Information processing method and device, electronic equipment and computer readable storage medium

By identifying the intent in user interaction information of the model group, the problem of not being able to recognize marketing intentions in natural language expressions in the prior art is solved, and more efficient user intention recognition is achieved.

CN119088595APending Publication Date: 2024-12-06BEIJING BAIDU NETCOM SCI & TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411170666.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-23
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

The prior art is difficult to effectively identify the marketing intention of users in natural language expression, especially in the absence of interactive structure sub-information, which leads to the inability of users to express their marketing demands freely.

Method used

By receiving user interaction information, detect whether there is interactive structure sub-information. If not, the pending information is extracted and multiple identification model groups (such as rule models, long and short-term memory models, RAG similarity models, large models, etc.) are used to identify users. to improve the accuracy and efficiency of intention recognition.

Benefits of technology

It realizes accurate identification of user intentions without interactive structure sub-information, improves the accuracy and efficiency of information recognition, and supports users to freely express marketing demands in natural language expression.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119088595A_ABST
    Figure CN119088595A_ABST
Patent Text Reader

Abstract

The invention provides an information processing method and device, and relates to the technical field of computers, in particular to the technical fields of large models, deep learning, natural language processing and the like. According to the specific implementation scheme, user interaction information sent by a front end is received; detecting whether interaction structure sub-information exists in the user interaction information or not; in response to the detection that the user interaction information does not have the interaction structure sub-information, obtaining to-be-processed information based on the user interaction information; and identifying the to-be-processed information by adopting at least one identification model group to obtain an intention identification result of the user interaction information.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the field of computers, specifically to technical fields such as large models, deep learning, and natural language processing, and in particular to an information processing method and device, an electronic device, a computer-readable storage medium, and a computer program product. Background Art

[0002] In the era of big models, marketing and promotion carriers have evolved from traditional marketing and promotion platforms to marketing intelligent entities, and advertisers' marketing intention expression has also changed from GUI (Graphical User Interface) form submission to LGUI (open source GUI system for embedded Linux). Among them, LGUI is mainly expressed in natural language, supplemented by GUI expression.

[0003] In traditional technology, one way to identify advertisers' intent is based on dialogue template configuration and word slot extraction. In this method, users can only have conversations based on the style of the template, and natural language expressed by users that deviates from the template style cannot be parsed. As a result, users cannot freely express their marketing demands. Summary of the invention

[0004] The present disclosure provides an information processing method and apparatus, an electronic device, a computer-readable storage medium, and a computer program product.

[0005] According to a first aspect, an information processing method is provided, the method comprising: receiving user interaction information sent by a front end; detecting whether the user interaction information contains interaction structure sub-information; in response to detecting that the user interaction information does not contain interaction structure sub-information, obtaining information to be processed based on the user interaction information; using at least one recognition model group to identify the information to be processed, and obtaining an intention recognition result of the user interaction information.

[0006] According to a second aspect, another information processing device is provided, which includes: a receiving unit, configured to receive user interaction information sent by a front end; a detection unit, configured to detect whether the user interaction information contains interaction structure sub-information; an information obtaining unit, configured to obtain information to be processed based on the user interaction information in response to detecting that the user interaction information does not contain interaction structure sub-information; and a result obtaining unit, configured to use at least one recognition model group to identify the information to be processed and obtain an intention recognition result of the user interaction information.

[0007] According to a third aspect, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method described in any implementation manner of the first aspect.

[0008] According to a fourth aspect, a non-transitory computer-readable storage medium storing computer instructions is provided, where the computer instructions are used to cause a computer to execute the method described in any implementation of the first aspect.

[0009] According to a fifth aspect, a computer program product is provided, comprising a computer program, which implements the method described in any implementation manner of the first aspect when executed by a processor.

[0010] The information processing method and device provided by the embodiments of the present disclosure firstly receive the user interaction information sent by the front end; secondly, detect whether there is interaction structure sub-information in the user interaction information; thirdly, in response to detecting that there is no interaction structure sub-information in the user interaction information, obtain information to be processed based on the user interaction information; finally, use at least one recognition model group to identify the information to be processed to obtain the intention recognition result of the user interaction information. Therefore, when it is recognized that there is no interaction structure sub-information in the user interaction information, the information to be processed after the user interaction information is processed is identified by at least one recognition model group, so that the user's intention can be accurately identified, thereby improving the accuracy and efficiency of user information recognition.

[0011] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0013] Figure 1 is a flow chart of an embodiment of an information processing method according to the present disclosure;

[0014] Figure 2 It is a structural diagram of information sent by the front end in the present disclosure;

[0015] Figure 3 is a flow chart of another embodiment of the information processing method according to the present disclosure;

[0016] Figure 4 is a schematic structural diagram of an embodiment of an information processing device according to the present disclosure;

[0017] Figure 5 The block diagram is a block diagram of an electronic device for implementing the information processing method according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0018] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0019] One way to identify user intent is to map the corresponding intent based on RAG (Relative Gain Array) similarity matching. However, RAG-based similarity matching relies on a large amount of labeled data. The quality and quantity of labeled data have a great impact on the accuracy of intent recognition, increasing the cost and difficulty of data preparation. In addition, RAG may face challenges in dealing with ambiguity and context dependency. If the retrieved information is not accurate or relevant, the generated results may deviate from the user's actual intent.

