Apparatus, system, and method for predicting user intent

Abstract

Various embodiments relate to a method for predicting user intent of a chat robot system, the method comprising the steps of: receiving or acquiring feature data from a plurality of feature sources; classifying the feature data into a plurality of feature groups; extracting the feature data by using a first machine learning layer; and modeling at least one relationship between one set of feature data and another set of feature data in the plurality of feature sets using a second machine learning layer.

Description

Technical Field

[0001] Various aspects of the present disclosure relate to devices, systems and methods for predicting user intent. The present disclosure is particularly applicable to, but not limited to, conversational interfaces, such as chatbot systems, and uses such chatbot systems to predict user intent. Background Art

[0002] The following discussion of the background technology is intended only to facilitate understanding of the present disclosure. It should be understood that the discussion does not acknowledge or admit that any of the material referred to was published, known or was part of the common general knowledge of a person skilled in the art in any jurisdiction before the priority date of the present disclosure.

[0003] Automated dialogue systems such as chatbot systems exist to facilitate human users' navigation on websites such as e-commerce platforms. A typical automated dialogue system may follow a decision tree to generate a machine-to-human chatbot conversation. Such systems typically follow a decision-driven menu when selecting a response to a user input. These systems may analyze one or more user inputs to determine the user's intent and respond using a prepared decision tree response.

[0004] However, current chatbot systems fail to consider the complexity of system behavior caused by multiple input sources, multiple optimization goals or objectives, and different or inconsistent human user behaviors.

[0005] Therefore, there is a need to provide a more robust and / or efficient chatbot system to at least partially address the above shortcomings. Summary of the invention

[0006] Various aspects of the present disclosure relate to methods and systems for predicting the intent of a user of a chatbot system.

[0007] According to one aspect of the present disclosure, a method for predicting user intention of a chatbot system is provided, the method comprising the following steps: receiving or acquiring feature data from multiple feature sources; classifying the feature data into multiple feature groups; extracting the feature data using a first machine learning layer; and modeling at least one relationship between one set of feature data and another set of feature data in the multiple feature groups using a second machine learning layer.

[0008] In some embodiments, the first machine learning layer comprises a first multilayer perceptron neural network and the second machine learning layer comprises a second multilayer perceptron neural network.

[0009] In some embodiments, the method further comprises the step of calculating a probability of intent based on the modeled at least one relationship.

[0010] In some embodiments, the method further includes displaying the intent to the user in the form of one or more suggestions for selection as input to the chatbot system.

[0011] In some embodiments, the method further comprises the step of detecting whether the user has entered the chatbot system based on predetermined user login data.

[0012] In some embodiments, upon a positive determination that the user has entered the chatbot system, the method further includes acquiring feature data associated with the user based on predetermined user login data.

[0013] In some embodiments, the characteristic data associated with a user may include an intent characteristic data set and a user-specific characteristic data set.

[0014] In some embodiments, the user-specific characteristic data includes the user's past transaction data, the user's profile data, and the user's credit status data.

[0015] In some embodiments, the intent feature data includes question text data, single-hotspot intent data, and single-hotspot category data.

[0016] In some embodiments, the feature data includes at least one of a real-time feature and an offline feature.

[0017] In some embodiments, the first machine learning layer is implemented as a multi-gate mixture of experts model.

[0018] According to another aspect of the present disclosure, a system for predicting user intentions of a chatbot system is provided, the system comprising at least one processor configured to model at least one relationship between one set of feature data in a feature group and another set of feature data using a machine learning layer; identify at least one intention of the user based on at least one relationship between one set of feature data in the feature group and another set of feature data; and display the at least one intention to the user in the form of one or more suggestions in a graphical user interface for selection as input to the chatbot system.

[0019] In some embodiments, the machine learning layer includes a multi-layer perceptron neural network.

[0020] In some embodiments, the processor includes a graphical user interface configured to display the intent to the user in the form of one or more suggestions for selection as input to the chatbot system.

[0021] In some embodiments, the processor is configured to detect whether the user has entered the chatbot system based on predetermined user login data.

