WO2024114374A1 - 服务处理方法及装置 - Google Patents

服务处理方法及装置 Download PDF

Info

Publication number
WO2024114374A1
WO2024114374A1 PCT/CN2023/131653 CN2023131653W WO2024114374A1 WO 2024114374 A1 WO2024114374 A1 WO 2024114374A1 CN 2023131653 W CN2023131653 W CN 2023131653W WO 2024114374 A1 WO2024114374 A1 WO 2024114374A1
Authority
WO
WIPO (PCT)
Prior art keywords
service
session
text
conversation
information
Prior art date
Application number
PCT/CN2023/131653
Other languages
English (en)
French (fr)
Inventor
杨崇
许婧
王永亮
杨帆
郑艳兰
Original Assignee
蚂蚁财富(上海)金融信息服务有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 蚂蚁财富(上海)金融信息服务有限公司 filed Critical 蚂蚁财富(上海)金融信息服务有限公司
Publication of WO2024114374A1 publication Critical patent/WO2024114374A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/332—Query formulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35—Clustering; Classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/205—Parsing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/06—Asset management; Financial planning or analysis

Definitions

  • This document relates to the field of data processing technology, and in particular to a service processing method and device.
  • One or more embodiments of the present specification provide a service processing method, including: obtaining historical session information and input session text of a user in a resource management service. Performing intent recognition on the session text to obtain service intent, and performing feature analysis on the historical session information and the session text to obtain service key features. Determining a service action sequence of the session text based on the service intent, the service key features, and the session behavior features of the user. Executing service processing corresponding to the service action sequence to respond to the session text.
  • One or more embodiments of the present specification provide a service processing device, including: a session information acquisition module, configured to acquire the user's historical session information and input session text in a resource management service.
  • a feature parsing module configured to perform intent recognition on the session text to obtain the service intent, and perform feature parsing on the historical session information and the session text to obtain service key features.
  • a service action determination module configured to determine a service action sequence of the session text based on the service intent, the service key features, and the user's session behavior features.
  • a service processing module configured to execute service processing corresponding to the service action sequence in response to the session text.
  • One or more embodiments of the present specification provide a service processing device, including: a processor; and a memory configured to store computer executable instructions, wherein when the computer executable instructions are executed, the processor: obtains the user's historical session information and input session text in the resource management service. Performs intent recognition on the session text to obtain the service intent, and performs feature analysis on the historical session information and the session text to obtain service key features. Determines a service action sequence for the session text based on the service intent, the service key features, and the user's session behavior features. Executes service processing corresponding to the service action sequence in response to the session text. Book.
  • One or more embodiments of the present specification provide a storage medium for storing computer executable instructions, which implement the following process when executed by a processor: obtaining the user's historical session information and input session text in a resource management service. Performing intent recognition on the session text to obtain the service intent, and performing feature analysis on the historical session information and the session text to obtain service key features. Determining the service action sequence of the session text based on the service intent, the service key features, and the user's session behavior features. Executing the service processing corresponding to the service action sequence to respond to the session text.
  • FIG1 is a processing flow chart of a service processing method provided by one or more embodiments of this specification.
  • FIG2 is a schematic diagram of a model training process of a service action model provided by one or more embodiments of this specification
  • FIG3 is a schematic diagram of a processing process of a service processing provided by one or more embodiments of this specification.
  • FIG4 is a processing flow chart of a service processing method applied to a financial service scenario provided by one or more embodiments of this specification;
  • FIG5 is a schematic diagram of a service processing device provided by one or more embodiments of this specification.
  • FIG. 6 is a schematic diagram of the structure of a service processing device provided by one or more embodiments of this specification.
  • a service processing method embodiment provided in this specification is based on the The service action sequence is determined based on the conversation behavior characteristics and the service intent and service key characteristics obtained by parsing the historical conversation information and conversation text of the user in the resource management service, and the service processing corresponding to the service action sequence is executed. Specifically, the historical conversation information and the input conversation text of the user in the resource management service are parsed to obtain the service intent and service key characteristics, and the service action sequence of the conversation text is determined based on the service intent, service key characteristics and conversation behavior characteristics, and the conversation text is responded to by executing the service processing corresponding to the service action sequence.
  • the service action sequence is determined based on the user's conversation behavior characteristics, and the flexibility and diversity of the determined service action sequence are improved. Different service action sequences can be determined for different users to meet the diverse needs of users. In addition, by determining the service action sequence from multiple aspects such as intent, key characteristics, and conversation behavior characteristics, the effectiveness and accuracy of the service action sequence are improved.
  • the service processing method provided in this embodiment specifically includes steps S102 to S108 .
  • Step S102 Acquire the user's historical session information and input session text in the resource management service.
  • the resource management service described in this embodiment refers to a service that manages the resources deposited by users so that users can obtain benefits.
  • the specific resource management service may be a financial management service, such as a financial management service of a financial institution that combines online and offline, a financial management service of an online financial institution, or a financial management service in a third-party payment platform.
  • the resource management service may include one or more sub-services, such as a balance inquiry service, a financial product service, a protection service (insurance service), etc.
  • the resource management service may also be other services related to resource management.
  • the conversation text refers to the conversation text of the current round input by the user during the entire conversation process.
  • the conversation process can be for a conversation unit, and the conversation unit includes a conversation from the beginning to the end.
  • multiple rounds of conversation interactions can be performed;
  • the conversation text may be the conversation text of the current round in multiple rounds of conversation interactions, for example, the conversation text of the current round input by the user is "What should I do if the new energy financial product I bought falls in value?";
  • the historical conversation information refers to the conversation information of the previous round or previous rounds of the conversation text of the user during the entire conversation process.
  • the historical conversation information includes at least one of the following: historical conversation text, historical service action sequence, historical response text, historical service intention, and historical service key features; in addition, the historical conversation information may also include other types of conversation information.
  • a conversation interaction interface can be set for the resource management service.
  • a conversation text input control can be configured in the conversation interaction interface. Users can input conversation text by triggering the conversation text input control.
  • the conversation interaction interface can also be configured with a conversation voice input control. The user triggers the conversation voice input control to input the conversation voice, perform voice recognition on the conversation voice, obtain the conversation text, and perform conversation interaction through multiple input methods such as text and voice to adapt to various conversation interaction scenarios.
  • the user's historical session information and input session text in the resource management service are obtained.
  • the currently input session text is processed, but also the historical session information of the user in the resource management service is processed, so as to improve the comprehensiveness and accuracy of session processing and enhance the user experience.
  • a user may enter the same conversation text several times for a certain matter in the resource management service. For example, the user enters the conversation text "What should I do if the price of my new energy financial product falls?" After the user receives the reply content to the conversation text, the user enters the conversation text "What should I do if the price of my new energy financial product falls?" again.
  • a response processing may be performed to the user for the conversation text based on the text status of the conversation text and the historical conversation text in the historical conversation information. Specifically, when the text status is that the conversation text is different from the historical conversation text, the following step S104 is executed.
  • the following operation is also performed: determining whether the session text and the historical session text in the historical session information are the same; if not, executing the following step S104; if so, calculating the number of global texts according to the first number of the session texts and the second number of the historical session texts, updating the historical service action sequence in the historical session information based on the global text number, and generating a response text of the session text according to the updated service action sequence.
  • the conversation text and the historical conversation text in the historical conversation information are the same, the sum of the first number of conversation texts and the second number of historical conversation texts is calculated as the global text number, the user's emotion category is determined based on the global text number, and the historical service action sequence of the historical conversation text adjacent to the conversation text in the historical conversation information is updated based on the emotion category, and a response text for the conversation text is generated according to the updated service action sequence.
  • the emotion category includes the first category, the second category and/or the third category.
  • the number of global texts corresponds to the emotion category.
  • the user's emotion category is the first category
  • m ⁇ the number of global texts ⁇ n the user's emotion category is the second category
  • the number of global texts is ⁇ o
  • the user's emotion category is the third category.
  • the conversation text is the same and the number of global texts is 2.
  • the user's emotion category is the second category
  • the historical service action sequence is "market interpretation->profit attribution->suggestions”.
  • the emotion category is the second category
  • the historical service action sequence is updated to obtain an updated service action sequence of "emotional comfort->market interpretation->profit attribution->suggestions"
  • a response text for the conversation text is generated according to the updated service action sequence.
  • the above-mentioned operation performed when the conversation text and the historical conversation text in the historical conversation information are the same can be replaced by calculating the global text quantity based on the first quantity of the conversation text and the second quantity of the historical conversation text; determining the user's emotion category based on the global text quantity, and updating the historical service action sequence in the historical conversation information based on the emotion category; generating a response text for the conversation text according to the updated service action sequence, and forming a new implementation method with other processing steps provided in this embodiment; or, it can also be replaced by calculating the global text quantity based on the first quantity of the conversation text and the second quantity of the historical conversation text, and updating the historical service action sequence in the historical conversation information based on the global text quantity, and executing the service processing corresponding to the updated service action sequence to respond to the conversation text, and forming a new implementation method with other processing steps provided in this embodiment.
  • step S102 can be replaced by obtaining the user ID, historical session information and input session text of the user in the resource management service, and forming a new implementation method with other processing steps provided in this embodiment; wherein the user ID includes the user account of the user in the resource management service.
  • Step S104 performing intent recognition on the conversation text to obtain service intent, and performing feature analysis on the historical conversation information and the conversation text to obtain service key features.
  • the session text is subjected to intent recognition to obtain the service intent, and the historical session information and session text are subjected to feature analysis to obtain the key service features.
  • the service intent described in this embodiment refers to the user's demand or purpose for the resource management service.
  • the service intent is related to the resource management service and obtains the user's demand from the "coarse-grained" level.
  • the conversation text is "What should I do if the price of my new energy financial product falls", and the service intent is to recommend new energy financial products.
  • the service key feature refers to the service intent based on the conversation text or the key feature information related to the resource management service obtained from the conversation text.
  • the service key feature includes service key items and/or service key information.
  • the conversation text is "What should I do if the price of my new energy financial product falls"
  • the service intent is to recommend new energy financial products
  • the service key item is the type of financial product
  • the service key information is new energy.
  • an intent recognition model in the process of performing intent recognition on conversation text to obtain service intent, in order to improve the recognition efficiency of intent recognition, an intent recognition model can be introduced, and the conversation text can be input into the intent recognition model for intent recognition to obtain service intent.
  • the conversation text and historical conversation information can also be input into the intent recognition model for intent recognition to obtain service intent, that is, the conversation text can also be subjected to intent recognition based on historical conversation information to obtain service intent.
  • a preset service intent set can be set for the resource management service, and the process of obtaining the service intent by performing intent recognition on the conversation text can also be achieved in the following way: extracting service keywords associated with the resource management service from the conversation text, and matching the service keywords with the preset service intent in the preset service intent set to obtain the preset service intent that matches the service keywords, and using the preset service intent as the service intent.
  • the conversation text is "What should I do if the value of my new energy financial product falls?"
  • the service keyword associated with the resource management service extracted from the conversation text is "new energy financial product”
  • the preset service intent matching the service keyword is "new energy financial product recommendations”, which is the service intent.
  • the key item set refers to the set of service key items corresponding to each service intent set for the resource management service
  • the service key items refer to the key items further mined at the "fine-grained" level, and the service key items include service filling items, such as financial product categories and financial product purchase time
  • the service key information refers to the key information corresponding to the service key items, and the service key information includes service filling information, such as the service key item is the financial product category, and the service key information is new energy.
  • the service intent is new energy financial product recommendations.
  • the key item set includes a set of service key items corresponding to liquor financial product recommendations and service key items corresponding to new energy financial product recommendations.
  • the service key items "financial product type, financial product purchase time, and financial product purchase amount” corresponding to the new energy financial product recommendations are searched in the key item set.