[0020] Another way is: intent recognition based on a general large language model. Although the large model can understand the context information and infer the user's intent, the large model performs poorly in understanding expressions in some specific business fields. In addition, it takes a long time to reason with the large model to recognize the intent. In a dialogue system that requires a quick response, delays may affect the user experience.

[0021] In view of the defects in the prior art, the present invention proposes an information processing method. Figure 1 A process 100 according to an embodiment of the information processing method of the present disclosure is shown. The information processing method comprises the following steps:

[0022] Step 101: receiving user interaction information sent by the front end.

[0023] In this embodiment, the front end is the part that receives user input information. The front end of the website is responsible for displaying things visible to website users, such as special effects on the web page, web page layout, pictures, videos and other content.

[0024] In this embodiment, the execution subject on which the information processing method runs can be the backend, which is responsible for the website background logic, interacting with the frontend, and saving and reading the user and website data after interacting with the user. For example, user registration and login are available on general websites. The user's registration information is sent to the backend through the frontend, and the backend saves it in the database. When the user logs in to the website, the backend needs to determine whether the user has the right to log in by whether the user name and password entered by the user are consistent with those in the database.

[0025] In this embodiment, the user interaction information is information sent by the user to the information processing method through the front end, and the user interaction information may include control information of the control operated by the user, text information input by the user, and the user's last intention.

[0026] In this embodiment, the front end and the back end communicate with each other through a specific data transmission protocol, and the user interaction information has information in a specific format. For example, the front end and the back end communicate through the LGUI protocol. Figure 2 As shown, user interaction information includes: text segment T, card segment C, suggestion segment S, and instruction segment I. Among them, text segment T is text type data, which is the most basic interaction type in input and output, and carries the content of natural language dialogue between users and systems. Card segment C is a custom style for users to interact with the system. There are many types of card segments C, including simple components such as single choice and multiple choice, as well as complex components such as reports, landing pages, and regions, which are used to represent complex structured information in different scenarios (marketing scenarios). Suggestion segment S is an intent recommendation. During the dialogue between the backend and the user, some intents will be recommended to the user at a certain time to guide the customer to complete the operation he wants to express. Instruction segment I is a set of instructions used to control the displayed content of the GUI area. Based on different marketing scenarios and intent intelligence, different GUI instructions will be sent to the front end.

[0027] In the technical solution disclosed in this disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user interaction information are carried out after authorization and in compliance with relevant laws and regulations. The information related to the user in the data operation request is obtained after the user's permission, and the processing of information related to the user is carried out under confidentiality conditions.

[0028] Step 102: Detect whether the user interaction information contains interaction structure sub-information.

[0029] In this embodiment, the interaction structure sub-information is the structured information in the user interaction information. The interaction structure sub-information can be used to determine the components and controls on the front-end displayed page operated by the user and the order of the components and controls operated. Specifically, the interaction structure sub-information can be the above Figure 2 The content in the middle card fragment C.

[0030] In this embodiment, the above step 102 includes: detecting whether there is content in the card segment C in the user interaction information; and in response to there being content in the card segment C in the user interaction information, determining that the user interaction information has interaction structure sub-information.

[0031] In this embodiment, the above step 102 may also include: matching the custom style information with the user interaction information, and in response to the custom style information successfully matching the user interaction information, determining that the user interaction information contains interaction structure sub-information. The custom style information is the style of the component or control displayed on the front-end page.

[0032] Step 103: in response to detecting that the user interaction information does not contain the interaction structure sub-information, obtaining information to be processed based on the user interaction information.

[0033] In this embodiment, the information to be processed is information that can be recognized by multiple recognition models. Generally, the information to be processed can be user input information obtained by extracting key information from user interaction information. The recognition model is used to perform intent recognition on the information to be processed to obtain an intent recognition result that reflects the user's current intention.

[0034] In this embodiment, the above step 103 includes: extracting text information input by the user in the user interaction information to obtain information to be processed.

[0035] Optionally, the user interaction information is information having a specific format, such as Figure 2 As shown, text-type data in the user interaction information can be directly extracted, and the extracted text-type data can be cleaned to obtain data to be processed.

[0036] Step 104: Use at least one recognition model group to recognize the information to be processed, and obtain an intention recognition result of the user interaction information.

[0037] In this embodiment, at least one recognition model group includes at least one recognition model group, and the recognition model group includes at least one recognition model. The recognition models in each recognition model group may have the same recognition principle or may not have the same recognition principle.

[0038] In this embodiment, the types of recognition models can be flexibly expanded according to business needs. Specifically, the recognition models may include the following types:

[0039] Rule model: Based on specific business rules in marketing scenarios, regular expressions, templates, word segmentation and other matching methods are used to identify intent. The recognition accuracy of this model is close to 100%, and the recognition time is at the millisecond level.

[0040] Long-short-term memory model: Build a long-short-term memory library, and obtain the corresponding intent from the memory data based on the user information and context information recognized by historical models such as large models or SFT small models. The recognition accuracy of this model is close to 100%, and the recognition time is at the ms level.

[0041] RAG similarity model: embeds the query and intent mapping relationship in the long and short-term memory and stores it in the vector library. It finds the corresponding intent after generalizing the user's information for similarity recognition. The recognition accuracy of this model is over 95%, and the recognition time is 1-2 seconds.

[0042] Big model: Based on the semantic understanding and contextual reasoning capabilities of the big model, the user's query is identified to correspond to the intent. The expert's recognition accuracy is over 99% and the recognition time is over 2s.

