Intention recognition model training method for retrieval, retrieval method and electronic equipment

Abstract

The embodiment of the invention relates to an intention recognition model training method for retrieval, a retrieval method and electronic device.In at least one embodiment of the invention, first information for retrieval is obtained, and the first information can be house finding appeal expression of a user; performing intention recognition on the first information on the basis of an intention recognition model obtained through training of an intention recognition model training method to obtain at least one retrieval condition character string so as to associate the house finding appeal expression of the user with the house source retrieval condition; therefore, retrieval is directly performed based on the at least one retrieval condition character string, the user does not need to input retrieval content for multiple times, the retrieval path is shortened, the house finding habit of the user can be met, and the house finding appeal of the user is fully expressed.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and in particular to an intent recognition model training method, retrieval method, and electronic device for retrieval. Background Technology

[0002] Currently, the method for searching properties involves users entering the name of a community in the search box on the homepage, and then further filtering based on more specific criteria. The backend will then search and match properties according to these criteria. The entire search process is lengthy, making it inconvenient for new users or the elderly, and it does not align with users' actual home-hunting habits, thus limiting the expression of users' home-hunting needs. Summary of the Invention

[0003] To address, or at least partially address, the aforementioned technical problems, this disclosure provides a method for training an intent recognition model for retrieval, a retrieval method, and an electronic device.

[0004] This disclosure provides a method for training an intent recognition model for retrieval, including:

[0005] An intent recognition model is obtained by jointly training a text generation model and a similarity calculation model.

[0006] The text generation model is used to generate text samples based on at least one first search term;

[0007] The similarity calculation model is used to calculate the similarity between the at least one first search term and at least one second search term, wherein the at least one second search term is a search term obtained by the intent recognition model from the intent recognition of the text sample.

[0008] In some embodiments, the intent recognition model obtained by jointly training the text generation model and the similarity calculation model includes:

[0009] Based on the text generation model, at least one first search term is processed to generate a text sample;

[0010] Based on the intent recognition model, the intent of the text sample is recognized to obtain at least one second search term;

[0011] The similarity between the at least one first search term and the at least one second search term is calculated based on the similarity calculation model.

[0012] The parameters of the intent recognition model are optimized based on the similarity until the similarity converges to 1, thus obtaining the optimized intent recognition model.

[0013] In some embodiments, the step of processing at least one first search term based on the text generation model to generate text samples includes:

[0014] The text generation model obtains at least one search condition string as the at least one first search term through a knowledge base, and obtains the natural language text corresponding to the at least one first search term; wherein, the knowledge base stores multiple correspondences between the search condition strings and natural language texts;

[0015] The text generation model generates a text sample of natural language expression based on the natural language text corresponding to at least one first search term.

[0016] In some embodiments, the step of performing intent recognition on the text sample based on the intent recognition model to obtain at least one second search term includes:

[0017] The intent recognition model performs intent recognition on the text sample to obtain at least one intent recognition result text.

[0018] Search the knowledge base for at least one of the search condition strings corresponding to the at least one intent recognition result text, and use the search condition strings as the second search terms to obtain at least one second search term.

[0019] This disclosure also provides a retrieval method, including:

[0020] Obtain the first information used for retrieval;

[0021] The intent recognition model trained based on the intent recognition model training method for retrieval performs intent recognition on the first information to obtain at least one retrieval condition string.

[0022] The search is performed based on at least one search condition string.

[0023] In some embodiments, the first information is text information or voice information.

[0024] In some embodiments, the step of performing intent recognition on the first information to obtain at least one search condition string includes:

[0025] If the first information is text information, then the first information is processed based on the intent recognition model to obtain at least one search term; and at least one search condition string is determined based on the at least one search term.

[0026] In some embodiments, the step of performing intent recognition on the first information to obtain at least one search condition string includes:

[0027] If the first information is voice information, then the first information is subjected to voice recognition to obtain the second information, which is the text information corresponding to the first information; the second information is processed based on the intent recognition model to obtain at least one search term; at least one search condition string is determined based on the at least one search term.

[0028] This disclosure also provides an electronic device, the electronic device comprising:

[0029] A storage device on which computer programs are stored;

[0030] A processing device is configured to execute the computer program in the storage device to implement the steps of the intent recognition model training method for retrieval provided in the embodiments of the present disclosure or to implement the steps of the retrieval method provided in the embodiments of the present disclosure.