  • the service keyword "new energy financial product” extracted from the conversation text based on the service intent is first converted to obtain the service key item "financial product type", and then the service key items “financial product purchase time and financial product purchase amount” corresponding to the new energy financial product recommendations are searched in the key item set.
  • Based on the historical conversation information, conversation text, and service key items "financial product type, financial product purchase time, and financial product purchase amount” Identify key service information.
  • this embodiment provides an optional implementation manner in which the service key information is determined in the following manner: extracting the service key information from the historical session information and the session text based on the service key items; if the service key information is not extracted, querying the corresponding service key information in the database based on the service key items; if the service key information is extracted, using the extracted service key information as the determined service key information.
  • the service key items are "financial product type, financial product purchase time, financial product purchase amount”.
  • the service key information "new energy” is extracted from the historical session information and session text.
  • the service key item “financial product purchase time, financial product purchase amount” is not extracted from the historical session information and session text. Then, the service key information corresponding to the service key item “financial product purchase time, financial product purchase amount” is queried in the database as "x month x day, xx million yuan”.
  • the process of performing intent recognition on conversation text to obtain service intent and performing feature analysis on historical conversation information and conversation text to obtain key service features can be implemented by an NLU (Natural Language Understanding) module.
  • the specific NLU module may include an intent recognition model and/or a feature analysis model.
  • the intent recognition model can perform intent recognition on conversation text to obtain service intent
  • the feature analysis model can perform feature analysis on historical conversation information and conversation text to obtain key service features.
  • an optional implementation method provided in this embodiment is to perform intent recognition on the session text to obtain service intent, and perform feature analysis on the historical session information and session text to obtain service key features.
  • the following operation is also performed: determine whether the service intent and the service key features match the historical session state in the historical session information; if not, perform correction processing on the historical session state in the historical session information according to the service intent and the service key features, and on this basis, perform the following step S106; if so, perform the following step S106.
  • the historical session state includes historical service intentions and/or historical service key features.
  • the historical session state may also include historical session texts and/or historical service action sequences.
  • the process of determining whether the service intent and service key features match the historical session state in the historical session information can be performed by determining whether the service intent and/or service key features match the historical session state in the historical conversation information.
  • the historical service intent and/or historical service key features are consistent; the process of correcting the historical session state in the historical session information according to the service intent and service key features can be achieved by correcting the historical service intent and/or historical service key features in the historical session state to the service intent and/or the service key features.
  • step S104 can be replaced by performing intent recognition on the conversation text based on historical conversation information to obtain the service intent, and performing feature analysis on the historical conversation information and/or conversation text to obtain the key service features, and forming a new implementation method with the other processing steps provided in this embodiment; or, it can also be replaced by parsing the historical conversation information and/or conversation text to obtain the service intent and/or key service features, and forming a new implementation method with the other processing steps provided in this embodiment.
  • Step S106 determining a service action sequence of the conversation text according to the service intent, the key service features and the conversation behavior features of the user.
  • the above-mentioned analysis of the conversation text and historical conversation information obtains the service intent and service key features.
  • the conversation text is used to identify the intent to obtain the service intent
  • the historical conversation information and conversation text are used to perform feature analysis to obtain the service key features.
  • the service action sequence of the conversation text is determined based on the service intent, service key features and the user's conversation behavior features.
  • the session behavior feature described in this embodiment refers to the behavior feature information of the user conducting a session in the resource management service.
  • the session behavior feature includes session preferences and/or historical session information.
  • the session preference refers to the resource management preference or service preference of the user conducting resource management in the resource management service or the session preference of the user conducting a session in the resource management service.
  • the session preference includes personality characteristics and/or emotional categories, such as session preferences are divided into aggressive and robust types, or session preferences are divided into optimistic and pessimistic types.
  • the session behavior feature is determined in the following manner: extracting keywords from user attribute information and resource management records; calculating a matching degree based on the keyword and the preference feature of the preset session preference, determining the user's session preference based on the matching degree, and using the session preference and the historical session information as the session behavior feature.
  • the process of determining the user's session preference based on the matching degree may be implemented by taking the preset session preference as the user's session preference if the matching degree is greater than a matching degree threshold.
  • the service action sequence refers to a sequence of service actions that respond to the conversation text in the resource management service.
  • the service action sequence may contain one or more service actions. For example, if the conversation text is "What should I do if my new energy financial product falls", the service action sequence is "fact acceptance->emotional soothing->market interpretation-> Suggestions", for another example, the session text is "open a (representing a subservice of the resource management service)", and the service action sequence is "request confirmation->rendering entry", “request confirmation->jump” or "jump”.
  • a service action model can be introduced to determine the service action sequence of the conversation text.
  • the following operations are performed: the service intent, the service key features, the user's conversation preferences and the historical conversation information are input into the service action model to determine the service action and obtain the service action sequence; optionally, the conversation preference is determined based on the user attribute information and the resource management record of the user in the resource management service, or the conversation preference is determined based on the user portrait information, and the user portrait information includes the user attribute information and/or the user's resource management record in the resource management service.
  • the user portrait information is obtained by reading based on the user identification, and specifically, it can be obtained by the DST (Dialogue State Tracking) module from an external database based on the user identification, and the user identification can be the user account of the user in the resource management service.
  • DST Dialogue State Tracking
  • the user attribute information refers to information related to user attributes, and the user attribute information includes but is not limited to: the length of survival from birth to the calculation time, occupation, and working hours;
  • the resource management record refers to the behavior information record of resource management in the resource management service, such as the financial management record in the financial management service.
  • the specific execution process of determining the service action sequence of the conversation text according to the service intent, service key features and conversation behavior features can be replaced by inputting the user portrait information into the first network of the service action model to determine the conversation preference, obtain the conversation preference, and input the service intent, service key features, the conversation preference and historical conversation information into the second network of the service action model to determine the service action, obtain the service action sequence of the conversation text, and form a new implementation method with other processing steps provided in this embodiment.
  • the model training of the service action model can be completed by using hierarchical reinforcement learning technology, and the service action model can use BCQ (Batch Constrained deep Q-learning, offline reinforcement learning algorithm).
  • model training can be performed in advance to obtain a service action model. Since the virtual environment construction or online training costs are relatively high, in order to reduce the training costs, the model to be trained can be trained by offline training to obtain a service action model.
  • the service action model is trained in the following manner: sample session information is input into the model to be trained to calculate session indicators to obtain session indicators; parameters of the model to be trained are updated according to the session indicators, sample session preferences and the sample session information; optionally, the sample session preferences are obtained after the sample session information is input into a preference detection model for preference detection.
  • the sample session information may include one or more session sequences, and the session indicator refers to An indicator of the confidence of the conversation sequence in the sample conversation information, such as the reward value.
  • the sample session information is obtained in the following manner: reading a session log obtained by performing user session interaction based on a session rule from a data warehouse; the session rule is cold-start deployed for the resource management service;
  • a candidate session sequence is constructed according to the session log, and a target session sequence is screened out from the candidate session sequence as the sample session information according to the sub-service category corresponding to the session log.
  • the data warehouse is set up for resource management services and is used to store session information.
  • the data warehouse is ODPS (Open Data Processing Service).
  • the session rules refer to the rules for responding to the session text input by the user; the user session interaction refers to the session interaction with the user; the model training process of the model to be trained as shown in Figure 2 collects session logs in an online manner, specifically collects or obtains session logs from the resource management service through a log collector, and the session log records session information (such as stay time, click-through rate, session text, response text).
  • the log collector sends the obtained session log to the data warehouse for storage.
  • the session sequence builder obtains the session log from the data warehouse, constructs candidate session sequences and filters out target session sequences as sample session information.
  • the sample session information is input into the model to be trained to determine the reward value and session preference, and the reward value and sample session preference are obtained.
  • the sample session information is input into the indicator function to calculate the reward value and obtain the target reward value.
  • the training loss is calculated according to the reward value and the target reward value as well as the sample session preference and the previous session preference, and the parameters of the model to be trained are updated according to the training loss.
  • the trained service action model is deployed in the resource management service.
  • the process of the log collector collecting session logs and sending the collected or obtained session logs to the data warehouse for storage is continuous, while the process of the session sequence builder obtaining session logs from the data warehouse, building candidate session sequences and screening out target session sequences as sample session information is only executed during the model training process of the model to be trained.
  • the candidate session sequences are sequence 1, sequence 2, sequence 3 and sequence 4, and the sub-service category corresponding to the session information is new energy financial products. Then, target session sequences related to new energy financial products are screened out from the candidate session sequences: sequence 1 and sequence 3.
  • an optional implementation manner provided by this embodiment performs the following operations: determining a user response parameter of a conversation factor in the conversation log, and using the conversation log and the user response parameter as the candidate conversation sequence; optionally, the conversation factor includes at least one of the following: conversation duration, conversation click rate, and number of conversations.
  • the session sequence refers to a sequence in which session information is presented in a sequence.
  • the session sequence is: Conversation text q1, response text a1, dwell time, click rate, user response parameter 1 for dwell time, user response parameter 0 for click rate.
  • the user response parameter refers to the response parameter of the user in response to the conversation factor. For example, if the conversation factor is dwell time, dwell time ⁇ T, the user response parameter is 1.
  • the user response parameter is determined in the following manner: determine whether the session factor in the session information satisfies a preset condition; if so, determine the user response parameter of the session factor as a first response parameter; if not, determine the user response parameter of the session factor as a second response parameter; wherein the preset condition includes that a duration parameter of the session factor exceeds a parameter threshold, for example, if the session factor is the stay duration, determine whether the duration parameter of the stay duration exceeds the parameter threshold; if so, determine the user response parameter of the stay duration as 1; if not, determine the user response parameter of the stay duration as 0.
  • a preset condition includes that a duration parameter of the session factor exceeds a parameter threshold, for example, if the session factor is the stay duration, determine whether the duration parameter of the stay duration exceeds the parameter threshold; if so, determine the user response parameter of the stay duration as 1; if not, determine the user response parameter of the stay duration as 0.
  • the following operations are performed: the preference loss is calculated according to the sample session preference and the previous session preference, and the session indicator loss is calculated according to the session indicator and the target session indicator; the parameter update is performed according to the preference loss and the session indicator loss; optionally, the target session indicator is obtained by calculating the session indicator of the sample session information based on an indicator function; the sample session preference and the previous session preference correspond to the same session unit in the sample session information.
  • the previous conversation preference refers to the conversation preference corresponding to the previous conversation text of the conversation text corresponding to the sample conversation preference in the same conversation unit, and the conversation unit refers to a conversation, which may include one or more rounds of conversations;
  • the conversation indicator refers to an indicator that characterizes the confidence of the conversation sequence in the sample conversation information or the conversation quality, such as a reward value.
  • the preference difference between the sample session preference and the previous session preference can be calculated as the preference loss
  • the indicator difference between the session indicator and the target session indicator can be calculated as the session indicator loss
  • the sum of the preference loss and the session indicator loss can be calculated as the training loss
  • the parameters of the training model can be updated according to the training loss
  • the process of updating the parameters of the training model according to the preference loss and the session indicator loss can also be achieved by calculating the training loss according to the preference loss, the session indicator loss and their respective assigned weights, and updating the parameters of the training model based on the training loss.
  • the above training process is repeated to train the model until the loss function converges. After the loss function converges, the training is completed to obtain the service action model.