[0043] SFT small model: Opens up the data flywheel, maps the historical conversation information to the intent through supervised training and publishes a small model expert to identify the intent corresponding to the user query. The expert has an accuracy rate of 95%+ and the recognition time is about 1 second.

[0044] In this embodiment, the priority of each recognition model group in at least one recognition model group can be set according to the system performance requirements, and each recognition model group can be controlled to recognize the information to be processed based on the priority of each recognition model group. For example, at least one recognition model group includes: a first recognition model group and a second recognition model group. The priority of the first recognition model group is higher than the priority of the second recognition model group. When recognizing the information to be processed, the recognition model in the first recognition model group can be used to preferentially recognize the information to be processed. When the first recognition model group cannot give an intention recognition result, the second recognition model group is controlled to recognize the information to be processed.

[0045] In this embodiment, the intent recognition result of the user interaction information is the result obtained after intent recognition of the user input information obtained by the front end. For different scenarios, the content of the intent recognition result is different. For example, for a marketing scenario, the intent recognition result of the user interaction information includes: the user's marketing intention.

[0046] The information processing method provided by the embodiment of the present disclosure first receives user interaction information sent by the front end; secondly, detects whether there is interaction structure sub-information in the user interaction information; thirdly, in response to detecting that there is no interaction structure sub-information in the user interaction information, obtains information to be processed based on the user interaction information; finally, uses at least one recognition model group to identify the information to be processed to obtain the intention recognition result of the user interaction information. Therefore, when it is recognized that there is no interaction structure sub-information in the user interaction information, the information to be processed after the user interaction information is processed is identified by at least one recognition model group, so that the user's intention can be accurately identified, thereby improving the accuracy and efficiency of user information recognition.

[0047] Figure 3 A process 300 according to an embodiment of the information processing method of the present disclosure is shown. The information processing method includes the following steps:

[0048] Step 301, receiving user interaction information sent by the front end, and then executing step 302.

[0049] Step 302 , detecting whether the user interaction information contains interaction structure sub-information; if it is detected that the user interaction information contains interaction structure sub-information, executing step 303 ; if it is detected that the user interaction information does not contain interaction structure sub-information, executing step 307 .

[0050] Step 303 , using a structured model to identify interaction structure sub-information, and then executing step 304 .

[0051] In this embodiment, the structured model is based on the expert who recognizes the content of the card fragment in the transmission message in the predefined front-end and back-end transmission protocol (LGUI protocol). Figure 2 As shown in the card fragment C shown in the figure, all the contents in the card fragment C are input into the structured model, and the structured model outputs the user's operation intention on the control or component on the interface displayed on the front end, that is, the user's intention recognition result. The recognition accuracy of the structured model is 100%, and the recognition time is at the ms level.

[0052] Step 304, detect whether the structured model outputs an intention recognition result; if it is detected that the structured model outputs an intention recognition result, execute step 305; if it is detected that the structured model does not output an intention recognition result, execute steps 307 to 308.

[0053] Step 305 , taking the intention recognition result output by the structured model as the intention recognition result of the user interaction information, and then executing step 306 .

[0054] Step 306, exit.

[0055] Step 307 , based on the user interaction information, obtain the information to be processed, and then execute step 308 .

[0056] Step 308 , using at least one recognition model group to recognize the information to be processed, obtaining an intention recognition result of the user interaction information, and executing step 306 .

[0057] In this embodiment, it should be understood that the operations and features in the above steps 301-302 and 307-308 correspond to the operations and features in steps 101-104, respectively. Therefore, the above descriptions of the operations and features in steps 101-104 are also applicable to steps 301-302 and 307-308 and will not be repeated here.

[0058] The information processing method provided in this embodiment adopts a structured model to identify the interaction structure sub-information when there is interaction structure sub-information in the user interaction information, and uses the intention recognition output by the structured model as the intention recognition result of the user interaction information, thereby providing a fast and reliable implementation method for obtaining the intention recognition result of the user interaction information; when the structured model does not output the intention recognition result, the information to be processed is obtained based on the user interaction information, and at least one recognition model group is used to identify the information to be processed, thereby ensuring the comprehensiveness and reliability of the user interaction information.

[0059] In some optional implementations of the present disclosure, the obtaining of the information to be processed based on the user interaction information includes: obtaining text type data based on the user interaction information; and preprocessing the text type data to obtain the information to be processed.

[0060] In this optional implementation, the text type data is data in text form, and the text type data is obtained by extracting information in the text field in the user interaction information. When the user interaction information is LGUI protocol data, the text segment T in the user interaction information is extracted to obtain the text type data.

[0061] In this optional implementation, preprocessing the text type data to obtain the information to be processed includes: performing word segmentation, text cleaning, standardization, and feature extraction on the text type data to obtain the information to be processed.

[0062] The method for obtaining the information to be processed provided by this optional implementation obtains text type data based on user interaction information; pre-processes the text type data to obtain the information to be processed, thereby providing a reliable implementation method for obtaining the information to be processed.

[0063] Optionally, the obtaining of the information to be processed based on the user interaction information includes: extracting text data in the user interaction information to obtain the information to be processed.

[0064] Optionally, the obtaining of the information to be processed based on the user interaction information includes: extracting subject-verb-object structure sentences in the user interaction information to obtain the information to be processed.

[0065] In some optional implementations of the present disclosure, the above-mentioned preprocessing of text type data to obtain information to be processed includes: removing ambiguous data in the text type data to obtain first text data; performing content normalization processing on the first text data to obtain second text data; changing the sentence structure of the second text data to obtain information to be processed.