[0031] This disclosure also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the intent recognition model training method for retrieval provided in this disclosure, or implements the retrieval method provided in this disclosure.

[0032] The technical solution provided in this disclosure obtains first information for retrieval, which may be a user's expression of their housing search needs; then, based on the intent recognition model trained by the intent recognition model training method, the intent recognition of the first information can be performed to obtain at least one search condition string, thereby associating the user's expression of their housing search needs with housing search conditions; and then, based on at least one search condition string, a retrieval can be performed directly without the user having to input search content multiple times, shortening the search path and conforming to the user's housing search habits, so that the user's housing search needs are fully expressed.

[0033] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0034] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0035] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 A schematic flowchart illustrating a housing search method provided in this embodiment of the disclosure;

[0037] Figure 2 A schematic flowchart illustrating a housing search method provided in this embodiment of the disclosure;

[0038] Figure 3 A flowchart illustrating an intent recognition model training method for retrieval provided in an embodiment of this disclosure;

[0039] Figure 4 This is a schematic diagram of the structure of a housing search device provided in an embodiment of the present disclosure;

[0040] Figure 5 A schematic structural diagram of an electronic device provided in an embodiment of the present disclosure. Detailed Implementation

[0041] To better understand the above-mentioned objectives, features, and advantages of this disclosure, the solutions disclosed herein will be further described below. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0042] Numerous specific details are set forth in the following description in order to provide a full understanding of this disclosure, but this disclosure may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only some, and not all, of the embodiments of this disclosure.

[0043] Currently, users typically search for properties based on location (city center, commercial area, near subway station, etc.), budget, property type, size, whether it's a five-year-old property and the only property owned by the seller, whether it has an elevator, and whether it's on a middle floor. However, the current property search method requires users to enter their search terms multiple times, resulting in a lengthy search process that doesn't align with users' search habits and limits the expression of their search needs.

[0044] Therefore, in order to shorten the search path, conform to users' house-hunting habits, and fully express users' house-hunting needs, at least one embodiment of this disclosure discloses an intent recognition model training method, a search method, or an electronic device for retrieval. By acquiring first information for retrieval, which may be the user's house-hunting needs; and then, based on the intent recognition model trained by the intent recognition model training method, the intent of the first information can be recognized to obtain at least one search condition string, thereby associating the user's house-hunting needs with the house search conditions; and then, based on at least one search condition string, the retrieval can be performed directly without the user having to input search content multiple times, thus shortening the search path and conforming to users' house-hunting habits, allowing users' house-hunting needs to be fully expressed.

[0045] To improve the accuracy of intent recognition, embodiments of this disclosure provide a method for training an intent recognition model for retrieval, the method comprising:

[0046] An intent recognition model is obtained by jointly training a text generation model and a similarity calculation model.

[0047] The text generation model generates text samples based on at least one first search term. The similarity calculation model calculates the similarity between at least one first search term and at least one second search term, where the at least one second search term is a search term obtained by the intent recognition model from the text sample through intent recognition.

[0048] For example, intent recognition models are obtained by jointly training a text generation model and a similarity calculation model, including the following (A) to (D):

[0049] (A) Process at least one first search term based on a text generation model to generate a text sample.

[0050] The text generation model can generate text samples of a user's housing search request based on at least one primary search term. For example, based on the primary search term "Dongcheng District," the text generation model can generate the text sample: "I want to find a house in Dongcheng District."

[0051] (B) Based on the intent recognition model, the intent of the text sample is recognized to obtain at least one second search term.

[0052] For example, the intent recognition model performs intent recognition on the text sample: "I want to find a house in Dongcheng District", and obtains the second search term: "Dongcheng District".

[0053] (C) Calculate the similarity between at least one first search term and at least one second search term based on the similarity calculation model.

[0054] The similarity calculation model can calculate the similarity between two texts. For example, it can calculate the similarity between a first search condition consisting of at least one first search term and a second search condition consisting of at least one second search term. The higher the similarity, the more accurate the intent recognition.

[0055] For example, the similarity calculation model calculates that the first search term (Dongcheng District) and the second search term (Dongcheng District) have a similarity of 1, indicating that the intent recognition is accurate.