[0022] In some embodiments, upon a positive determination that the user has entered the chatbot system, the processor is configured to obtain feature data associated with the user based on predetermined user login data.

[0023] In some embodiments, the characteristic data associated with the user includes an intent characteristic data set and a user-specific characteristic data set.

[0024] In some embodiments, the user-specific characteristic data includes the user's past transaction data, the user's profile data, and the user's credit status data.

[0025] In some embodiments, the intent feature data includes question text data, one-hot intent data, and one-hot category data.

[0026] According to another aspect of the present disclosure, there is provided a computer program element comprising program instructions, which, when executed by one or more processors, cause the one or more processors to perform the method.

[0027] According to another aspect of the present disclosure, a computer-readable medium including program instructions is provided. When executed by one or more processors, the program instructions cause the one or more processors to perform the method.

[0028] According to another aspect of the present disclosure, a user interaction system is provided, which includes a chatbot system, and the chatbot system includes a system as described, which is configured as an intention prediction module, the intention prediction module is used to detect a user entering the chatbot system, and provide the intention to the user in the form of one or more suggestions; a typing suggestion module, the typing suggestion module is configured to detect a selected suggestion, and provide one or more further suggestions when the user enters or types the selected suggestion; an intention recognition module, the intention recognition module is configured to detect the user's selected suggestion; the intention recognition module is further configured to generate a response based on the selected suggestion; and a related item suggestion module, the related item suggestion module is configured to display a list of related items, and the related item list is related to the selected suggestion and the response.

[0029] In some embodiments, the related item suggestion module includes a relevance model that is trained to generate a ranked output set based on a degree of relevance between the list of related items and the selected suggestion or response.

[0030] In some embodiments, the related item suggestion module includes a re-ranking module configured to re-rank the ranked output set based on at least one user feedback.

[0031] In some embodiments, the at least one user feedback includes an affirmative or negative determination that the user subsequently selected any one of the related items from the ranked output set.

[0032] In some embodiments, the re-ranking module may be configured to obtain user behavior data for re-ranking.

[0033] It should be noted that embodiments described in the context of a method for predicting user intent are equally valid for a prediction engine, device and / or system, and vice versa. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The invention will be better understood with reference to the detailed description when considered in conjunction with the non-limiting examples and accompanying drawings, in which:

[0035] Figure 1 is a schematic flow chart depicting a method for predicting user intent;

[0036] Figure 2A is a schematic diagram of a system for predicting user intent;

[0037] Figure 2B is a schematic diagram of an intent prediction service arranged to communicate with a chatbot system;

[0038] Figure 2C is a schematic diagram of an example of a feature service configured to receive and / or send feature data from one or more sources of a chatbot system;

[0039] Figure 3 is an example of a graphical user interface for a user of a chatbot system of an e-commerce platform;

[0040] Figure 4A is a schematic diagram of an embodiment of an intent prediction service including two machine learning layers;

[0041] Figure 4B is a schematic diagram showing another embodiment of an intent prediction service modeled as a multi-task model;

[0042] Figure 4C is a schematic diagram illustrating an embodiment of an automatic training pipeline of an automatic training function of an intent prediction training module;

[0043] Figure 4D is a schematic diagram illustrating another embodiment of a prediction service, wherein feature data may be grouped according to an intent feature data group or a user-specific feature data group; and

[0044] Figure 5 is a schematic diagram of data connections between the intent prediction service and one or more other modules of the chatbot system. DETAILED DESCRIPTION

[0045] The following detailed description refers to the accompanying drawings, which illustrate the specific details and embodiments that can practice the present disclosure by way of illustration. These embodiments are described in sufficient detail to enable those skilled in the art to practice the present disclosure. Other embodiments may be utilized and structural or logical changes may be made without departing from the scope of the present disclosure. Various embodiments are not necessarily mutually exclusive, as some embodiments may be combined with one or more other embodiments to form new embodiments.

[0046] Embodiments described in the context of one of an apparatus or method are analogously valid for the other apparatus or method. Similarly, embodiments described in the context of an apparatus are analogously valid for the method as well, and vice versa.