[0066] In this optional implementation, ambiguous data refers to noisy and ambiguous data in text type data. For example, in text type data, "remove" and "eliminate", "remove" is removed. The purpose of disambiguation and noise reduction can be achieved by removing the ambiguous data in the text type data.

[0067] In this optional implementation, performing content normalization processing on the first text data means: expressing texts in the first text data that have the same meaning but different forms of expression in the same data form, and modifying the numbers, dates, and regions in the first text data in a unified manner. For example, all numbers are expressed in Arabic numerals, and all dates are expressed in the form of day, hour, and second.

[0068] In this optional implementation, changing the sentence structure of the second text data includes: detecting whether the sentence structure of the second text data is a subject-object structure; if not, modifying the sentence structure of the second text data to a subject-object structure. For example, the second text data is: I want to modify the budget to 100 yuan, and the normalized content is: Modify the budget.

[0069] In this optional implementation, if the sentence structure of the second text data is a subject-object structure, the subject of the second text data is removed to obtain the information to be processed.

[0070] The method for obtaining the information to be processed provided by this optional implementation method first removes ambiguous data in the text type data to obtain the first text data; performs content normalization processing on all the text data to obtain the second text data; changes the sentence structure of the second text data to obtain the information to be processed, thereby providing a reliable implementation method for obtaining the information to be processed.

[0071] Optionally, the above-mentioned preprocessing of the text type data to obtain the information to be processed includes: segmenting the text type data to obtain segmentation results; extracting verbs and recipients of actions corresponding to the verbs in the segmentation results, and taking each verb and recipient as the information to be processed.

[0072] In some optional implementations of the present disclosure, the above-mentioned use of at least one recognition model group to identify the information to be processed and obtain the intention recognition result of the user interaction information includes: based on the information to be processed, dynamically determining at least one recognition model group and the priority of each recognition model group from the recognition model set; controlling each recognition model group to identify the information to be processed according to its respective priority, and obtaining the intention recognition result of the user interaction information.

[0073] In this optional implementation, the execution subject on which the information processing method runs can dynamically determine the recognition model group and the priority of each recognition model group based on the content of the information to be processed. For example, when the information structure of the information to be processed matches the template structure of the rule recognition model, the rule recognition model and the long short-term memory expert are used as a recognition model group, and the first priority is set for the recognition model group.

[0074] The method for obtaining the intention recognition result of user interaction information provided by this optional implementation method determines the recognition model group and the priority of each recognition model group in the recognition model set based on the information to be processed; controls each recognition model group to work according to its own priority, and dynamically selects the recognition model based on the information to be processed, thereby improving the accuracy of the intention recognition result of user interaction information.

[0075] Optionally, the above-mentioned use of at least one recognition model group to identify the information to be processed and obtain the intention recognition result of the user interaction information includes: based on the recognition speed and efficiency of each recognition model in the recognition model set, dividing the recognition models in the recognition model set into at least one recognition model group, and pre-assigning corresponding priorities to the recognition model groups in at least one recognition model group, controlling the recognition model groups in at least one recognition model group to identify the information to be processed according to their respective priorities, and obtaining the intention recognition result of the user interaction information.

[0076] In some optional implementations of the present disclosure, the above-mentioned control of each recognition model group to identify the information to be processed according to their respective priorities to obtain the intention recognition result of the user interaction information includes: determining the execution order and confidence weight of the recognition models in each recognition model group; for the currently executed recognition model group, controlling the recognition models in the recognition model group to sequentially identify the information to be processed according to their respective execution orders; in response to detecting that the recognition model with a higher execution order in the recognition model group obtains the intention recognition result of the information to be processed, multiplying the intention recognition result by the confidence weight of the recognition model to obtain the intention recognition result of the user interaction information.

[0077] In this optional implementation, the currently executed recognition model group refers to the recognition model group that is determined according to the priority of each recognition model and that needs to be used to recognize the information to be processed. It should be noted that the higher the priority of the recognition model group, the earlier the information to be processed will be recognized. When the recognition model with a high priority cannot recognize the information to be processed, the recognition model with an adjacent priority to the high priority will be controlled to recognize the information to be processed.

[0078] In this optional implementation, the execution order is the number of times each recognition model recognizes the information to be processed in its respective recognition model group. For example, a recognition model group includes three recognition models: a first recognition model, a second recognition model, and a third recognition model. The execution orders of the three recognition models are execution order 1, execution order 1, and execution order 2, respectively. It can be determined that the first recognition model and the second recognition model recognize the information to be processed at the same time. When the first recognition model and the second recognition model are executed or have been executed for a preset time, the third recognition model is controlled to recognize the information to be processed.

[0079] In this optional implementation, the recognition model with a high execution order refers to the recognition model that needs to be executed in advance in the recognition model group. The high execution order is determined by the execution order and confidence weight of the recognition model. For example, when the execution order is in the top three and the confidence weight is more than 50%, the recognition model with a high execution order is the recognition model with a high execution order. Optionally, the recognition model with a high execution order can also be determined only by the execution order. For example, the recognition models with the top three execution orders are the recognition models with a high execution order.

[0080] In this optional implementation, the confidence weight is the proportion of the intent recognition results obtained by each recognition model in the recognition model group to the intent recognition results of the entire recognition model group.