  • step S106 can be replaced by determining the service action sequence of the conversation text according to any one or more of the service intent, service key features, conversation behavior features and/or the conversation text, and forming a new implementation method with other processing steps provided in this embodiment, or can also be replaced by determining the service action sequence of the conversation text according to the service intent, service key features, conversation behavior features and/or the conversation text.
  • the service key features, service intent and historical session information are used to determine the service action sequence of the session text, and together with other processing steps provided in this embodiment, a new implementation method is formed.
  • Step S108 executing the service processing corresponding to the service action sequence to respond to the conversation text.
  • the above determines the service action sequence of the conversation text based on the service intent, service key characteristics and conversation behavior characteristics.
  • the service processing corresponding to the service action sequence is executed to respond to the conversation text.
  • the service processing is performed according to the service action sequence to improve the effectiveness of the service processing, thereby improving the user experience.
  • the following operations are performed: generating a response text for the session text according to the service action sequence, rendering a service entry for a sub-service of the resource management service based on the service action sequence, and jumping from the resource management service to the target service based on the service action sequence.
  • the response text refers to the text corresponding to the conversation text generated in response to the conversation text.
  • the response text can be generated by inputting the service action sequence, service intention, service key features and/or conversation behavior features into the text generation model to obtain the response text of the conversation text, or it can also be achieved by generating the response text of the conversation text according to the service key items, service key information and/or service action sequence and text template; wherein, in the process of generating the response text of the conversation text according to the service key items, service key information and/or service action sequence and text template, the text template corresponding to the service action sequence can be obtained, and the service key information is filled into the filling position of the service key item corresponding to the text template to obtain the response text of the conversation text.
  • the conversation text is "What should I do if the new energy financial product I bought falls"
  • the response text is "Recently, new energy financial products have indeed fallen a little. Don't worry. Although new energy financial products have fallen recently, the overall valuation is still low. Please wait patiently.”
  • the sub-service refers to a sub-application that is mounted on the resource management service for operation, for example, the resource management service is a financial service, and the sub-service of the resource management service is a financial product service; for example, the session text is "Open a (representing the sub-service of the resource management service)", and the service action sequence is "Request confirmation->Render entry”, then the response text of the session text can be generated according to the service action of "Request confirmation", and the service entry of a can be rendered based on the service action of "Render entry”.
  • the target service includes a sub-service of the resource management service or a third-party application service; wherein the third-party application service refers to an application program that runs independently on the terminal device; for example, the session text is "Open b (representing the target service)", and the service action sequence is "Jump”, and based on the "Jump” service action sequence, it jumps from the resource management service to the target service b.
  • the third-party application service refers to an application program that runs independently on the terminal device; for example, the session text is "Open b (representing the target service)", and the service action sequence is "Jump”, and based on the "Jump” service action sequence, it jumps from the resource management service to the target service b.
  • the above-mentioned service processing process corresponding to the execution of the service action sequence can be replaced by generating a response text for the session text according to the service action sequence, rendering the service entrance of the sub-service of the resource management service based on the service action sequence, or jumping from the resource management service to the target service based on the service action sequence, and forming a new implementation method with other processing steps provided in this embodiment.
  • the processing procedure of the service processing obtains the conversation text and historical conversation information of the user u in the financial management service, inputs the conversation text and historical conversation information into the NLU module, performs intent recognition and feature analysis in the NLU module, obtains the service intent and service key features, and inputs the user ID of the user u in the financial management service into the DST module.
  • the DST module reads the user portrait information based on the user ID, thereby determining the user's conversation preference according to the user portrait information, inputs the service intent, service key features and conversation preference into the service action model to determine the service action, and obtains the service action sequence.
  • service intent may also be input into the service action model to determine the service action, and obtain the service action sequence, and the service action sequence is input into the NLG (Natural Language Generation) module, and the NLG module generates a response text for the conversation text according to the service action sequence.
  • NLG Natural Language Generation
  • the service processing method firstly performs intent recognition on the conversation text input by the user in the resource management service to obtain the service intent, and performs feature analysis on the historical conversation information and conversation text of the user in the resource management service to obtain the key service features.
  • the service action is determined for the service intent, the key service features, the user's conversation preferences and the historical conversation information through the service action model, and a response text of the conversation text is generated according to the service action sequence obtained by performing the service action determination, and the service entrance of the sub-service of the resource management service is rendered based on the service action sequence obtained by performing the service action determination, and/or, based on the service action sequence obtained by performing the service action determination, the service is jumped from the resource management service to the target service.
  • the following further illustrates the service processing method provided by this embodiment by taking the application of a service processing method provided by this embodiment in a financial service scenario as an example.
  • the service processing method applied to the financial service scenario specifically includes the following steps.
  • Step S402 obtaining the user's historical session information and input session text in the financial management service.
  • Step S404 perform intent recognition on the conversation text to obtain service intent.
  • Step S406 searching for a service key item corresponding to the service intention in the key item set of the financial management service.
  • Step S408 determining service key information according to the historical session information, session text and service key items.
  • Step S410 input the service intention, service key items, service key information, user's session preference and historical session information into the service action model to determine the service action and obtain a service action sequence.
  • Step S412 Generate a response text of the conversation text according to the service action sequence.
  • step S412 can be replaced by generating a response text of the conversation text according to the service action sequence, rendering the service entrance of the sub-service of the financial management service based on the service action sequence, and/or jumping from the financial management service to the target service based on the service action sequence, and forming a new implementation method with other processing steps provided in this embodiment.
  • An embodiment of a service processing device provided in this specification is as follows: In the above embodiment, a service processing method is provided, and correspondingly, a service processing device is also provided, which is described below in conjunction with the accompanying drawings.
  • FIG. 5 there is shown a schematic diagram of a service processing device provided by this embodiment.
  • the description is relatively simple, and the relevant parts can refer to the corresponding description of the method embodiment provided above.
  • the device embodiment described below is only illustrative.
  • This embodiment provides a service processing device, including: a session information acquisition module 502, configured to acquire the user's historical session information and input session text in the resource management service; a feature analysis module 504, configured to perform intent recognition on the session text to obtain the service intent, and perform feature analysis on the historical session information and the session text to obtain service key features; a service action determination module 506, configured to determine the service action sequence of the session text according to the service intent, the service key features and the user's session behavior features; a service processing module 508, configured to execute service processing corresponding to the service action sequence to respond to the session text.
  • a session information acquisition module 502 configured to acquire the user's historical session information and input session text in the resource management service
  • a feature analysis module 504 configured to perform intent recognition on the session text to obtain the service intent, and perform feature analysis on the historical session information and the session text to obtain service key features
  • a service action determination module 506 configured to determine the service action sequence of the session text according to the service intent, the service key features and the user's session behavior features
  • FIG. 6 is a structural diagram of a service processing device provided by one or more embodiments of this specification.
  • a service processing device provided in this embodiment includes: As shown in FIG. 6, the service processing device may have relatively large differences due to different configurations or performances, and may include one or more processors 601 and a memory 602, and the memory 602 may store one or more storage applications or data.
  • the memory 602 may be a temporary storage or a permanent storage.
  • the application stored in the memory 602 may include one or more modules (FIG.
  • the service processing device 601 may include a series of computer executable instructions in the service processing device.
  • the processor 601 may be configured to communicate with the memory 602, and execute a series of computer executable instructions in the memory 602 on the service processing device.
  • the service processing device may also include one or more power supplies 603, one or more wired or wireless network interfaces 604, one or more input/output interfaces 605, one or more keyboards 606, etc.
  • a service processing device includes a memory and one or more programs, wherein the one or more programs are stored in the memory, and the one or more programs may include one or more modules, and each module may include a series of computer executable instructions in the service processing device, and is configured to be executed by one or more processors.
  • the one or more programs include computer executable instructions for performing the following: obtaining historical session information and input session text of a user in a resource management service; performing intent recognition on the session text to obtain service intent, and performing feature analysis on the historical session information and the session text to obtain service key features; determining a service action sequence of the session text based on the service intent, the service key features and the user's session behavior features; and executing service processing corresponding to the service action sequence in response to the session text.
  • An embodiment of a storage medium provided in this specification is as follows: Corresponding to the service processing method described above, based on the same technical concept, one or more embodiments of this specification also provide a storage medium.
  • the storage medium provided in this embodiment is used to store computer-executable instructions, and the computer-executable instructions implement the following process when executed by a processor: obtaining the user's historical session information and input session text in the resource management service; performing intent recognition on the session text to obtain the service intent, and performing feature analysis on the historical session information and the session text to obtain service key features; determining the service action sequence of the session text based on the service intent, the service key features and the user's session behavior features; and executing service processing corresponding to the service action sequence to respond to the session text.
  • a programmable logic device such as a field programmable gate array (FPGA)
  • FPGA field programmable gate array
  • HDL Hardware Description Language
  • HDL Very-High-Speed Integrated Circuit Hardware Description Language
  • ABEL Advanced Boolean Expression Language
  • AHDL Altera Hardware Description Language
  • HDCal Joint CHDL
  • JHDL Java Hardware Description Language
  • Lava Lava
  • Lola MyHDL
  • PALASM RHDL
  • VHDL Very-High-Speed Integrated Circuit Hardware Description Language
  • Verilog Verilog
  • the controller may be implemented in any suitable manner, for example, the controller may take the form of a microprocessor or processor and a computer readable medium storing a computer readable program code (e.g., software or firmware) executable by the (micro)processor, a logic gate, a switch, an application specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller, examples of which include but are not limited to the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320, and the memory controller may also be implemented as part of the control logic of the memory.
  • a computer readable program code e.g., software or firmware
  • the controller may be implemented in the form of a logic gate, a switch, an application specific integrated circuit, a programmable logic controller, and an embedded microcontroller by logically programming the method steps. Therefore, such a controller may be considered as a hardware component, and the means for implementing various functions included therein may also be considered as a structure within the hardware component. Or even, the means for implementing various functions may be considered as both a software module for implementing the method and a structure within the hardware component.
  • the systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities. Or it is implemented by a product with a certain function.
  • a typical implementation device is a computer.
  • the computer can be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.
  • one or more embodiments of this specification may be provided as a method, system or computer program product. Therefore, one or more embodiments of this specification may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
  • computer-usable storage media including but not limited to disk storage, CD-ROM, optical storage, etc.
  • These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device that implements the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.
  • These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.
  • a computing device includes one or more processors (CPU), input/output interfaces, network interfaces, and memory.
  • processors CPU
  • input/output interfaces network interfaces
  • memory volatile and non-volatile memory
  • Memory may include non-permanent storage in computer-readable media, random access memory (RAM), and/or Or non-volatile memory such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.
  • Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information.
  • Information can be computer readable instructions, data structures, program modules or other data.
  • Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.
  • computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.
  • One or more embodiments of the present specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules.
  • program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types.
  • One or more embodiments of the present specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network.
  • program modules may be located in local and remote computer storage media, including storage devices.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Business, Economics & Management (AREA)
  • Finance (AREA)
  • Development Economics (AREA)
  • Accounting & Taxation (AREA)
  • General Health & Medical Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Mathematical Physics (AREA)
  • Game Theory and Decision Science (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Operations Research (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Strategic Management (AREA)
  • Technology Law (AREA)
  • General Business, Economics & Management (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本说明书实施例提供了服务处理方法及装置，其中，一种服务处理方法包括：获取用户在资源管理服务的历史会话信息和输入的会话文本；对所述会话文本进行意图识别获得服务意图，并对所述历史会话信息和所述会话文本进行特征解析，获得服务关键特征；根据所述服务意图、所述服务关键特征和所述用户的会话行为特征，确定所述会话文本的服务动作序列；执行所述服务动作序列对应的服务处理，以响应所述会话文本。