[0047] Features described in the context of one embodiment may be correspondingly applicable to the same or similar features in other embodiments. Features described in the context of one embodiment may be correspondingly applicable to other embodiments, even if not explicitly described in these other embodiments. In addition, additions and / or combinations and / or substitutions described for features in the context of one embodiment may be correspondingly applicable to the same or similar features in other embodiments.

[0048] In the context of various embodiments, the articles “a,” “an,” and “the” used with respect to features or elements include reference to one or more features or elements.

[0049] As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0050] As used herein, the term "chatbot" broadly refers to a computer program or computer executable instructions that simulates an intelligent conversation interface that can conduct an interactive chat session with a human user via an auditory or text method. The chatbot may include one or more virtual agents that are used to conduct intelligent conversations through an online platform for various practical purposes, such as customer service and information delivery. The chatbot may include an independent application or may be integrated with various websites as a platform / tool ​​for providing help to users. In some embodiments, the chatbot may provide a conversational experience that interacts with the user. For example, a user may type a question, and the chatbot will try to explain it and then provide an answer. In order to maximize the relevance and / or accuracy of the answer, the chatbot may need to be trained on different types of inputs received from the user to modify the chatbot's response, which will enhance the customer experience when interacting with the chatbot. In some embodiments, the training of the chatbot may include artificial intelligence, such as a machine learning algorithm or method, and such training may further include supervised learning, unsupervised learning, and / or hybrid learning methods.

[0051] As used herein, the term "processor" includes one or more circuits capable of processing data, such as processing circuits. The processor may include analog circuits or elements, digital circuits or elements, or hybrid circuits or elements. Any other type of implementation of the various functions described in more detail below may also be understood as "circuitry" according to alternative embodiments. A digital circuit may be understood as any type of logic implementation entity, which may be a dedicated circuit or a processor executing software stored in a memory, or firmware.

[0052] As used herein, the term "data" may be understood to include information in any suitable analog or digital form, for example, information provided as a file, a portion of a file, a group of files, a signal or stream, a portion of a signal or stream, a group of signals or streams, etc. However, the term data is not limited to the above examples and may take some form and represent any information understood in the art.

[0053] As used herein, the term "acquire" refers to a processor that actively acquires input or passively receives input from one or more sensors or data sources. The term "acquire" may also refer to a processor that receives or acquires input from a communication interface (e.g., a user interface). The processor may also receive or acquire input via memory, registers, and / or analog-to-digital ports.

[0054] As used herein, the term "intent prediction" includes various forms of prediction, including the use of one or more artificial intelligence-based prediction modules to optimize one or more goals. The one or more goals may include click-through rate prediction, solution prediction, and category prediction.

[0055] As used herein, the term "module" refers to an application specific integrated circuit (ASIC); an electrical / electronic circuit; a combinational logic circuit; a field programmable gate array (FPGA); a processor (shared processor, dedicated processor, or group processor) that executes code; other suitable hardware elements that provide the above functions; or a combination of some or all of the above, such as in a system on a chip, or forming part of or including the above circuits and devices. The term module may include a memory (shared memory, dedicated memory, or group memory) that stores code executed by a processor.

[0056] Hereinafter, each embodiment will be described in detail.

[0057] According to one aspect of the present disclosure and with reference to Figure 1 , a method 100 for predicting the intention of a user or one or more users of a chatbot system is provided, the method comprising the following steps:

[0058] Step S102: receiving or acquiring feature data from multiple feature sources;

[0059] Step S104: classifying the feature data into multiple feature groups;

[0060] Step S106: extracting the feature data using the first machine learning layer; and

[0061] Step S108: using a second machine learning layer to model at least one relationship between one set of feature data and another set of feature data in the plurality of feature groups.

[0062] Method 100 can be implemented in a chatbot system of an e-commerce platform. In some embodiments, method 100 can be executed in an intention prediction module of a chatbot system.

[0063] Figure 2A is a schematic diagram of a system 200 for predicting user intent of a chatbot system, the system comprising at least one processor 202, the processor 202 being configured to receive or acquire feature data from a plurality of feature sources; classify the feature data into a plurality of feature groups; extract the feature data using a first machine learning layer; and model at least one relationship between one set of feature data in a feature group and another set of feature data using a second machine learning layer.