[0081] The method for obtaining the intention recognition result of user interaction information provided by this optional implementation controls the recognition models in the recognition model group to sequentially identify the information to be processed according to their respective execution orders; when it is detected that the recognition model with a higher execution order in the recognition model group obtains the intention recognition result of the information to be processed, the intention recognition result obtained by the recognition model with a higher execution order is multiplied by the confidence weight of the recognition model to obtain the intention recognition result of the user interaction information, thereby improving the reliability of the intention recognition result of the user interaction information.

[0082] Optionally, each of the above-mentioned recognition model groups identifies the information to be processed according to its own priority, and obtains the intention recognition result of the user interaction information, including: determining the execution order and confidence weight of the recognition models in each recognition model group; for the currently executed recognition model group, in response to the fact that the execution order of each recognition model in the recognition model group is the same, controlling the recognition models in the recognition model group to simultaneously recognize the information to be processed, multiplying the intention recognition results of each recognition model in the recognition model group by their respective confidence weights and then adding them together to obtain the intention recognition result of the user interaction information.

[0083] In some optional implementations of the present disclosure, the above-mentioned control of each recognition model group to identify the information to be processed according to its respective priority level to obtain the intention recognition result of the user interaction information also includes: in response to detecting that a recognition model with a later execution order in the recognition model group obtains an intention recognition result, waiting for a recognition model with an earlier execution order in the recognition model group until the recognition model with an earlier execution order is detected to obtain the intention recognition result; multiplying the intention recognition result of the recognition model with a later execution order by the confidence weight of the recognition model with a later execution order plus the intention recognition result of the recognition model with an earlier execution order multiplied by the confidence weight of the recognition model with an earlier execution order to obtain the intention recognition result of the user interaction information.

[0084] In this optional implementation, the execution subject dynamically adjusts the expert execution order and confidence weight, and controls the recognition model to perform intent recognition according to the execution order and confidence weight. The one with an earlier execution order is executed first. If the intent can be recognized, it returns directly and ends the process; otherwise, it is transferred to the recognition model with an earlier execution order, and the recognition models with the same execution order are executed in parallel. If the recognition model with a larger confidence weight returns the intent recognition result first, it returns directly and ends the process, otherwise it waits for other recognition models to return, and so on, and finally identifies the intent recognition result of the user interaction information.

[0085] The method for obtaining the intention recognition result of user interaction information provided by this optional implementation manner, after detecting that the recognition model with a later execution order in the recognition model group has obtained the intention recognition result, waits for the recognition model with an earlier execution order in the recognition model group until the recognition model with the earlier execution order is detected to have obtained the intention recognition result, and multiplies the intention recognition result of the recognition model with a later execution order by the confidence weight of the recognition model with a later execution order plus the intention recognition result of the recognition model with an earlier execution order multiplied by the confidence weight of the recognition model with a later execution order, to obtain the intention recognition result of the user interaction information. As a result, the intention recognition result of the recognition model with an earlier execution order is emphasized, and the intention recognition results of the recognition models with an earlier execution order and a later execution order are comprehensively considered, thereby improving the reliability of the intention recognition result of the user interaction information.

[0086] Optionally, the above-mentioned control of at least one recognition model group in the recognition model group identifies the information to be processed according to their respective priorities, and obtaining the intention recognition result of the user interaction information also includes: in response to at least two recognition models with the same execution order both feeding back intention recognition results, comparing the confidence weights of the at least two recognition models with the same execution order, and taking the intention recognition result of the recognition model with the largest confidence weight as the intention recognition result of the user interaction information.

[0087] In some optional implementations of the present disclosure, the above-mentioned control of at least one recognition model group to identify the information to be processed according to their respective priorities to obtain the intention recognition result of the user interaction information includes: using a recognition model group with a high priority in at least one recognition model group to identify the information to be processed; detecting whether the recognition model group with a high priority outputs the intention recognition result; in response to detecting that the recognition model group with a high priority outputs the intention recognition result, using the intention recognition result output by the recognition model group with a high priority as the intention recognition result of the user interaction information.

[0088] In this optional implementation, the recognition model group with a high priority is the priority relative to the priorities of all recognition model groups in at least one recognition model group. The priorities of all recognition model groups in at least one recognition model group are sorted, and the knowledge expert group with a higher priority is a recognition model group with a higher priority relative to the knowledge expert group with a lower priority, and the knowledge expert group with a lower priority is a recognition model group with a lower priority relative to the knowledge expert group with a higher priority.

[0089] In this optional implementation, when the recognition model group with the highest priority among at least one recognition model group outputs an intent recognition result, other recognition model groups are no longer allowed to perform intent recognition on the information to be processed; when only the recognition model with the highest priority does not output an intent recognition result, recognition models with a lower priority than the highest priority are controlled to perform intent recognition on the information to be processed. Similarly, when a recognition model group with a high priority outputs an intent recognition result, a recognition model group with a low priority is no longer controlled to perform intent recognition on the information to be processed.

[0090] The method for obtaining the intention recognition result of user interaction information provided by this optional implementation first adopts a high-priority recognition model group to identify the information to be processed. When the high-priority recognition model outputs an intention recognition result, the intention recognition result output by the high-priority recognition model group is used as the intention recognition result of the user interaction information. Therefore, a reliable implementation method is provided for obtaining the intention recognition result of user interaction information.