Description

服务处理方法及装置 技术领域
本文件涉及数据处理技术领域，尤其涉及一种服务处理方法及装置。
背景技术
随着互联网技术的不断发展，互联网平台推出越来越多的互联网服务，利用互联网技术来实现对应的线上服务也愈发普遍，而在用户使用线上服务的过程中，往往存在一定的服务问题，针对于此，可在线上服务中设置对话服务，用户针对存在的服务问题通过对话服务进行提出，从而得到相应的反馈，这也使得对话服务逐渐成为研究热点。
发明内容
本说明书一个或多个实施例提供了一种服务处理方法，包括：获取用户在资源管理服务的历史会话信息和输入的会话文本。对所述会话文本进行意图识别获得服务意图，并对所述历史会话信息和所述会话文本进行特征解析，获得服务关键特征。根据所述服务意图、所述服务关键特征和所述用户的会话行为特征，确定所述会话文本的服务动作序列。执行所述服务动作序列对应的服务处理，以响应所述会话文本。
本说明书一个或多个实施例提供了一种服务处理装置，包括：会话信息获取模块，被配置为获取用户在资源管理服务的历史会话信息和输入的会话文本。特征解析模块，被配置为对所述会话文本进行意图识别获得服务意图，并对所述历史会话信息和所述会话文本进行特征解析，获得服务关键特征。服务动作确定模块，被配置为根据所述服务意图、所述服务关键特征和所述用户的会话行为特征，确定所述会话文本的服务动作序列。服务处理模块，被配置为执行所述服务动作序列对应的服务处理，以响应所述会话文本。
本说明书一个或多个实施例提供了一种服务处理设备，包括：处理器；以及，被配置为存储计算机可执行指令的存储器，所述计算机可执行指令在被执行时使所述处理器：获取用户在资源管理服务的历史会话信息和输入的会话文本。对所述会话文本进行意图识别获得服务意图，并对所述历史会话信息和所述会话文本进行特征解析，获得服务关键特征。根据所述服务意图、所述服务关键特征和所述用户的会话行为特征，确定所述会话文本的服务动作序列。执行所述服务动作序列对应的服务处理，以响应所述会话文 本。
本说明书一个或多个实施例提供了一种存储介质，用于存储计算机可执行指令，所述计算机可执行指令在被处理器执行时实现以下流程：获取用户在资源管理服务的历史会话信息和输入的会话文本。对所述会话文本进行意图识别获得服务意图，并对所述历史会话信息和所述会话文本进行特征解析，获得服务关键特征。根据所述服务意图、所述服务关键特征和所述用户的会话行为特征，确定所述会话文本的服务动作序列。执行所述服务动作序列对应的服务处理，以响应所述会话文本。
附图说明
为了更清楚地说明本说明书一个或多个实施例或相关技术中的技术方案，下面将对实施例或相关技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本说明书中记载的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图；
图1为本说明书一个或多个实施例提供的一种服务处理方法处理流程图；
图2为本说明书一个或多个实施例提供的一种服务动作模型的模型训练过程示意图；
图3为本说明书一个或多个实施例提供的一种服务处理的处理过程示意图；
图4为本说明书一个或多个实施例提供的一种应用于理财服务场景的服务处理方法处理流程图；
图5为本说明书一个或多个实施例提供的一种服务处理装置示意图；
图6为本说明书一个或多个实施例提供的一种服务处理设备的结构示意图。
具体实施方式
为了使本技术领域的人员更好地理解本说明书一个或多个实施例中的技术方案，下面将结合本说明书一个或多个实施例中的附图，对本说明书一个或多个实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本说明书的一部分实施例，而不是全部的实施例。基于本说明书一个或多个实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都应当属于本文件的保护范围。
本说明书提供的一种服务处理方法实施例：本实施例提供的服务处理方法，根据会 话行为特征以及对用户在资源管理服务的历史会话信息和会话文本进行解析获得的服务意图、服务关键特征，确定服务动作序列，并执行服务动作序列对应的服务处理。具体的，对用户在资源管理服务的历史会话信息和输入的会话文本进行解析获得服务意图和服务关键特征，以服务意图、服务关键特征和会话行为特征为依据，确定会话文本的服务动作序列，并通过执行服务动作序列对应的服务处理，来响应会话文本，以此，通过在确定服务动作序列的过程中引入会话行为特征，来确保针对用户的会话行为特征决策出服务动作序列，提升确定的服务动作序列的灵活性和多样性，能够针对不同的用户决策出不同的服务动作序列，满足用户的多样化需求，并且，通过从意图、关键特征、会话行为特征等多个方面确定服务动作序列，提升服务动作序列的有效性和精确度。
参照图1，本实施例提供的服务处理方法，具体包括步骤S102至步骤S108。
步骤S102，获取用户在资源管理服务的历史会话信息和输入的会话文本。
本实施例所述资源管理服务，是指针对用户存入的资源进行管理以使用户获得收益的服务，具体资源管理服务可以是理财服务，比如可以是线上和线下结合的金融机构的理财服务、线上金融机构的理财服务、或者第三方支付平台中的理财服务，资源管理服务中可以包括一个或者多个子服务，比如子服务可以是余额查询服务、理财产品服务、保障服务(保险服务)等，此外，资源管理服务还可以是与资源管理相关的其他服务。
所述会话文本是指用户在整个会话过程中输入的当前一轮的对话文本，该会话过程可针对一个会话单元，会话单元包括从开始到结束的一通会话，在一个会话单元中，可以进行多轮会话交互；会话文本可为多轮会话交互中当前一轮的会话文本，比如用户输入的当前轮的会话文本为“我买的新能源理财产品跌了怎么办”；所述历史会话信息，是指用户在整个会话过程中会话文本的前一轮或者前几轮的会话信息，可选的，历史会话信息，包括下述至少一项：历史会话文本、历史服务动作序列、历史响应文本、历史服务意图、历史服务关键特征；此外，历史会话信息中也可包括其他类型的会话信息。
实际应用中，用户在访问资源管理服务的过程中，可能针对资源管理服务中的子服务存在疑虑困惑，比如用户想要了解保险信息或者用户想要了解自己在资源管理服务的资源余额，在此情况下，用户可将自己存在的疑虑困惑通过会话文本的形式在资源管理服务进行输入，输入会话文本的渠道可资源管理服务的请求渠道，提升用户体验。
为了提升会话交互的便捷性，针对资源管理服务可设置会话交互界面，在会话交互界面配置会话文本输入控件，用户通过触发该会话文本输入控件可输入会话文本，此外， 会话交互界面也可配置会话语音输入控件，用户通过触发该会话语音输入控件输入会话语音，对会话语音进行语音识别，获得会话文本，通过文本和语音多种输入方式进行会话交互，适应多种会话交互场景。
具体实施时，获取用户在资源管理服务的历史会话信息和输入的会话文本，以此，不仅从当前输入的会话文本进行处理，还结合用户在资源管理服务的历史会话信息进行处理，提升会话处理的全面性和精确度，提升用户体验。
实际应用中，用户可能在资源管理服务针对某个事项输入若干次相同的会话文本，比如用户输入会话文本“我的新能源理财产品跌了怎么办”，在用户接收到针对该会话文本的回复内容后，再次输入会话文本“我的新能源理财产品跌了怎么办”，针对于此，为了提升对会话文本进行回复的有效性，可根据会话文本和历史会话信息中的历史会话文本的文本状态，向用户针对该会话文本进行响应处理，具体在文本状态为会话文本与历史会话文本不相同的情况下，执行下述步骤S104。
本实施例提供的一种可选实施方式中，在获取用户在资源管理服务的历史会话信息和输入的会话文本执行之后，还执行如下操作：判断所述会话文本和所述历史会话信息中的历史会话文本是否相同；若否，执行下述步骤S104；若是，根据所述会话文本的第一数量和所述历史会话文本的第二数量计算全局文本数量，基于所述全局文本数量对所述历史会话信息中的历史服务动作序列进行更新处理，并按照更新后的服务动作序列生成所述会话文本的响应文本。
具体的，在会话文本和历史会话信息中的历史会话文本相同的情况下，计算会话文本的第一数量和历史会话文本的第二数量之和作为全局文本数量，根据全局文本数量确定用户的情绪类别，并基于该情绪类别对历史对话信息中与会话文本相邻的历史会话文本的历史服务动作序列进行更新处理，按照更新后的服务动作序列生成会话文本的响应文本。
其中，情绪类别包括第一类别、第二类别和/或第三类别，可选的，全局文本数量与情绪类别具有对应关系，比如全局文本数量≤m时，用户的情绪类别为第一类别；m＜全局文本数量≤n时，用户的情绪类别为第二类别；全局文本数量≥o时，用户的情绪类别为第三类别。
例如，存在1条历史会话文本为“我的新能源理财产品跌了怎么办”，用户当前输入的会话文本也是“我的新能源理财产品跌了怎么办”，当前输入的会话文本和历史会 话文本相同，并且全局文本数量为2，在1＜全局文本数量≤3时，用户的情绪类别为第二类别，历史服务动作序列为“行情解读->收益归因->建议”，在情绪类别为第二类别的情况下，对历史服务动作序列进行更新处理，获得更新后的服务动作序列为“情绪安抚->行情解读->收益归因->建议”，并按照更新后的服务动作序列生成会话文本的响应文本。
此外，上述在会话文本和历史对话信息中的历史会话文本相同的情况下执行的操作，可被替换为根据所述会话文本的第一数量和所述历史会话文本的第二数量计算全局文本数量；根据所述全局文本数量确定所述用户的情绪类别，并基于情绪类别对所述历史会话信息中的历史服务动作序列进行更新处理；按照更新后的服务动作序列生成所述会话文本的响应文本，并与本实施例提供的其他处理步骤组成新的实现方式；或者，也可被替换为根据会话文本的第一数量和历史会话文本的第二数量计算全局文本数量，并基于全局文本数量对历史会话信息中的历史服务动作序列进行更新处理，并执行更新后的服务动作序列对应的服务处理，以响应所述会话文本，并与本实施例提供的其他处理步骤组成新的实现方式。
需要补充的，步骤S102可被替换为获取用户在资源管理服务的用户标识、历史会话信息和输入的会话文本，并与本实施例提供的其他处理步骤组成新的实现方式；其中，用户标识包括用户在资源管理服务的用户账号。
步骤S104，对所述会话文本进行意图识别获得服务意图，并对所述历史会话信息和所述会话文本进行特征解析，获得服务关键特征。
上述获取用户在资源管理服务的历史会话信息和输入的会话文本，本步骤中，为了针对性地对用户输入的会话文本进行处理，对会话文本进行意图识别获得服务意图，并对历史会话信息和会话文本进行特征解析获得服务关键特征。