[0064] Processor 202 may be a server computer or a portion thereof. The server computer may be a single server or include multiple distributed servers. In some embodiments, the server computer may be a remote server, such as a cloud server.

[0065] The processor 202 may include an input / output interface 204, a prediction engine 206, and one or more databases 208. The I / O interface 204 may include various hardware and / or software to obtain or receive feature data from one or more feature data sources 210. In some embodiments, the I / O interface 204 may include a wired or wireless communication interface to obtain or receive feature data 212 from the feature data source 210.

[0066] The prediction engine 206 may be an artificial intelligence-based predictor, in particular a machine learning-based predictor. The prediction engine 206 may include a first machine learning layer that performs step 106 and a second machine learning layer that performs step 108.

[0067] The processor 202 may include one or more databases 208 for storing different types of data. In some embodiments, the one or more databases 208 may be configured to store at least one of the following: feature data 212 used as input to the prediction engine 206, training data for training the first machine learning layer and / or the second machine learning layer, system-related data, etc.

[0068] Figure 2BAn embodiment of a prediction engine 206 in communication with a chatbot system is shown. The chatbot system may include various components numbered 211 to 220. The chatbot system may include a front end 212 for users 1000 to interact with and access the chatbot system. The front end 212 may be linked to a back end 213. The back end 213 may be configured to send and receive feature data from a feature service 215. The chatbot system may include a testing system 211, such as an AB testing system, for testing the online performance of different strategies or models associated with different modules of the chatbot system.

[0069] The feature service 215 may be configured to send and receive feature data from the API service 214 and the key value (KV) store (Redis) database 216. The KV store database may in turn be configured to receive streaming data from a data processing framework and / or embeddable library such as, but not limited to, the Flink & Kafka framework 217. The backend 213 may also be arranged to perform data communication with the Flink & Kafka framework 217 to receive streaming data from it and / or send streaming data to it.

[0070] The intent prediction service 206 may be configured to be trained periodically, at each predetermined time interval, or when needed. In some embodiments, the intent prediction service 206 may be trained continuously. The prediction service 206 may obtain data from the model storage database 220. The model storage database 220 may be configured to store corresponding predicted output intents for various input feature data. The model storage database 220 may be configured to receive trained data from the intent prediction training module 219.

[0071] The intention prediction training module 219 may be configured to receive input data for automatic training. The input data may be stored in the log database 218 .

[0072] The log database 218 may include historical chatbot log entries. In some embodiments, the log database 218 may be a distributed fault-tolerant data warehouse system. The log database 218 may be configured to facilitate relatively large-scale analysis performance. In some embodiments, the log database 218 may be configured as an Apache TM HIVE database.

[0073] Figure 2CAn example of a feature service 215 is shown. The feature service 215 can be configured to send feature data to one or more modules of the chatbot system in response to one or more data requests from one or more modules. One or more modules may include an intent prediction module 206, a typing suggestion module 232, and other modules 234, and other modules 234 may include an intent recognition module, a related item module, etc. The feature service 215 can be configured to process data from multiple sources in one or more different data formats. For example, feature data from an interface such as an application programming interface (API) 236, a real-time data stream 240 (preprocessed using Flink 239), and offline data 241 (preprocessed using Spark 242) can be stored in a KV repository 238.

[0074] The feature service 215 also provides an easy-to-use interface 236 that is connected in data communication with one or more other shared services 243 of the chatbot system service to send and / or collect data from the other services 243.

[0075] In some embodiments, feature service 215 can handle various I / O complexities based on a unified data structure obtained from real-time data stream 240, offline data 241, and application programming interface (API) 236. In some embodiments, this unified data structure can be implemented using a high-level programming language, such as but not limited to the Python programming language.

[0076] Figure 3 A graphical user interface (GUI) 300 of a user of a chatbot system of an e-commerce platform is shown. Before a user enters the chatbot system via the GUI 300, a login page may be provided to the user to provide necessary user identifiers, passwords, added security information / further authentication information.