[0091] In some optional implementations of the present disclosure, the above-mentioned control of at least one recognition model group in the recognition model group identifies the information to be processed according to their respective priorities to obtain the intention recognition result of the user interaction information also includes: in response to detecting that the recognition model group with a high priority does not output the intention recognition result, using a recognition model group with a low priority in at least one recognition model group to identify the information to be processed; detecting whether the recognition model group with a low priority outputs the intention recognition result; in response to detecting that the recognition model group with a low priority outputs the intention recognition result, using the intention recognition result output by the recognition model group with a low priority as the intention recognition result of the user interaction information.

[0092] This optional implementation provides a method for obtaining intent recognition results of user interaction information. When a high-priority recognition model group does not output an intent recognition result, a low-priority recognition model group is used to identify the information to be processed. After the low-priority recognition model group outputs the intent recognition result, the intent recognition result output by the low-priority recognition model group is used as the intent recognition result of the user interaction information, thereby ensuring the comprehensiveness of the intent recognition results of the user interaction information.

[0093] In one example of the present disclosure, the information processing method of the present disclosure can be applied in a marketing promotion business scenario. The main idea of ​​the method is to construct multiple recognition model groups, each recognition model group processes a specific type of input or subtask; dynamically select or weight the intention recognition results output by different recognition model groups according to user interaction information, and be responsible for summarizing and deriving the intention recognition results of user interaction information.

[0094] Different from traditional conversational robots that perform intent recognition through a single method such as rule configuration, similarity matching, and large models, the information processing method provided by the present invention supports intent recognition of text input, structured information, recommendation, and instruction information; by distributing different types of information to a group of recognition models for intent recognition, it has stronger recognition capabilities and higher accuracy than traditional single intent recognition models.

[0095] Further references Figure 4 As an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of an information processing device, which is similar to Figure 1 Corresponding to the method embodiment shown, the device can be specifically implemented in various electronic devices.

[0096] like Figure 4As shown, the information processing device 400 provided in this embodiment includes: a receiving unit 401, a detection unit 402, an information obtaining unit 403, and a result obtaining unit 404. Among them, the above-mentioned receiving unit 401 can be configured to receive user interaction information sent by the front end. The above-mentioned detection unit 402 can be configured to detect whether there is interaction structure sub-information in the user interaction information. The above-mentioned information obtaining unit 403 can be configured to obtain information to be processed based on the user interaction information in response to detecting that there is no interaction structure sub-information in the user interaction information. The above-mentioned result obtaining unit 404 can be configured to use at least one recognition model group to identify the information to be processed and obtain the intention recognition result of the user interaction information.

[0097] In this embodiment, the specific processing of the receiving unit 401, the detecting unit 402, the information obtaining unit 403, and the result obtaining unit 404 and the technical effects thereof can be referred to in detail. Figure 1 The relevant descriptions of step 101, step 102, step 103, and step 104 in the corresponding embodiment are not repeated here.

[0098] In some optional implementations of the present embodiment, the above-mentioned device 400 also includes: an identification unit (not shown in the figure), which is configured to, in response to detecting that the user interaction information contains interaction structure sub-information, use a structured model to identify the interaction structure sub-information; detect whether the structured model outputs an intention recognition result; in response to detecting that the structured model outputs an intention recognition result, use the intention recognition result output by the structured model as the intention recognition result of the user interaction information.

[0099] In some optional implementations of the present embodiment, the above-mentioned device 400 also includes: an intention obtaining unit (not shown in the figure), and the above-mentioned intention obtaining unit is configured to obtain information to be processed based on user interaction information in response to detecting that the structured model does not output an intention recognition result; use at least one recognition model group to identify the information to be processed, and obtain the intention recognition result of the user interaction information.

[0100] In some optional implementations of this embodiment, the information obtaining unit 403 is configured to: obtain text type data based on user interaction information; and perform normalization processing on the text type data to obtain information to be processed.

[0101] In some optional implementations of this embodiment, the above-mentioned information acquisition unit 403 is further configured to: remove ambiguous data in the text type data to obtain first text data; perform content normalization processing on the first text data to obtain second text data; change the sentence structure of the second text data to obtain information to be processed.

[0102] In some optional implementations of the present embodiment, the above-mentioned result obtaining unit 404 is configured to: dynamically determine at least one recognition model group and the priority of each recognition model group from the recognition model set based on the information to be processed; control each recognition model group to identify the information to be processed according to its respective priority, and obtain the intention recognition result of the user interaction information.

[0103] In some optional implementations of the present embodiment, the above-mentioned result obtaining unit 404 is further configured to: determine the execution order and confidence weight of the recognition models in each recognition model group; for the currently executed recognition model group, control the recognition models in the recognition model group to sequentially identify the information to be processed according to their respective execution orders; in response to detecting that the recognition model with a higher execution order in the recognition model group obtains the intention recognition result of the information to be processed, multiply the intention recognition result by the confidence weight of the recognition model to obtain the intention recognition result of the user interaction information.

[0104] In some optional implementations of the present embodiment, the result obtaining unit 404 is further configured to: in response to detecting that a recognition model with a later execution order in the recognition model group obtains an intention recognition result, wait for a recognition model with an earlier execution order in the recognition model group until it is detected that the recognition model with an earlier execution order obtains an intention recognition result; multiply the intention recognition result of the recognition model with a later execution order by the confidence weight of the recognition model with a later execution order plus the intention recognition result of the recognition model with an earlier execution order multiplied by the confidence weight of the recognition model with an earlier execution order to obtain the intention recognition result of the user interaction information.