本实施例所述服务意图，是指用户在资源管理服务的需求或者目的，该服务意图是与资源管理服务相关的，是从“粗粒度”层面获取用户的需求，比如会话文本为“我的新能源理财产品跌了怎么办”，服务意图即为新能源理财产品建议；所述服务关键特征，是指基于会话文本的服务意图或者从会话文本中获得的与资源管理服务相关的关键特征信息，可选的，所述服务关键特征，包括服务关键项和/或服务关键信息，沿用上例，会话文本为“我的新能源理财产品跌了怎么办”，服务意图为新能源理财产品建议，服务关键项为理财产品类型，服务关键信息为新能源。
具体实施时，在对会话文本进行意图识别获得服务意图的过程中，为了提升意图识别的识别效率，可引入意图识别模型，将会话文本输入意图识别模型进行意图识别获得服务意图，此外，也可将会话文本和历史会话信息输入意图识别模型进行意图识别获得服务意图，即也可基于历史会话信息对所述会话文本进行意图识别获得服务意图。
另外，针对资源管理服务可设置预设服务意图集，则对会话文本进行意图识别获得服务意图的过程，还可通过如下方式实现：从会话文本中提取与资源管理服务关联的服务关键词，并将该服务关键词与预设服务意图集中的预设服务意图进行匹配处理，获得服务关键词匹配的预设服务意图，将该预设服务意图作为所述服务意图。
沿用上例，会话文本为“我的新能源理财产品跌了怎么办”，从会话文本中提取与资源管理服务关联的服务关键词为“新能源理财产品”，服务关键词匹配的预设服务意图为“新能源理财产品建议”，该预设服务意图即为所述服务意图。
在具体的执行过程中，为了更深入地了解用户输入的会话文本的真实意图，挖掘用户内心的真实需求，从而能够为用户提供更有效的服务处理，本实施例提供的一种可选实施方式中，在对历史会话信息和会话文本进行特征解析，获得服务关键特征的过程中，执行如下操作：在所述资源管理服务的关键项集中查找所述服务意图对应的服务关键项，和/或，对基于服务意图在会话文本中提取的服务关键词进行转换获得的服务关键项；
根据所述历史会话信息、所述会话文本和所述服务关键项，确定服务关键信息。
其中，所述关键项集，是指针对资源管理服务设置的每个服务意图对应的服务关键项组成的集合；所述服务关键项，是指在“细粒度”层面进一步挖掘的关键项，所述服务关键项包括服务填充项，比如理财产品类别、理财产品买入时间；所述服务关键信息，是指服务关键项对应的关键信息，所述服务关键信息包括服务填充信息，比如服务关键项为理财产品类别，服务关键信息为新能源。
沿用上例，服务意图为新能源理财产品建议，关键项集中包括白酒理财产品建议对应的服务关键项和新能源理财产品建议对应的服务关键项组成的集合，在关键项集中查找新能源理财产品建议对应的服务关键项“理财产品类型、理财产品买入时间、理财产品买入金额”，或者，先对基于服务意图在会话文本中提取的服务关键词“新能源理财产品”进行转换获得服务关键项“理财产品类型”，再在关键项集中查找新能源理财产品建议对应的服务关键项“理财产品买入时间、理财产品买入金额”，根据历史会话信息、会话文本和服务关键项“理财产品类型、理财产品买入时间、理财产品买入金额”， 确定服务关键信息。
进一步，在上述根据历史会话信息、会话文本和服务关键项，确定服务关键信息的过程中，本实施例提供的一种可选实施方式中，通过如下方式确定服务关键信息：基于所述服务关键项在所述历史会话信息和所述会话文本中提取服务关键信息；若未提取到服务关键信息，基于所述服务关键项在数据库中查询对应的服务关键信息；若提取到服务关键信息，将提取的服务关键信息作为确定的服务关键信息。
沿用上例，服务关键项为“理财产品类型、理财产品买入时间、理财产品买入金额”，基于服务关键项“理财产品类型”在历史会话信息和会话文本中提取服务关键信息“新能源”，而服务关键项“理财产品买入时间、理财产品买入金额”在历史会话信息和会话文本中未提取到，则在数据库中查询服务关键项“理财产品买入时间、理财产品买入金额”对应的服务关键信息为“x月x日、xx万元”。
需要说明的是，对会话文本进行意图识别获得服务意图、以及对历史会话信息和会话文本进行特征解析获得服务关键特征的过程可由NLU(Natural Language Understanding，自然语言理解)模块实现，具体NLU模块中可以包括意图识别模型和/或特征解析模型，意图识别模型可对会话文本进行意图识别获得服务意图，特征解析模型可对历史会话信息和会话文本进行特征解析获得服务关键特征。
实际应用中，在基于会话文本和历史会话信息获得服务意图和服务关键特征之后，可能历史会话信息中的历史会话状态和当前的服务意图、服务关键项不匹配，即历史会话状态中的历史服务意图、历史服务关键特征和当前的服务意图、服务关键特征不一致，为了提升服务意图和服务关键特征的精确度，本实施例提供的一种可选实施方式中，在对会话文本进行意图识别获得服务意图，并对历史会话信息和会话文本进行特征解析，获得服务关键特征执行之后，还执行如下操作：判断所述服务意图和所述服务关键特征与所述历史会话信息中的历史会话状态是否匹配；若否，根据所述服务意图和所述服务关键特征，对所述历史会话信息中的历史会话状态进行修正处理，在此基础上，可执行下述步骤S106；若是，执行下述步骤S106。
可选的，所述历史会话状态包括历史服务意图和/或历史服务关键特征，此外，所述历史会话状态还可包括历史会话文本和/或历史服务动作序列。
具体的，判断服务意图和服务关键特征与历史会话信息中的历史会话状态是否匹配的过程，可通过判断服务意图和/或服务关键特征与历史对话信息中的历史会话状态中的 历史服务意图和/或历史服务关键特征是否一致的方式实现；根据服务意图和服务关键特征，对历史会话信息中的历史会话状态进行修正处理的过程，可通过将历史会话状态中的历史服务意图和/或历史服务关键特征修正为所述服务意图和/或所述服务关键特征的方式实现。
需要说明的是，步骤S104可被替换为基于历史会话信息对会话文本进行意图识别获得服务意图，并对历史会话信息和/或会话文本进行特征解析，获得服务关键特征，并与本实施例提供的其他处理步骤组成新的实现方式；或者，也可被替换为对历史会话信息和/或会话文本进行解析，获得服务意图和/或服务关键特征，并与本实施例提供的其他处理步骤组成新的实现方式。
步骤S106，根据所述服务意图、所述服务关键特征和所述用户的会话行为特征，确定所述会话文本的服务动作序列。
上述对会话文本和历史会话信息进行解析获得服务意图和服务关键特征，具体对会话文本进行意图识别获得服务意图，并对历史会话信息和会话文本进行特征解析，获得服务关键特征，本步骤中，以服务意图、服务关键特征和用户的会话行为特征为依据，确定会话文本的服务动作序列。
本实施例所述会话行为特征，是指用户在资源管理服务进行会话的行为特征信息，可选的，所述会话行为特征包括会话偏好和/或历史会话信息；所述会话偏好，是指用户在资源管理服务进行资源管理的资源管理偏好或者服务偏好或者用户在资源管理服务进行会话的会话偏好，可选的，所述会话偏好包括性格特征和/或情绪类别，比如会话偏好分为激进型、稳健型，或者会话偏好分为乐观型、悲观型；本实施例提供的一种可选实施方式中，所述会话行为特征，通过如下方式确定：从用户属性信息和资源管理记录中提取关键词；根据所述关键词和预设会话偏好的偏好特征计算匹配度，并基于所述匹配度确定所述用户的会话偏好，将所述会话偏好和所述历史会话信息作为所述会话行为特征。
具体的，基于匹配度确定所述用户的会话偏好的过程，可通过若所述匹配度大于匹配度阈值，将所述预设会话偏好作为所述用户的会话偏好的方式实现。
所述服务动作序列，是指针对会话文本在资源管理服务进行响应的服务动作组成的序列，该服务动作序列中包含的服务动作可以是一者或者多者，比如，会话文本为“我的新能源理财产品跌了怎么办”，服务动作序列为“事实承接->情绪安抚->行情解读-> 建议”，再比如，会话文本为“打开a(代表资源管理服务的子服务)”，服务动作序列为“请求确认->渲染入口”、“请求确认->跳转”或者“跳转”。
具体实施时，为了提升服务动作序列的确定效率和确定精确度，可引入服务动作模型，用于确定会话文本的服务动作序列，本实施例提供的一种可选实施方式中，在根据服务意图、服务关键特征和所述用户的会话行为特征，确定会话文本的服务动作序列的过程中，执行如下操作：将所述服务意图、所述服务关键特征、用户的会话偏好和所述历史会话信息输入服务动作模型进行服务动作确定，获得所述服务动作序列；可选的，所述会话偏好基于用户属性信息和所述用户在所述资源管理服务的资源管理记录确定，或者，所述会话偏好基于用户画像信息确定，所述用户画像信息包括用户属性信息和/或用户在资源管理服务的资源管理记录，可选的，所述用户画像信息基于用户标识读取获得，具体的，可由DST(Dialogue State Tracking，对话状态追踪)模块基于用户标识从外部数据库读取获得，用户标识可以是用户在资源管理服务的用户账号。
其中，所述用户属性信息，是指与用户属性相关的信息，所述用户属性信息包括但不限于：从出生时起到计算时止生存的时间长度、职业、工作时长；所述资源管理记录，是指在资源管理服务进行资源管理的行为信息记录，比如在理财服务的理财记录。
此外，上述根据服务意图、服务关键特征和会话行为特征，确定会话文本的服务动作序列的具体执行过程，可被替换为将用户画像信息输入服务动作模型的第一网络进行会话偏好确定，获得会话偏好，并将服务意图、服务关键特征、所述会话偏好和历史会话信息输入服务动作模型的第二网络进行服务动作确定，获得会话文本的服务动作序列，并与本实施例提供的其他处理步骤组成新的实现方式。需要补充的是，服务动作模型的模型训练可采用分层强化学习技术完成，服务动作模型可采用BCQ(Batch Constrained deep Q-learning，离线强化学习算法)。
在具体的执行过程中，可提前进行模型训练获得服务动作模型，由于虚拟环境构造或者线上训练成本较大，所以为了减少训练成本，可通过离线训练的方式对待训练模型进行模型训练获得服务动作模型，本实施例提供的一种可选实施方式中，所述服务动作模型，采用如下方式训练：将样本会话信息输入待训练模型进行会话指标计算，获得会话指标；根据所述会话指标、样本会话偏好和所述样本会话信息对所述待训练模型进行参数更新；可选的，所述样本会话偏好在将所述样本会话信息输入偏好检测模型进行偏好检测后获得。
其中，所述样本会话信息中可包含有一个或者多个会话序列，所述会话指标是指表 征样本会话信息中会话序列的置信度的指标，比如奖赏值。
可选的，所述样本会话信息，采用如下方式获得：从数据仓库读取基于会话规则进行用户会话交互获得的会话日志；所述会话规则针对所述资源管理服务进行冷启动部署；