[0056] (D) Optimize the parameters of the intent recognition model based on similarity until the similarity converges to 1, and obtain the optimized intent recognition model.

[0057] The above A to D can be understood in the context of question setting, examination, and scoring. Question setting refers to the production of test papers and questions based on standard answers, which is the expression of users' housing search needs completed by the text generation model. A standard answer (the answers to various test questions are actually understood as various search conditions) can generate multiple test papers by combination. The intent recognition model is the question-taking (intent recognition). Finally, the similarity calculation model is the scoring (judging the similarity between the result after intent recognition and the standard answer). The higher the score (the higher the similarity), the better the learning (the better the intent recognition effect).

[0058] As can be seen, the text generation model can continuously (e.g., periodically) generate text samples based on at least one first search term; the intent recognition model completes intent recognition to obtain at least one second search term; finally, the similarity calculation model completes similarity calculation, and the parameters of the intent recognition model can be optimized based on the similarity. This allows for a more comprehensive simulation of the housing search needs of different users, and enables the evaluation of the intent recognition effect.

[0059] In some embodiments, in (A) above, at least one first search term is processed based on a text generation model to generate text samples, including the following (A1) and (A2):

[0060] (A1) The text generation model obtains at least one search condition string as at least one first search term from the knowledge base, and obtains the natural language text corresponding to at least one first search term.

[0061] The knowledge base stores multiple mappings between search condition strings and natural language text. For example, the search condition string "d23008614" corresponds to the natural language text "Dongcheng District".

[0062] In this embodiment, the text generation model can automatically obtain the search condition strings and their corresponding natural speech text from the knowledge base, eliminating the need for manual generation of search terms and improving model training efficiency.

[0063] (A2) The text generation model generates a text sample of natural language expression based on the natural language text corresponding to at least one first search term.

[0064] For example, the text generation model generates a text sample expressing the following natural language expression based on the natural language text corresponding to the first search term (e.g., d23008614, i.e., Dongcheng District): I want to find a house in Dongcheng District.

[0065] In some embodiments, in (B) above, intent recognition is performed on the text sample based on the intent recognition model to obtain at least one second search term, including (B1) and (B2):

[0066] (B1) The intent recognition model performs intent recognition on the text sample and obtains at least one intent recognition result text.

[0067] For example, when an intent recognition model performs intent recognition on a text sample: "I want to find a house in Dongcheng District", it gets three intent recognition results: "I want to find", "Dongcheng District", and "house".

[0068] (B2) Search for at least one search condition string corresponding to at least one intent recognition result text in the knowledge base, and use the search condition string as the second search term to obtain at least one second search term.

[0069] For example, the intent recognition model searches the knowledge base for the search condition strings corresponding to "I want to find", "Dongcheng District", and "house". After searching, it only finds the search condition string corresponding to "Dongcheng District": d23008614. Therefore, d23008614 is used as the second search term.

[0070] Figure 1 This is a flowchart illustrating a retrieval method provided in an embodiment of the present disclosure. This method can be executed by a retrieval device, which can be implemented using software and / or hardware, and is generally integrated into an electronic device. For example... Figure 1 As shown, the method includes, but is not limited to, steps 101 to 103:

[0071] In step 101, the first information used for retrieval is obtained.

[0072] The first piece of information can be the user's expression of their housing search needs. For example, the first piece of information can be a natural language input expression of their housing search needs (i.e., text information), or the first piece of information can be a user's voice input expression of their housing search needs (i.e., voice information).

[0073] Users can describe their housing needs in one go according to their own housing search habits, without having to enter search content multiple times, which shortens the search path and lowers the threshold for new users or the elderly to find housing.

[0074] In step 102, the intent recognition model trained is used to perform intent recognition on the first information to obtain at least one retrieval condition string.

[0075] In this embodiment, the intent recognition model trained based on the intent recognition model training method for retrieval provided in the foregoing embodiment is used to perform intent recognition on the first information, which can obtain at least one search condition string. This enables the association of the user's house-finding request with the house search conditions, eliminating the need for the user to input search content multiple times to determine the house search conditions and improving search efficiency.

[0076] In step 103, a search is performed based on at least one search condition string.