[0077] GUI 300 can display chatbot information 302 and display a customized list of predicted intents. In some embodiments, the predicted intent can be one or more suggestions in the form of one or more questions 304a as shown in GUI 300. It should be understood that the user of the chatbot system can select the predicted intent to enter into the input text box 306. Any entry selected from the predicted intent can be marked as an "expected entry" and used for future intent predictions. In some embodiments, the predicted intent may include one or more interactive buttons 304b as shown in GUI 300, or may include other interface elements 304c shown just above the input text box 306.

[0078] It is worth noting that the predicted intent is displayed before the user enters any input into the input text box 306.

[0079] Next, we will refer to Figure 2A , Figure 2B and Figure 3 Embodiments of the operation of method 100 and system 200 are described. In some embodiments, the chatbot system may form part of an e-commerce platform. Therefore, the predicted intent may be related to an action, function, product or service provided by the e-commerce platform.

[0080] The user 1000 can access the front end 212 of the chatbot system using his / her login data / credentials, which can include two or more levels of authentication. Such authentication can include biometric authentication, password-based authentication, etc. The back end 213 can be configured to detect whether the user has entered the chatbot system based on the login data. In a positive determination that the user 1000 has entered the chatbot system, feature data associated with the user based on the login data can be used.

[0081] After the user 1000 successfully logs into the chatbot system, the front end 212 is configured to display a GUI 300 with chatbot information / details 302 (such as the name of the chatbot, e.g., "Sophie"). The back end 213 is configured to retrieve feature data related to the user 1000 from the feature service 215. Examples of feature data may include historical transactions performed by the user 1000 using the e-commerce platform via the chatbot system and / or other interfaces, previous records of the user 1000 using the chatbot, one or more intents (e.g., clicks) previously selected by the user 1000, user actions, or transactions on the e-commerce platform, categories of goods and / or services preferred by the user, and the user's current payment status, such as whether the user 1000 has an overall credit or debit with respect to the e-commerce platform.

[0082] Typically, feature data can be grouped between intent feature data groups, user-specific feature data groups, or other groupings. User-specific feature data includes, for example, a user's past transaction data, user profile data, and a user's credit status data. Intent feature data can include, for example, question text data, single-hotspot intent data, and single-hotspot category data.

[0083] It is worth noting that the feature data may include real-time feature data and offline feature data.

[0084] In some embodiments, the retrieved feature data may be sent to the intent prediction service 206 for predicting one or more intents 304. Prediction may include comparing one or more candidate predictions with the data retrieved from the model repository 220. In some embodiments, the intent prediction service may include an artificial neural network, and the predicted intent may tag the data, the tag data including one or more questions that the user may think of based on the input feature data. In some embodiments, based on the comparison with the tag data, a set of candidate predictions with the minimum error or error metric will be considered for displaying the intent 304 to the user for selection.

[0085] Figure 4A 4 is a specific embodiment 400 of the intent prediction service 206, including two machine learning layers 420, 440.

[0086] The first machine learning layer 420 may be a first multi-layer perceptron neural network (MLP), and the second machine learning layer 440 may be a second MLP 440 .

[0087] The first machine learning layer 420 may include a plurality of MLPs, labeled 420a, 420b, 420c, etc. The MLP 420a may be configured to receive feature data 410 associated with various categories or categorical feature data, such as feature data associated with order category 410a, feature data associated with checkout category 410b, and chatbot log 410c, etc.

[0088] The feature data 410a, 410b may be continuous data based on attention to an order or checkout. MLP 420a, 420b may be used to extract feature data from each group 410a, 410b, 410c. In some embodiments, for feature data related to user profiles and other basic features, an embedding layer may be used to extract a vectorized representation.

[0089] Extracting feature data from each group 410a, 410b, 410c can be performed via corresponding attention pooling steps 430a, 430b, 430c. The extracted feature data can then be cascaded to form an input set that enters the second machine learning layer 440.

[0090] The second machine learning layer 440 may include an MLP 440 for modeling at least one relationship between one set of feature data and another set of feature data in the plurality of feature groups 440a, 440b, 440c.

