable storage medium, wherein the non-transitory computer readable storage medium stores instructions, which when executed by a computer system, cause the computer system to perform a method of:
A) receiving, in electronic form, a first message from a user to engage with a corresponding session of an automated content logic configured, at least in part, to facilitate electronic communication between with the user and a plurality of agents associated with the automated content logic;
B) applying the first message as an input to a first set of agents, in the plurality of agents, to receive a corresponding output for each respective agent in the first set of agents, wherein each corresponding output identifies an action, in a plurality of actions, performed, at least in part, by the automated content logic, wherein the plurality of actions is defined in a registered set of actions, each action in the registered set of actions comprising a typed schema specifying a plurality of parameters, one or more authentication types, and one or more return types;
C) using the corresponding output for each respective agent in the first set of agents to identify a first action in the plurality of actions; and
D) performing a process in accordance with the first action to generate a second message to engage with the corresponding session.
29 . A server system, comprising:
one or more processors;
memory; and
one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, the one or more programs including instructions for:
A) receiving, in electronic form, a first message from a user to engage with a corresponding session of an automated content logic configured, at least in part, to facilitate electronic communication between with the user and a plurality of agents associated with the automated content logic;
B) applying the first message as an input to a first set of agents, in the plurality of agents, to receive a corresponding output for each respective agent in the first set of agents, wherein each corresponding output identifies an action, in a plurality of actions, performed, at least in part, by the automated content logic, wherein the plurality of actions is defined in a registered set of actions, each action in the registered set of actions comprising a typed schema specifying a plurality of parameters, one or more authentication types, and one or more return types;
C) using the corresponding output for each respective agent in the first set of agents to identify a first action in the plurality of actions; and
D) performing a process in accordance with the first action to generate a second message to engage with the corresponding session.