根据所述会话日志构建候选会话序列，并根据所述会话日志对应的子服务类别从所述候选会话序列中筛选出目标会话序列作为所述样本会话信息。
其中，所述数据仓库是面向资源管理服务设置的，用于存储会话信息，比如数据仓库为ODPS(Open Data Processing Service，开发数据处理服务)，会话规则是指针对用户输入的会话文本进行响应的规则；所述用户会话交互是指与用户进行会话交互；如图2所示的待训练模型的模型训练过程，通过在线方式收集会话日志，具体通过日志收集器从资源管理服务收集或者获取会话日志，会话日志中记录有会话信息(比如停留时长、点击率、会话文本、响应文本)，日志收集器将获取的会话日志发送至数据仓库进行存储，会话序列构建器在模型训练的过程中，从数据仓库获取会话日志，构建候选会话序列并筛选出目标会话序列作为样本会话信息。一方面将样本会话信息输入待训练模型进行奖赏值和会话偏好确定，获得奖赏值和样本会话偏好，另一方面将样本会话信息输入指标函数进行奖赏值计算，获得目标奖赏值，根据奖赏值和目标奖赏值以及样本会话偏好和前一会话偏好计算训练损失，并根据训练损失进行待训练模型的参数更新，将训练完成的服务动作模型部署于资源管理服务。
需要说明的是，日志收集器收集会话日志以及将收集或者获取的会话日志发送至数据仓库进行存储的过程是持续进行的，而会话序列构建器从数据仓库获取会话日志，构建候选会话序列并筛选出目标会话序列作为样本会话信息的过程仅在待训练模型的模型训练过程中执行。
例如，候选会话序列有序列1、序列2、序列3和序列4，会话信息对应的子服务类别为新能源理财产品，则从候选会话序列中筛选出与新能源理财产品相关的目标会话序列：序列1和序列3。
在根据会话日志构建候选会话序列的过程中，本实施例提供的一种可选实施方式中，执行如下操作：确定所述会话日志中会话因子的用户响应参数，并将所述会话日志和所述用户响应参数作为所述候选会话序列；可选的，所述会话因子，包括下述至少一项：会话停留时长、会话点击率、会话次数。
其中，会话序列是指将会话信息以序列的方式进行呈现的序列，比如会话序列为： 会话文本q1、响应文本a1、停留时长、点击率、停留时长的用户响应参数1，点击率的用户响应参数0。所述用户响应参数是指用户针对会话因子进行响应的响应参数，比如会话因子为停留时长，停留时长≥T，用户响应参数为1。
具体的，所述用户响应参数，通过如下方式确定：判断会话信息中会话因子是否满足预设条件，若是，确定会话因子的用户响应参数为第一响应参数，若否，确定会话因子的用户响应参数为第二响应参数；其中，所述预设条件包括会话因子的时长参数超出参数阈值，比如会话因子为停留时长，判断停留时长的时长参数是否超出参数阈值，若是，确定停留时长的用户响应参数为1，若否，确定停留时长的用户响应参数为0。
在上述根据会话指标、样本会话偏好和样本会话信息对待训练模型进行参数更新的过程中，本实施例提供的一种可选实施方式中，执行如下操作：根据所述样本会话偏好和前一会话偏好计算偏好损失，并根据所述会话指标和目标会话指标计算会话指标损失；根据所述偏好损失和所述会话指标损失进行所述参数更新；可选的，所述目标会话指标基于指标函数对所述样本会话信息进行会话指标计算获得；所述样本会话偏好和所述前一会话偏好对应于所述样本会话信息中的同一会话单元。
其中，所述前一会话偏好是指同一会话单元中样本会话偏好对应的会话文本的前一会话文本对应的会话偏好，所述会话单元是指一通会话，该会话单元中可包括一轮或者多轮会话；所述会话指标是指表征样本会话信息中会话序列的置信度或者会话质量的指标，比如奖赏值。
具体的，可计算样本会话偏好和前一会话偏好的偏好差值作为偏好损失，计算会话指标和目标会话指标的指标差值作为会话指标损失，计算偏好损失和会话指标损失之和作为训练损失，并根据训练损失对待训练模型进行参数更新；根据偏好损失和会话指标损失对待训练模型进行参数更新的过程，还可通过根据偏好损失、会话指标损失以及各自的分配权重计算训练损失，并基于训练损失对待训练模型进行参数更新的方式实现。
参见上述基于样本会话信息对待训练模型进行模型训练获得服务动作模型的训练过程，重复上述训练过程进行模型训练，直至损失函数收敛，在损失函数收敛后即完成训练，获得服务动作模型。
需要补充的是，上述步骤S106可被替换为根据服务意图、服务关键特征、会话行为特征和/或所述会话文本中的任意一者或者多者，确定会话文本的服务动作序列，并与本实施例提供的其他处理步骤组成新的实现方式，或者，也可被替换为根据服务意图、服 务关键特征、服务意图和历史会话信息，确定会话文本的服务动作序列，并与本实施例提供的其他处理步骤组成新的实现方式。
步骤S108，执行所述服务动作序列对应的服务处理，以响应所述会话文本。
上述根据服务意图、服务关键特征和会话行为特征，确定会话文本的服务动作序列，本步骤中，执行服务动作序列对应的服务处理，以响应会话文本，以此，按照服务动作序列执行服务处理，提升服务处理的有效性，从而提升用户体验。
本实施例提供的一种可选实施方式中，在执行服务动作序列对应的服务处理的过程中，执行如下操作：按照所述服务动作序列生成所述会话文本的响应文本，基于所述服务动作序列渲染所述资源管理服务的子服务的服务入口，并基于所述服务动作序列从所述资源管理服务跳转至目标服务。
其中，所述响应文本是指为响应会话文本生成的会话文本对应的文本，在按照服务动作序列生成会话文本的响应文本的过程中，可通过将服务动作序列、服务意图、服务关键特征和/或会话行为特征输入文本生成模型进行响应文本生成，获得会话文本的响应文本的方式实现，或者，也可通过根据服务关键项、服务关键信息和/或服务动作序列以及文本模板，生成会话文本的响应文本的方式实现；其中，在根据服务关键项、服务关键信息和/或服务动作序列以及文本模板，生成会话文本的响应文本的过程中，可获取服务动作序列对应的文本模板，将服务关键信息填充至文本模板对应的服务关键项的填充位，获得会话文本的响应文本。比如会话文本为“我买的新能源理财产品跌了怎么办”，响应文本为“最近新能源理财产品确实有少许下跌，您先不要着急，虽然新能源理财产品最近下跌了，但是整体估值还较低，您耐心等待一下”。
所述子服务是指搭载于资源管理服务以进行运行的子应用程序，比如资源管理服务为理财服务，资源管理服务的子服务为理财产品服务；比如会话文本为“打开a(代表资源管理服务的子服务)”，服务动作序列为“请求确认->渲染入口”，则可按照“请求确认”的服务动作生成会话文本的响应文本，并基于“渲染入口”的服务动作渲染a的服务入口。
可选的，所述目标服务包括资源管理服务的子服务或者第三方应用服务；其中，所述第三方应用服务是指独立运行于终端设备的应用程序；比如，会话文本为“打开b(代表目标服务)”，服务动作序列为“跳转”，基于“跳转”服务动作序列从资源管理服务跳转至目标服务b。
需要说明的是，上述执行服务动作序列对应的服务处理的过程，可被替换为按照所述服务动作序列生成所述会话文本的响应文本，基于所述服务动作序列渲染所述资源管理服务的子服务的服务入口，或者，基于所述服务动作序列从所述资源管理服务跳转至目标服务，并与本实施例提供的其他处理步骤组成新的实现方式。
如图3所示的服务处理的处理过程，获取用户u在理财服务输入的会话文本和在理财服务的历史会话信息，将会话文本和历史会话信息输入NLU模块，在NLU模块进行意图识别和特征解析，获取服务意图和服务关键特征，同时将用户u在理财服务的用户标识输入DST模块，DST模块基于用户标识读取用户画像信息，从而根据用户画像信息确定用户的会话偏好，将服务意图、服务关键特征和会话偏好输入服务动作模型进行服务动作确定，获得服务动作序列，此外，也可将服务意图、服务关键特征、用户画像信息和历史会话信息输入服务动作模型进行服务动作确定，获得服务动作序列，并将服务动作序列输入NLG(Natural Language Generation，自然语言生成)模块，由NLG模块按照服务动作序列生成会话文本的响应文本。
综上所述，本实施例提供的服务处理方法，首先对用户在资源管理服务输入的会话文本进行意图识别获得服务意图，并对用户在资源管理服务的历史会话信息和会话文本进行特征解析，获得服务关键特征，其次通过服务动作模型对服务意图、服务关键特征、用户的会话偏好和历史会话信息进行服务动作确定，并按照进行服务动作确定获得的服务动作序列生成会话文本的响应文本，基于进行服务动作确定获得的服务动作序列渲染资源管理服务的子服务的服务入口，和/或，基于进行服务动作确定获得的服务动作序列从资源管理服务跳转至目标服务，以此，以此，通过在确定服务动作序列的过程中引入会话偏好，来确保针对用户的会话偏好决策出服务动作序列，提升确定的服务动作序列的灵活性和多样性，能够针对不同的用户决策出不同的服务动作序列，满足用户的多样化需求，并且，通过从多个方面确定服务动作序列，提升服务动作序列的有效性和精确度。
下述以本实施例提供的一种服务处理方法在理财服务场景的应用为例，对本实施例提供的服务处理方法进行进一步说明，参见图4，应用于理财服务场景的服务处理方法，具体包括如下步骤。
步骤S402，获取用户在理财服务的历史会话信息和输入的会话文本。
步骤S404，对会话文本进行意图识别获得服务意图。
步骤S406，在理财服务的关键项集中查找服务意图对应的服务关键项。
步骤S408，根据历史会话信息、会话文本和服务关键项，确定服务关键信息。
步骤S410，将服务意图、服务关键项、服务关键信息、用户的会话偏好和历史会话信息输入服务动作模型进行服务动作确定，获得服务动作序列。
步骤S412，按照服务动作序列生成会话文本的响应文本。
上述步骤S412可被替换为按照服务动作序列生成会话文本的响应文本，基于服务动作序列渲染理财服务的子服务的服务入口，和/或，基于服务动作序列从理财服务跳转至目标服务，并与本实施例提供的其他处理步骤组成新的实现方式。
本说明书提供的一种服务处理装置实施例如下：在上述的实施例中，提供了一种服务处理方法，与之相对应的，还提供了一种服务处理装置，下面结合附图进行说明。
参照图5，其示出了本实施例提供的一种服务处理装置示意图。
由于装置实施例对应于方法实施例，所以描述得比较简单，相关的部分请参见上述提供的方法实施例的对应说明即可。下述描述的装置实施例仅仅是示意性的。
本实施例提供一种服务处理装置，包括：会话信息获取模块502，被配置为获取用户在资源管理服务的历史会话信息和输入的会话文本；特征解析模块504，被配置为对所述会话文本进行意图识别获得服务意图，并对所述历史会话信息和所述会话文本进行特征解析，获得服务关键特征；服务动作确定模块506，被配置为根据所述服务意图、所述服务关键特征和所述用户的会话行为特征，确定所述会话文本的服务动作序列；服务处理模块508，被配置为执行所述服务动作序列对应的服务处理，以响应所述会话文本。