[0105] In some optional implementations of the present embodiment, the above-mentioned result obtaining unit 404 is configured to: adopt a recognition model group with a high priority in at least one recognition model group to identify the information to be processed; detect whether the recognition model group with a high priority outputs an intention recognition result; in response to detecting that the recognition model group with a high priority outputs an intention recognition result, use the intention recognition result output by the recognition model group with a high priority as the intention recognition result of the user interaction information.

[0106] In some optional implementations of the present disclosure, the above-mentioned result obtaining unit 404 is further configured to: in response to detecting that the recognition model group with a high priority does not output the intention recognition result, adopt a recognition model group with a low priority in at least one recognition model group to identify the information to be processed; detect whether the recognition model group with a low priority outputs the intention recognition result; in response to detecting that the recognition model group with a low priority outputs the intention recognition result, use the intention recognition result output by the recognition model group with a low priority as the intention recognition result of the user interaction information.

[0107] The information processing method provided by the embodiment of the present disclosure is as follows: first, the receiving unit 401 receives the user interaction information sent by the front end; secondly, the detecting unit 402 detects whether there is interaction structure sub-information in the user interaction information; thirdly, the information obtaining unit 403 obtains the information to be processed based on the user interaction information in response to detecting that there is no interaction structure sub-information in the user interaction information; finally, the result obtaining unit 404 uses at least one recognition model group to identify the information to be processed and obtains the intention recognition result of the user interaction information. Therefore, when it is recognized that there is no interaction structure sub-information in the user interaction information, the information to be processed after the user interaction information is processed is identified by at least one recognition model group, so that the user's intention can be accurately identified, thereby improving the accuracy and efficiency of user information recognition.

[0108] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0109] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0110] Figure 5 A schematic block diagram of an example electronic device 500 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0111] like Figure 5 As shown, the device 500 includes a computing unit 501, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 502 or a computer program loaded from a storage unit 508 into a random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the device 500 can also be stored. The computing unit 501, the ROM 502, and the RAM 503 are connected to each other via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0112] A number of components in the device 500 are connected to the I / O interface 505, including: an input unit 506, such as a keyboard, a mouse, etc.; an output unit 507, such as various types of displays, speakers, etc.; a storage unit 508, such as a disk, an optical disk, etc.; and a communication unit 509, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 509 allows the device 500 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0113] The computing unit 501 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 501 performs the various methods and processes described above, such as information processing methods. For example, in some embodiments, the information processing method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 508. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 500 via ROM 502 and / or communication unit 509. When the computer program is loaded into RAM 503 and executed by the computing unit 501, one or more steps of the information processing method described above may be performed. Alternatively, in other embodiments, the computing unit 501 may be configured to perform the information processing method in any other appropriate manner (e.g., by means of firmware).

[0114] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0115] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable information processing device, so that the program code, when executed by the processor or controller, implements the functions / operations specified in the flow chart and / or block diagram. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0116] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0117] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0118] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an information server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communications network). Examples of communications networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0119] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0120] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0121] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. An information processing method, the method comprising: Receive user interaction information sent by the front end; Detecting whether the user interaction information contains interaction structure sub-information; In response to detecting that the user interaction information does not contain the interaction structure sub-information, obtaining information to be processed based on the user interaction information; At least one recognition model group is used to recognize the information to be processed, and an intention recognition result of the user interaction information is obtained.

2. The method according to claim 1, further comprising: In response to detecting that the user interaction information contains interaction structure sub-information, using a structured model to identify the interaction structure sub-information; Detecting whether the structured model outputs an intention recognition result; In response to detecting that the structured model outputs an intent recognition result, the intent recognition result output by the structured model is used as the intent recognition result of the user interaction information.

3. The method according to claim 2, further comprising: In response to detecting that the structured model does not output an intention recognition result, obtaining information to be processed based on the user interaction information; At least one recognition model group is used to recognize the information to be processed, and an intention recognition result of the user interaction information is obtained.

4. The method according to any one of claims 1 to 3, wherein: The obtaining of information to be processed based on the user interaction information includes: Based on the user interaction information, obtaining text type data; The text type data is preprocessed to obtain information to be processed.

5. The method according to claim 4, wherein: The preprocessing of the text type data to obtain the information to be processed includes: removing ambiguous data in the text type data to obtain first text data; Performing content normalization processing on the first text data to obtain second text data; The sentence structure of the second text data is changed to obtain information to be processed.

6. The method according to any one of claims 1 to 3, wherein: The step of using at least one recognition model group to recognize the information to be processed and obtaining an intention recognition result of the user interaction information includes: Based on the information to be processed, dynamically determine at least one recognition model group and a priority of each recognition model group from the recognition model set; Control each recognition model group to recognize the information to be processed according to their respective priorities, and obtain the intention recognition result of the user interaction information.

7. The method according to claim 6, wherein: The controlling each recognition model group to recognize the information to be processed according to their respective priorities to obtain the intention recognition result of the user interaction information includes: Determining the execution order and confidence weight of the recognition models in each recognition model group; For the currently executed recognition model group, control the recognition models in the recognition model group to sequentially recognize the information to be processed according to their respective execution order; In response to detecting that a recognition model with a higher execution order in the recognition model group obtains an intention recognition result of the information to be processed, the intention recognition result is multiplied by the confidence weight of the recognition model to obtain the intention recognition result of the user interaction information.