[0091] The output of the second MLP 440 may be input to a softmax function module 450 to calculate a probability value to select the most likely intent set.

[0092] Figure 4B Another embodiment of the intent prediction service 206 modeled as a multi-task model 400A is shown. The multi-task model 400A can be referred to as a multi-gate mixture of experts model and includes feature data 410, a first MLP in the form of an attention gate encoder 426, and second MLPs 440a, 440b, etc.

[0093] The attention-based encoder 426 is configured to extract feature representations, and the MLPs 440a, 440b can be configured to model feature interactions to provide a final output. In some embodiments, each of the multiple feature encoders 426 can be considered an expert. In a system of N experts (N is an integer greater than 0), the N experts can be configured to output N different feature representations. For different tasks, the gate attention 430 can be used to extract representations from different experts 426. If the tasks are highly related, each gate attention 430 can be trained to use features from each expert 426.

[0094] Figure 4C FIG. 2 shows an embodiment of an automatic training pipeline of the automatic training function of the intention prediction training module 219. The automatic training function may include an online process and an offline process. Figure 4C As shown, the offline process can start by acquiring feature data from the log database 218. The feature data is used for data extraction, model training, and performance checking before being uploaded to the model repository 220. The intent prediction service 206 can then download the model and load it into its data pipeline to replace the old (current) pipeline after performing a performance check.

[0095] In some embodiments, the performance check may include statistical measurements, which may include accuracy indicators, such as root mean square (RMS) correlation measurements.

[0096] In some embodiments, intent prediction models are developed using non-limiting examples of such accuracy metrics, including Python functions such as precision@k, recall@k, NDCG@k, and the like.

[0097] Figure 4D Another embodiment of the prediction service 206 is shown, where feature data may be grouped according to intent feature data groups or user-specific feature data groups.

[0098] The intent feature data group may include hot question text data 413, single hotspot intent data 411, and single hotspot category data 412. These data may be labeled according to whether the user 1000 has clicked on these data before. In some embodiments, any predicted intent including label data may be preprocessed to eliminate any label data bias.

[0099] Each of the intent group data 411, 412, 413 may be further pre-processed before being input into the MLP 420. Figure 4D In the illustrated embodiment, the single-hotspot intent vector 411 and the single-hotspot category vector 412 may be embedded via an embedding layer 414 to form intent embedding data 416 and category embedding data 417. The hot question text data 413 may be input to a pre-trained language model 415 to form text embedding data 418. The three embedding data 416, 417, 418 may be combined as input to the MLP 420.

[0100] The feature data extracted from the MLP 420 may be combined with the user-side features and input into the MLP 440 .

[0101] It is worth noting that the training of MLP 420, 440 may require a large amount of high-quality training data. In some embodiments, the training data can be significantly reduced or simplified based on feature data associated with click-through rate (CTR). In some embodiments, CTR can be used to measure the performance of the predicted intent. For example, if a large number of users of the chatbot system always select a specific intent, then the intent will have a relatively high CTR, which will make the predicted intent more likely to be displayed as an intent on GUI 300.

[0102] In some embodiments, predicted intent 304 may include a set of questions categorized based on order, transaction, delivery, or other issues.

[0103] In some embodiments, the predicted intents 304 may be further ranked for display (ranked intent list), where higher ranked tag data in the predicted intents 304 may be displayed in higher priority slots within the GUI 300. In some embodiments, the ranked intent list may also be sorted according to a predefined policy. In one such embodiment, hot issues may be prioritized over shortcuts, followed by category hot issue slots.

[0104] In some embodiments, the intent prediction service 206 may be configured to receive the user's selected intent (label data). The selected intent may be used to update the database 218 to train the intent prediction training module 219.

[0105] In some embodiments, the intent prediction service 206 can be configured to receive any input from any user 1000 that is not selected from the list of label data. Such input can be in the form of typed input in the text box 306. The typed input can be used to update the database 218 to train the intent prediction training module 219.