本说明书提供的一种服务处理设备实施例如下：对应上述描述的一种服务处理方法，基于相同的技术构思，本说明书一个或多个实施例还提供一种服务处理设备，该服务处理设备用于执行上述提供的服务处理方法，图6为本说明书一个或多个实施例提供的一种服务处理设备的结构示意图。
本实施例提供的一种服务处理设备，包括：如图6所示，服务处理设备可因配置或性能不同而产生比较大的差异，可以包括一个或一个以上的处理器601和存储器602，存储器602中可以存储有一个或一个以上存储应用程序或数据。其中，存储器602可以是短暂存储或持久存储。存储在存储器602的应用程序可以包括一个或一个以上模块(图 示未示出)，每个模块可以包括服务处理设备中的一系列计算机可执行指令。更进一步地，处理器601可以设置为与存储器602通信，在服务处理设备上执行存储器602中的一系列计算机可执行指令。服务处理设备还可以包括一个或一个以上电源603，一个或一个以上有线或无线网络接口604，一个或一个以上输入/输出接口605，一个或一个以上键盘606等。
在一个具体的实施例中，服务处理设备包括有存储器，以及一个或一个以上的程序，其中一个或者一个以上程序存储于存储器中，且一个或者一个以上程序可以包括一个或一个以上模块，且每个模块可以包括对服务处理设备中的一系列计算机可执行指令，且经配置以由一个或者一个以上处理器执行该一个或者一个以上程序包含用于进行以下计算机可执行指令：获取用户在资源管理服务的历史会话信息和输入的会话文本；对所述会话文本进行意图识别获得服务意图，并对所述历史会话信息和所述会话文本进行特征解析，获得服务关键特征；根据所述服务意图、所述服务关键特征和所述用户的会话行为特征，确定所述会话文本的服务动作序列；执行所述服务动作序列对应的服务处理，以响应所述会话文本。
本说明书提供的一种存储介质实施例如下：对应上述描述的一种服务处理方法，基于相同的技术构思，本说明书一个或多个实施例还提供一种存储介质。
本实施例提供的存储介质，用于存储计算机可执行指令，所述计算机可执行指令在被处理器执行时实现以下流程：获取用户在资源管理服务的历史会话信息和输入的会话文本；对所述会话文本进行意图识别获得服务意图，并对所述历史会话信息和所述会话文本进行特征解析，获得服务关键特征；根据所述服务意图、所述服务关键特征和所述用户的会话行为特征，确定所述会话文本的服务动作序列；执行所述服务动作序列对应的服务处理，以响应所述会话文本。
需要说明的是，本说明书中关于一种存储介质的实施例与本说明书中关于一种服务处理方法的实施例基于同一发明构思，因此该实施例的具体实施可以参见前述对应方法的实施，重复之处不再赘述。
上述对本说明书特定实施例进行了描述。其它实施例在所附权利要求书的范围内。在一些情况下，在权利要求书中记载的动作或步骤可以按照不同于实施例中的顺序来执行并且仍然可以实现期望的结果。另外，在附图中描绘的过程不一定要求示出的特定顺序或者连续顺序才能实现期望的结果。在某些实施方式中，多任务处理和并行处理也是可以的或者可能是有利的。
在20世纪30年代，对于一个技术的改进可以很明显地区分是硬件上的改进(例如，对二极管、晶体管、开关等电路结构的改进)还是软件上的改进(对于方法流程的改进)。然而，随着技术的发展，当今的很多方法流程的改进已经可以视为硬件电路结构的直接改进。设计人员几乎都通过将改进的方法流程编程到硬件电路中来得到相应的硬件电路结构。因此，不能说一个方法流程的改进就不能用硬件实体模块来实现。例如，可编程逻辑器件(Programmable Logic Device，PLD)(例如现场可编程门阵列(Field Programmable Gate Array，FPGA))就是这样一种集成电路，其逻辑功能由用户对器件编程来确定。由设计人员自行编程来把一个数字***“集成”在一片PLD上，而不需要请芯片制造厂商来设计和制作专用的集成电路芯片。而且，如今，取代手工地制作集成电路芯片，这种编程也多半改用“逻辑编译器(logic compiler)”软件来实现，它与程序开发撰写时所用的软件编译器相类似，而要编译之前的原始代码也得用特定的编程语言来撰写，此称之为硬件描述语言(Hardware Description Language，HDL)，而HDL也并非仅有一种，而是有许多种，如ABEL(Advanced Boolean Expression Language)、AHDL(Altera Hardware Description Language)、Confluence、CUPL(Cornell University Programming Language)、HDCal、JHDL(Java Hardware Description Language)、Lava、Lola、MyHDL、PALASM、RHDL(Ruby Hardware Description Language)等，目前最普遍使用的是VHDL(Very-High-Speed Integrated Circuit Hardware Description Language)与Verilog。本领域技术人员也应该清楚，只需要将方法流程用上述几种硬件描述语言稍作逻辑编程并编程到集成电路中，就可以很容易得到实现该逻辑方法流程的硬件电路。
控制器可以按任何适当的方式实现，例如，控制器可以采取例如微处理器或处理器以及存储可由该(微)处理器执行的计算机可读程序代码(例如软件或固件)的计算机可读介质、逻辑门、开关、专用集成电路(Application Specific Integrated Circuit，ASIC)、可编程逻辑控制器和嵌入微控制器的形式，控制器的例子包括但不限于以下微控制器：ARC 625D、Atmel AT91SAM、Microchip PIC18F26K20以及Silicone Labs C8051F320，存储器控制器还可以被实现为存储器的控制逻辑的一部分。本领域技术人员也知道，除了以纯计算机可读程序代码方式实现控制器以外，完全可以通过将方法步骤进行逻辑编程来使得控制器以逻辑门、开关、专用集成电路、可编程逻辑控制器和嵌入微控制器等的形式来实现相同功能。因此这种控制器可以被认为是一种硬件部件，而对其内包括的用于实现各种功能的装置也可以视为硬件部件内的结构。或者甚至，可以将用于实现各种功能的装置视为既可以是实现方法的软件模块又可以是硬件部件内的结构。
上述实施例阐明的***、装置、模块或单元，具体可以由计算机芯片或实体实现， 或者由具有某种功能的产品来实现。一种典型的实现设备为计算机。具体的，计算机例如可以为个人计算机、膝上型计算机、蜂窝电话、相机电话、智能电话、个人数字助理、媒体播放器、导航设备、电子邮件设备、游戏控制台、平板计算机、可穿戴设备或者这些设备中的任何设备的组合。
为了描述的方便，描述以上装置时以功能分为各种单元分别描述。当然，在实施本说明书实施例时可以把各单元的功能在同一个或多个软件和/或硬件中实现。
本领域内的技术人员应明白，本说明书一个或多个实施例可提供为方法、***或计算机程序产品。因此，本说明书一个或多个实施例可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且，本说明书可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
本说明书是参照根据本说明书实施例的方法、设备(***)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。
这些计算机程序指令也可装载到计算机或其他可编程数据处理设备上，使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。
在一个典型的配置中，计算设备包括一个或多个处理器(CPU)、输入/输出接口、网络接口和内存。
内存可能包括计算机可读介质中的非永久性存储器，随机存取存储器(RAM)和/ 或非易失性内存等形式，如只读存储器(ROM)或闪存(flash RAM)。内存是计算机可读介质的示例。
计算机可读介质包括永久性和非永久性、可移动和非可移动媒体可以由任何方法或技术来实现信息存储。信息可以是计算机可读指令、数据结构、程序的模块或其他数据。计算机的存储介质的例子包括，但不限于相变内存(PRAM)、静态随机存取存储器(SRAM)、动态随机存取存储器(DRAM)、其他类型的随机存取存储器(RAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、快闪记忆体或其他内存技术、只读光盘只读存储器(CD-ROM)、数字多功能光盘(DVD)或其他光学存储、磁盒式磁带，磁盘存储或其他磁性存储设备或任何其他非传输介质，可用于存储可以被计算设备访问的信息。按照本文中的界定，计算机可读介质不包括暂存电脑可读媒体(transitory media)，如调制的数据信号和载波。
还需要说明的是，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、商品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、商品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括所述要素的过程、方法、商品或者设备中还存在另外的相同要素。
本说明书一个或多个实施例可以在由计算机执行的计算机可执行指令的一般上下文中描述，例如程序模块。一般地，程序模块包括执行特定任务或实现特定抽象数据类型的例程、程序、对象、组件、数据结构等等。也可以在分布式计算环境中实践本说明书的一个或多个实施例，在这些分布式计算环境中，由通过通信网络而被连接的远程处理设备来执行任务。在分布式计算环境中，程序模块可以位于包括存储设备在内的本地和远程计算机存储介质中。
本说明书中的各个实施例均采用递进的方式描述，各个实施例之间相同相似的部分互相参见即可，每个实施例重点说明的都是与其他实施例的不同之处。尤其，对于***实施例而言，由于其基本相似于方法实施例，所以描述的比较简单，相关之处参见方法实施例的部分说明即可。
以上所述仅为本文件的实施例而已，并不用于限制本文件。对于本领域技术人员来说，本文件可以有各种更改和变化。凡在本文件的精神和原理之内所作的任何修改、等同替换、改进等，均应包含在本文件的权利要求范围之内。