[0077] In this embodiment, housing search is performed directly based on at least one search condition string. For users, they only need to input their housing search requirements (i.e., the first information in step 101) to obtain housing search results. Users do not need to input search content multiple times, which shortens the search path and conforms to users' housing search habits, allowing users to fully express their housing search requirements.

[0078] In some embodiments, in step 102, intent recognition is performed on the first information to obtain at least one search condition string, including the following methods (1) or (2):

[0079] (1) If the first information is text information, the first information is processed based on the intent recognition model to obtain at least one search term; at least one search condition string is determined based on the at least one search term.

[0080] (2) If the first information is speech information, then the first information is speech recognized to obtain the second information, which is the text information corresponding to the first information; the second information is processed based on the intent recognition model to obtain at least one search term; at least one search condition string is determined based on at least one search term.

[0081] The difference from (1) is that in (2), the voice information is first converted into text information, and then the same process as (1) is performed.

[0082] Figure 2 This is a flowchart illustrating a housing search method provided in an embodiment of the present disclosure. Figure 2 In China, the process of searching for properties includes the following steps 1 to 5:

[0083] 1. Users can input natural language through their client devices.

[0084] In this example, prompts can be displayed to users to guide them in entering their housing search needs.

[0085] 2. The user's natural language can be processed through speech recognition to obtain the request text.

[0086] Those skilled in the art will understand that the natural language input by the user may not be in the form of speech. For example, it can be directly input as text, in which case speech recognition is not required, and the natural language input by the user can be used as the request text.

[0087] 3. The intent recognition model can acquire the request text in real time and perform intent recognition on the request text based on prompt words and knowledge base, and output at least one search term.

[0088] Among these, prompt words help the intent recognition model better understand user input and improve the accuracy of intent recognition. The knowledge base of the intent recognition model is built based on property attributes and property tags. The knowledge base includes at least one search term (i.e., ToC (Transaction Oriented to Consumer) expression) used to construct search conditions and the search service expression of the search term (i.e., search condition string). A search term can be a property attribute or a property tag.

[0089] For example, the knowledge base of the intent recognition model includes ToC expressions and retrieval service expressions for different retrieval conditions. For example, the ToC expression is Dongcheng District, and the retrieval service expression is d23008614.

[0090] 4. The search recall layer directly searches for properties based on at least one property search condition determined by at least one search term, and recalls at least one property information.

[0091] 5. The business layer then returns at least one property listing to the user. The business layer is used for ToC (ToC) logic control.

[0092] Figure 3 This is a flowchart illustrating a method for training an intent recognition model for retrieval, provided in an embodiment of this disclosure. The intent recognition model, text generation model, and similarity calculation model are all artificial intelligence (AI) models.

[0093] In this embodiment, the intent recognition model is obtained by jointly training the text generation model and the similarity calculation model.

[0094] The text generation model can generate a text sample reflecting a user's housing search request based on at least one primary search term. For example, based on the primary search term "Dongcheng District," the model can generate the text sample: "I want to find a house in Dongcheng District."

[0095] Intent recognition models can identify the intent of text samples and obtain at least one second search term. For example, for the text sample "I want to find houses in Dongcheng District," the intent recognition model can identify the intent and obtain the second search term: "Dongcheng District."

[0096] Similarity calculation models can calculate the similarity between two texts. For example, a similarity calculation model can calculate the similarity between a first search condition consisting of at least one first search term and a second search condition consisting of at least one second search term. The higher the similarity, the more accurate the intent recognition.

[0097] For example, the similarity calculation model calculates that the first search term (Dongcheng District) and the second search term (Dongcheng District) have a similarity of 1, indicating that the intent recognition is accurate.

[0098] At this point, there is no need to optimize the parameters of the intent recognition model. Different text samples are generated again by the text generation model, and the intent recognition model performs intent recognition on the different text samples. The similarity calculation model calculates the similarity again, and then optimizes the parameters of the intent recognition model based on the similarity until the similarity converges to 1, thus obtaining the optimized intent recognition model.

[0099] It can be seen that, in Figure 3 In this process, the text generation model can continuously (e.g., periodically) generate text samples and at least one corresponding first search term; the intent recognition model performs intent recognition to obtain at least one second search term; finally, the similarity calculation model performs similarity calculation, and the parameters of the intent recognition model can be optimized based on the similarity. This allows for a more comprehensive simulation of the housing search needs of different users, and enables the evaluation of the intent recognition effect.