[0106] In some embodiments, upon detecting that the user 1000 logs into the chatbot system using his / her login credentials, real-time feature data may be retrieved from the feature service 215. Such real-time feature data may include the latest order status and account status associated with the user 1000. It is contemplated that the real-time feature data of the present disclosure may primarily include structured data.

[0107] In some embodiments, when user 1000 conducts multiple transactions in the form of multiple orders at the same time, user 1000 may be related to only one or a few of the orders. In such an operation example, model 400A may be appropriate. In particular, the attention mechanism of model 400A may be applicable to multiple orders. For each order, the data is extracted and mapped using an embedding layer, and a vector is used to represent the order status. In the case of several orders, there is a list of vectors, each representing an order. Then, attention pooling can be used to select one or two or several of the most relevant order information.

[0108] In such Figure 5 In some of the illustrated embodiments, the intent prediction system 200 may be configured to communicate data with at least one of the following modules: a typing suggestion module 502, an intent recognition module 504, and a related item suggestion module 506. The intent prediction system 200 may communicate data directly with each of the modules 502, 504, 506, as shown, or may be arranged to communicate data indirectly via a remote network (e.g., a cloud service). Each of the typing suggestion module 502, the intent recognition module 504, and the related item suggestion module 506 may be configured to be data connected to each other.

[0109] The typing suggestion module 502 may be configured to detect an entry in the text box 306 and provide one or more further suggestions as the user 1000 enters or types his / her entry.

[0110] When user 1000 selects an intent to output from intent prediction system 200 or types into text box 306 , intent recognition module 504 may be used to detect his / her selected intent. Intent recognition module 504 is configured to generate a response based on the data entered into text box 306 .

[0111] The related item suggestion module 506 can be configured to display a list of further suggestions (e.g., related items) related to the user entry in the text box 306. After the chatbot provides one or more responses to the user, the list of further suggestions can be displayed to the user. In some embodiments, the related item suggestion module 506 can be configured to detect and parse the responses provided by the chatbot to the user. In some embodiments, one or more responses can be solutions to solve problems faced by the user. The related item suggestion module 506 can be configured to extract one or more keywords from the user entry in the text box 306 to provide further suggestions that may be of interest to the user.

[0112] In some embodiments, the related item suggestion module 506 may include an offline or real-time database for text mining or identifying related items based on user input. In some embodiments, the related item suggestion module 506 may include a relevance model that is trained to generate a ranked output set based on the degree of relevance between the related items and the input data entered into the text box 306.

[0113] In some embodiments, the related item suggestion module 506 may include a re-ranking module. The re-ranking module may be configured to re-rank the ranked output set based on at least one user related feedback, such as whether the user subsequently considers any related items to be relevant and enters the same. Any related items that the user subsequently does not select after a predetermined number of displays for selection, or any related items that the user explicitly indicates are not relevant, such as via a feedback form, may be re-ranked to a lower ranking.

[0114] In some embodiments, the re-ranking module may be configured to obtain user behavior data for ranking / re-ranking. Such behavior data may be based on feature data associated with the user profile.

[0115] The methods described herein may be performed, and the various processing or computing units and devices and computing entities described herein may be implemented by one or more circuits. In one embodiment, a "circuit" may be understood as any type of logic implementation entity, which may be hardware, software, firmware, or any combination thereof. Thus, in one embodiment, a "circuit" may be a hardwired logic circuit or a programmable logic circuit, such as a programmable processor, e.g., a microprocessor. A "circuit" may also be software implemented or executed by a processor, such as any type of computer program, such as a computer program using virtual machine code. Any other type of implementation of the various functions described herein may also be understood as a "circuit" according to an alternative embodiment.

[0116] Although the present disclosure has been specifically shown and described with reference to various specific embodiments, it will be understood by those skilled in the art that various changes may be made in form and detail without departing from the spirit and scope of the present invention as defined by the appended claims. Therefore, the scope of the present invention is indicated by the appended claims and all changes in the meaning and equivalent scope of the claims are therefore intended to be included therein.

Claims

1. A method for predicting the intention of a user of a chatbot system, the method comprising the following steps: receiving or acquiring feature data from a plurality of feature sources; classifying the feature data into a plurality of feature groups; Extracting the feature data using a first machine learning layer; and At least one relationship between one set of feature data and another set of feature data in the plurality of feature groups is modeled using a second machine learning layer.