Claims (16)

  1. 一种服务处理方法，包括：
    获取用户在资源管理服务的历史会话信息和输入的会话文本；
    对所述会话文本进行意图识别获得服务意图，并对所述历史会话信息和所述会话文本进行特征解析，获得服务关键特征；
    根据所述服务意图、所述服务关键特征和所述用户的会话行为特征，确定所述会话文本的服务动作序列；
    执行所述服务动作序列对应的服务处理，以响应所述会话文本。
  2. 根据权利要求1所述的服务处理方法，所述执行所述服务动作序列对应的服务处理，包括：
    按照所述服务动作序列生成所述会话文本的响应文本，基于所述服务动作序列渲染所述资源管理服务的子服务的服务入口，和/或，基于所述服务动作序列从所述资源管理服务跳转至目标服务。
  3. 根据权利要求1所述的服务处理方法，所述对所述历史会话信息和所述会话文本进行特征解析，获得服务关键特征，包括：
    在所述资源管理服务的关键项集中查找所述服务意图对应的服务关键项；
    根据所述历史会话信息、所述会话文本和所述服务关键项，确定服务关键信息。
  4. 根据权利要求1所述的服务处理方法，所述根据所述服务意图、所述服务关键特征和所述用户的会话行为特征，确定所述会话文本的服务动作序列，包括：
    将所述服务意图、所述服务关键特征、所述用户的会话偏好和所述历史会话信息输入服务动作模型进行服务动作确定，获得所述服务动作序列；
    其中，所述会话偏好基于用户属性信息和所述用户在所述资源管理服务的资源管理记录确定。
  5. 根据权利要求4所述的服务处理方法，所述服务动作模型，采用如下方式训练：
    将样本会话信息输入待训练模型进行会话指标计算，获得会话指标；
    根据所述会话指标、样本会话偏好和所述样本会话信息对所述待训练模型进行参数更新；所述样本会话偏好在将所述样本会话信息输入偏好检测模型进行偏好检测后获得。
  6. 根据权利要求5所述的服务处理方法，所述根据所述会话指标、样本会话偏好和所述样本会话信息对所述待训练模型进行参数更新，包括：
    根据所述样本会话偏好和前一会话偏好计算偏好损失，并根据所述会话指标和目标会话指标计算会话指标损失；
    根据所述偏好损失和所述会话指标损失进行所述参数更新；
    其中，所述目标会话指标基于指标函数对所述样本会话信息进行会话指标计算获得；所述样本会话偏好和所述前一会话偏好对应于所述样本会话信息中的同一会话单元。
  7. 根据权利要求5所述的服务处理方法，所述样本会话信息，采用如下方式获得：
    从数据仓库读取基于会话规则进行用户会话交互获得的会话日志；所述会话规则针对所述资源管理服务进行冷启动部署；
    根据所述会话日志构建候选会话序列，并根据所述会话日志对应的子服务类别从所述候选会话序列中筛选出目标会话序列作为所述样本会话信息。
  8. 根据权利要求7所述的服务处理方法，所述根据所述会话日志构建候选会话序列，包括：
    确定所述会话日志中会话因子的用户响应参数，并将所述会话日志和所述用户响应参数作为所述候选会话序列；
    其中，所述会话因子，包括下述至少一项：会话停留时长、会话点击率、会话次数。
  9. 根据权利要求1所述的服务处理方法，所述对所述会话文本进行意图识别获得服务意图，并对所述历史会话信息和所述会话文本进行特征解析，获得服务关键特征步骤执行之后，还包括：
    判断所述服务意图和所述服务关键特征与所述历史会话信息中的历史会话状态是否匹配；
    若否，根据所述服务意图和所述服务关键特征，对所述历史会话信息中的历史会话状态进行修正处理；
    若是，执行所述根据所述服务意图、所述服务关键特征和所述用户的会话行为特征，确定所述会话文本的服务动作序列步骤。
  10. 根据权利要求1所述的服务处理方法，所述会话行为特征，通过如下方式确定：
    从用户属性信息和资源管理记录中提取关键词；
    根据所述关键词和预设会话偏好的偏好特征计算匹配度，并基于所述匹配度确定所述用户的会话偏好，将所述会话偏好和所述历史会话信息作为所述会话行为特征。
  11. 根据权利要求3所述的服务处理方法，所述根据所述历史会话信息、所述会话文本和所述服务关键项，确定服务关键信息，包括：
    基于所述服务关键项在所述历史会话信息和所述会话文本中提取服务关键信息；
    若未提取到服务关键信息，基于所述服务关键项在数据库中查询对应的服务关键信息。
  12. 根据权利要求1所述的服务处理方法，所述获取用户在资源管理服务的历史会话信息和输入的会话文本步骤执行之后，还包括：
    判断所述会话文本和所述历史会话信息中的历史会话文本是否相同；
    若否，执行所述对所述会话文本进行意图识别获得服务意图，并对所述历史会话信息和所述会话文本进行特征解析，获得服务关键特征步骤。
  13. 根据权利要求12所述的服务处理方法，若所述判断所述会话文本和所述历史会话信息中的历史会话文本是否相同操作执行之后的执行结果为是，执行如下操作：
    根据所述会话文本的第一数量和所述历史会话文本的第二数量计算全局文本数量；
    基于所述全局文本数量对所述历史会话信息中的历史服务动作序列进行更新处理，并按照更新后的服务动作序列生成所述会话文本的响应文本。
  14. 一种服务处理装置，包括：
    会话信息获取模块，被配置为获取用户在资源管理服务的历史会话信息和输入的会话文本；
    特征解析模块，被配置为对所述会话文本进行意图识别获得服务意图，并对所述历史会话信息和所述会话文本进行特征解析，获得服务关键特征；
    服务动作确定模块，被配置为根据所述服务意图、所述服务关键特征和所述用户的会话行为特征，确定所述会话文本的服务动作序列；
    服务处理模块，被配置为执行所述服务动作序列对应的服务处理，以响应所述会话文本。
  15. 一种服务处理设备，包括：
    处理器；以及，被配置为存储计算机可执行指令的存储器，所述计算机可执行指令在被执行时使所述处理器：
    获取用户在资源管理服务的历史会话信息和输入的会话文本；
    对所述会话文本进行意图识别获得服务意图，并对所述历史会话信息和所述会话文本进行特征解析，获得服务关键特征；
    根据所述服务意图、所述服务关键特征和所述用户的会话行为特征，确定所述会话文本的服务动作序列；
    执行所述服务动作序列对应的服务处理，以响应所述会话文本。
  16. 一种存储介质，用于存储计算机可执行指令，所述计算机可执行指令在被处理器执行时实现以下流程：
    获取用户在资源管理服务的历史会话信息和输入的会话文本；
    对所述会话文本进行意图识别获得服务意图，并对所述历史会话信息和所述会话文本进行特征解析，获得服务关键特征；
    根据所述服务意图、所述服务关键特征和所述用户的会话行为特征，确定所述会话文本的服务动作序列；
    执行所述服务动作序列对应的服务处理，以响应所述会话文本。
PCT/CN2023/131653 2022-11-28 2023-11-14 服务处理方法及装置 WO2024114374A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211501052.8 2022-11-28
CN202211501052.8A CN115840802A (zh) 2022-11-28 2022-11-28 服务处理方法及装置

Publications (1)

Publication Number Publication Date
WO2024114374A1 true WO2024114374A1 (zh) 2024-06-06

Family

ID=85576149

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/131653 WO2024114374A1 (zh) 2022-11-28 2023-11-14 服务处理方法及装置

Country Status (2)

Country Link
CN (1) CN115840802A (zh)
WO (1) WO2024114374A1 (zh)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115840802A (zh) * 2022-11-28 2023-03-24 蚂蚁财富(上海)金融信息服务有限公司 服务处理方法及装置
CN116861861A (zh) * 2023-07-06 2023-10-10 百度(中国)有限公司 文本处理方法及装置、电子设备和存储介质
CN116881429B (zh) * 2023-09-07 2023-12-01 四川蜀天信息技术有限公司 一种基于多租户的对话模型交互方法、装置及存储介质
CN117094376B (zh) * 2023-10-19 2024-02-23 浪潮电子信息产业股份有限公司 一种任务处理方法、装置、***、设备及可读存储介质

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106326452A (zh) * 2016-08-26 2017-01-11 宁波薄言信息技术有限公司 一种基于上下文的人机对话方法
CN111782792A (zh) * 2020-08-05 2020-10-16 支付宝(杭州)信息技术有限公司 用于信息处理的方法和装置
CN112735374A (zh) * 2020-12-29 2021-04-30 北京三快在线科技有限公司 一种自动语音交互的方法及装置
CN114528386A (zh) * 2022-01-14 2022-05-24 北京健康之家科技有限公司 机器人的外呼控制方法、装置、存储介质和终端
CN114862286A (zh) * 2022-07-06 2022-08-05 北京云迹科技股份有限公司 酒店客房服务判断方法、装置、电子设备及存储介质
CN115170330A (zh) * 2022-07-19 2022-10-11 支付宝(杭州)信息技术有限公司 应用于资源管理的服务处理方法及装置
CN115840802A (zh) * 2022-11-28 2023-03-24 蚂蚁财富(上海)金融信息服务有限公司 服务处理方法及装置

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106326452A (zh) * 2016-08-26 2017-01-11 宁波薄言信息技术有限公司 一种基于上下文的人机对话方法
CN111782792A (zh) * 2020-08-05 2020-10-16 支付宝(杭州)信息技术有限公司 用于信息处理的方法和装置
CN112735374A (zh) * 2020-12-29 2021-04-30 北京三快在线科技有限公司 一种自动语音交互的方法及装置
CN114528386A (zh) * 2022-01-14 2022-05-24 北京健康之家科技有限公司 机器人的外呼控制方法、装置、存储介质和终端
CN114862286A (zh) * 2022-07-06 2022-08-05 北京云迹科技股份有限公司 酒店客房服务判断方法、装置、电子设备及存储介质
CN115170330A (zh) * 2022-07-19 2022-10-11 支付宝(杭州)信息技术有限公司 应用于资源管理的服务处理方法及装置
CN115840802A (zh) * 2022-11-28 2023-03-24 蚂蚁财富(上海)金融信息服务有限公司 服务处理方法及装置

Also Published As

Publication number Publication date
CN115840802A (zh) 2023-03-24

Similar Documents

Publication Publication Date Title
WO2024114374A1 (zh) 服务处理方法及装置
KR102288249B1 (ko) 정보 처리 방법, 단말기, 및 컴퓨터 저장 매체
US10831811B2 (en) Resolution of ambiguous and implicit references using contextual information
CN112799747A (zh) 智能助理评价、推荐方法、***、终端及可读存储介质
TW202020691A (zh) 特徵詞的確定方法、裝置和伺服器
WO2018040069A1 (zh) 信息推荐***及方法
US11586689B2 (en) Electronic apparatus and controlling method thereof
JP2012118977A (ja) 文書類似性計算の機械学習に基づく最適化およびカスタマイズのための方法およびシステム
US20190114711A1 (en) Financial analysis system and method for unstructured text data
CN110162780A (zh) 用户意图的识别方法和装置
CN116127020A (zh) 生成式大语言模型训练方法以及基于模型的搜索方法
EP4060517A1 (en) System and method for designing artificial intelligence (ai) based hierarchical multi-conversation system
CN111813900B (zh) 多轮对话处理方法、装置、电子设备及存储介质
CN111767394A (zh) 一种基于人工智能专家***的摘要提取方法及装置
CN113516480A (zh) 一种支付风险识别方法、装置及设备
CN109710732A (zh) 信息查询方法、装置、存储介质和电子设备
CN113886033A (zh) 一种任务处理的方法及装置
CN114580402A (zh) 一种面向企业的产品信息获取方法、装置、服务器和存储介质
CN117575008A (zh) 训练样本生成方法、模型训练方法、知识问答方法及设备
CN108595395B (zh) 一种昵称的生成方法、装置及设备
CN115186179A (zh) 保险产品推送方法和装置
CN112507214B (zh) 基于用户名的数据处理方法、装置、设备及介质
CN115017914A (zh) 语言处理方法、装置、电子设备以及存储介质
CN114115878A (zh) 一种工作流节点推荐方法及装置
CN110428814B (zh) 一种语音识别的方法及装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23896526

Country of ref document: EP

Kind code of ref document: A1