8. The method according to claim 7, wherein: The controlling each recognition model group to recognize the information to be processed according to their respective priorities to obtain the intention recognition result of the user interaction information also includes: In response to detecting that a recognition model with a later execution order in the recognition model group obtains an intention recognition result, waiting for a recognition model with an earlier execution order in the recognition model group until detecting that the recognition model with an earlier execution order obtains an intention recognition result; The intention recognition result of the recognition model with a later execution order is multiplied by the confidence weight of the recognition model with a later execution order, and the intention recognition result of the recognition model with an earlier execution order is multiplied by the confidence weight of the recognition model with an earlier execution order to obtain the intention recognition result of the user interaction information.

9. The method according to claim 6, wherein: The controlling the recognition model groups in the at least one recognition model group to recognize the information to be processed according to their respective priorities to obtain the intention recognition result of the user interaction information comprises: Using a recognition model group with a high priority among at least one recognition model group to recognize the information to be processed; Detecting whether the recognition model group with a high priority outputs an intention recognition result; In response to detecting that the recognition model group with a high priority outputs an intention recognition result, the intention recognition result output by the recognition model group with a high priority is used as the intention recognition result of the user interaction information.

10. The method according to claim 9, wherein: The controlling the recognition model groups in the at least one recognition model group to recognize the information to be processed according to their respective priorities to obtain the intention recognition result of the user interaction information further comprises: In response to detecting that the recognition model group with a high priority does not output an intention recognition result, using a recognition model group with a low priority among at least one recognition model group to recognize the information to be processed; Detecting whether the recognition model group with a low priority outputs an intention recognition result; In response to detecting that the recognition model group with a low priority outputs an intention recognition result, the intention recognition result output by the recognition model group with a low priority is used as the intention recognition result of the user interaction information.

11. An information processing device, the device comprising: A receiving unit, configured to receive user interaction information sent by the front end; a detection unit, configured to detect whether the user interaction information contains interaction structure sub-information; an information obtaining unit, configured to obtain information to be processed based on the user interaction information in response to detecting that the user interaction information does not contain the interaction structure sub-information; The result obtaining unit is configured to use at least one recognition model group to identify the information to be processed and obtain the intention recognition result of the user interaction information.

12. The device according to claim 11, further comprising: an identification unit, configured to, in response to detecting that the user interaction information contains interaction structure sub-information, identify the interaction structure sub-information using a structured model; Detecting whether the structured model outputs an intention recognition result; In response to detecting that the structured model outputs an intent recognition result, the intent recognition result output by the structured model is used as the intent recognition result of the user interaction information.

13. The device according to claim 12, further comprising: an intention obtaining unit, configured to obtain information to be processed based on the user interaction information in response to detecting that the structured model does not output an intention recognition result; At least one recognition model group is used to recognize the information to be processed, and an intention recognition result of the user interaction information is obtained.

14. The device according to any one of claims 11 to 13, wherein: The information obtaining unit is configured to: obtain text type data based on the user interaction information; and perform normalization processing on the text type data to obtain information to be processed.

15. The device according to claim 14, wherein: The information obtaining unit is further configured to: remove ambiguous data in the text type data to obtain first text data; perform content normalization processing on the first text data to obtain second text data; and change the sentence structure of the second text data to obtain information to be processed.

16. The device according to any one of claims 11 to 13, wherein: The result obtaining unit is configured to: dynamically determine at least one recognition model group and the priority of each recognition model group from the recognition model set based on the information to be processed; control each recognition model group to recognize the information to be processed according to its respective priority, and obtain the intention recognition result of the user interaction information.

17. The device according to claim 16, wherein: The result obtaining unit is further configured to: determine the execution order and confidence weight of the recognition models in each recognition model group; for the currently executed recognition model group, control the recognition models in the recognition model group to sequentially recognize the information to be processed according to their respective execution orders; In response to detecting that a recognition model with a higher execution order in the recognition model group obtains an intention recognition result of the information to be processed, the intention recognition result is multiplied by the confidence weight of the recognition model to obtain the intention recognition result of the user interaction information.

18. The device according to claim 17, wherein: The result obtaining unit is further configured to: in response to detecting that a recognition model with a later execution order in the recognition model group obtains an intention recognition result, wait for a recognition model with an earlier execution order in the recognition model group until it is detected that the recognition model with an earlier execution order obtains an intention recognition result; multiply the intention recognition result of the recognition model with a later execution order by the confidence weight of the recognition model with a later execution order plus the intention recognition result of the recognition model with an earlier execution order multiplied by the confidence weight of the recognition model with an earlier execution order, to obtain the intention recognition result of the user interaction information.

19. The device according to claim 16, wherein: The result obtaining unit is configured to: use a recognition model group with a high priority among at least one recognition model group to recognize the information to be processed; Detecting whether the recognition model group with a high priority outputs an intention recognition result; In response to detecting that the recognition model group with a high priority outputs an intention recognition result, the intention recognition result output by the recognition model group with a high priority is used as the intention recognition result of the user interaction information.

20. The device according to claim 19, wherein The result obtaining unit is further configured to: in response to detecting that the recognition model group with a high priority does not output an intention recognition result, use a recognition model group with a low priority among at least one recognition model group to recognize the information to be processed; Detecting whether the recognition model group with a low priority outputs an intention recognition result; In response to detecting that the recognition model group with a low priority outputs an intention recognition result, the intention recognition result output by the recognition model group with a low priority is used as the intention recognition result of the user interaction information.

21. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 10.

22. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to enable the computer to execute the method according to any one of claims 1 to 10.

23. A computer program product, comprising a computer program, which, when executed by a processor, implements the method of any one of claims 1 to 10.