2. The method according to claim 1, wherein: The first machine learning layer includes a first multi-layer perceptron neural network, and the second machine learning layer includes a second multi-layer perceptron neural network.

3. The method according to claim 1 or 2 further comprises the step of calculating the probability of the intention based on at least one relationship modeled.

4. The method of claim 3, further comprising displaying the intent to the user in the form of one or more suggestions for selection as input to the chatbot system.

5. The method according to any one of the preceding claims, further comprising the step of detecting whether the user has entered the chatbot system based on predetermined user login data.

6. The method according to claim 5, wherein: In a positive determination that the user has entered the chatbot system, the method further includes acquiring feature data associated with the user based on the predetermined user login data.

7. The method according to claim 6, wherein: The characteristic data associated with the user includes an intention characteristic data set and a user-specific characteristic data set.

8. The method according to claim 7, wherein: The user-specific characteristic data includes the user's past transaction data, user profile data, and the user's credit status data.

9. The method according to claim 7 or 8, wherein: The intent feature data includes question text data, single hotspot intent data and single hotspot category data.

10. The method according to any one of claims 1 or 2, wherein: The feature data includes at least one of a real-time feature and an offline feature.

11. A method according to any one of the preceding claims, wherein: The first machine learning layer is implemented as a multi-gate mixture of experts model.

12. A system for predicting the intent of a user of a chatbot system, the system comprising at least one processor configured to modeling, using a machine learning layer, at least one relationship between one set of feature data and another set of feature data in the set of features; identifying at least one intention of the user based on the at least one relationship between one set of feature data and another set of feature data in the feature group; and The at least one intent is displayed to the user in the form of one or more suggestions in a graphical user interface for selection as input to the chatbot system.

13. The system according to claim 12, wherein: The machine learning layer includes a multi-layer perceptron neural network.

14. A system according to any one of claims 12 or 13, wherein: The processor is configured to detect whether the user has entered the chatbot system based on predetermined user login data.

15. The system of claim 14, wherein: In an affirmative determination that the user has entered the chatbot system, the processor is configured to obtain feature data associated with the user based on the predetermined user login data.

16. The system of claim 15, wherein: The characteristic data associated with the user includes an intention characteristic data set and a user-specific characteristic data set.

17. The system of claim 16, wherein: The user-specific characteristic data includes the user's past transaction data, user profile data, and the user's credit status data.

18. The system according to claim 16 or 17, wherein: The intent feature data includes question text data, single hotspot intent data and single hotspot category data.

19. A computer program element comprising program instructions which, when executed by one or more processors, cause the one or more processors to perform the method of any one of claims 1 to 11.

20. A computer readable medium comprising program instructions which, when executed by one or more processors, cause the one or more processors to perform the method of any one of claims 1 to 11.

21. A user interaction system comprising a chat robot system, the chat robot system comprising The system according to claims 12 to 18, the system being configured as follows: an intention prediction module, the intention prediction module being configured to detect a user entering the chatbot system and providing the intention to the user in the form of one or more suggestions; a typing suggestion module configured to detect the selected suggestion and provide one or more further suggestions when the user enters or types the selected suggestion; an intent recognition module configured to detect a selected suggestion of the user, the intent recognition module further configured to generate a response based on the selected suggestion; as well as A related item suggestion module is configured to display a list of related items that are related to the selected suggestion and the response.

22. The system of claim 21, wherein: The related item suggestion module includes a relevance model trained to generate a ranked output set according to a degree of relevance between the list of related items and the selected suggestion or the response.

23. The system of claim 22, wherein: The related item suggestion module includes a re-ranking module configured to re-rank the ranked output set based on at least one user feedback.

24. The system of claim 23, wherein: The at least one user feedback includes a positive or negative determination that the user subsequently selected any one of the related items from the ranked output set.

25. A system according to claim 23 or 24, wherein: The re-ranking module may be configured to obtain user behavior data for the re-ranking.

Images