[0100] exist Figure 3 Based on this, this embodiment illustrates the training method for the intent recognition model with an example:

[0101] Intent Recognition Model: This model recognizes user intent by identifying their housing search requests and converting them into specific search criteria. Through its capabilities and the creation of prompts, it loads a knowledge base containing relevant Chinese housing search expressions and corresponding search criteria. The model can then perform recognition and matching from Chinese text to search terms based on these prompts. An example of the knowledge base content is shown below:

[0102] Dongcheng District: d23008614;

[0103] Xicheng District: d23008626;

[0104] 1-bedroom apartment: l1;

[0105] 2-bedroom: l2;

[0106] 3-bedroom: l3.

[0107] For example: If a user inputs "I want to find a 3-bedroom apartment in Xicheng District", the intent recognition output will be "d23008626l3".

[0108] Text generation model: This model generates Chinese expressions of housing search needs, i.e., content automatically generated by artificial intelligence (AIGC). Based on the text generation model, prompt words, and a knowledge base, it arranges and combines different search criteria, and then generates numerous different expressions of housing search needs guided by prompt words. Examples of knowledge base content are as follows:

[0109] li652s20578: Xizhimen Station;

[0110] li652s43143232: Dazhongsi Station;

[0111] li652s20628: Zhichun Road Station.

[0112] For example, the text generation model generates: "I want to find a 3-bedroom apartment near Xizhimen subway station" or "I want to find a 3-bedroom or 2-bedroom apartment near Xizhimen subway station".

[0113] Similarity calculation model: Receive text input from the intent recognition model and text generation model, convert the text into vectors using the similarity calculation model and prompt words, and then calculate the similarity using the cosine similarity algorithm. The higher the value, the higher the accuracy of intent recognition.

[0114] For example, the intent recognition result is:

[0115] li120027908698063s1120027970561993.

[0116] The standard result (i.e., the result of the text generation model) is:

[0117] li120027908698063s1120027970561993l2p1a2su1.

[0118] The similarity calculation model yielded a similarity of 80%, with the difference being the absence of l2p1a2su1. Therefore, identifying these differences allowed for optimization of the intent recognition model.

[0119] Corresponding to the aforementioned intent recognition model training method for retrieval, this disclosure further provides an intent recognition model training device for retrieval, which can be implemented by software and / or hardware and is generally integrated into an electronic device.

[0120] The device includes a joint training module for: jointly training an intent recognition model based on a text generation model and a similarity calculation model.

[0121] The text generation model is used to generate text samples based on at least one first search term; the similarity calculation model is used to calculate the similarity between at least one first search term and at least one second search term, wherein at least one second search term is a search term obtained by the intent recognition model from the text sample through intent recognition.

[0122] In some embodiments, the joint training module is used for:

[0123] The text generation model processes at least one first search term to generate a text sample.

[0124] The intent is identified by the intent recognition model to obtain at least one second search term from the text sample.

[0125] The similarity between at least one first search term and at least one second search term is calculated based on a similarity calculation model.

[0126] The parameters of the intent recognition model are optimized based on similarity until the similarity converges to 1, resulting in the optimized intent recognition model.

[0127] In some embodiments, the joint training module processes at least one first search term based on a text generation model to generate text samples, including:

[0128] The text generation model obtains at least one search condition string as at least one first search term from a knowledge base, and obtains the natural language text corresponding to at least one first search term; wherein, the knowledge base stores the correspondence between multiple search condition strings and natural language texts;

[0129] The text generation model generates a text sample of a natural language expression based on the natural language text corresponding to at least one first search term.

[0130] In some embodiments, the joint training module performs intent recognition on text samples based on the intent recognition model to obtain at least one second search term, including:

[0131] The intent recognition model performs intent recognition on text samples and obtains at least one intent recognition result text.

[0132] Search the knowledge base for at least one search condition string corresponding to at least one intent recognition result text, and use the search condition string as the second search term to obtain at least one second search term.

[0133] Corresponding to the aforementioned retrieval method, this disclosure further provides a retrieval device. Figure 4 This is a schematic diagram of a retrieval device provided in an embodiment of the present disclosure. The device can be implemented by software and / or hardware, and is generally integrated into an electronic device, such as... Figure 4As shown, the retrieval device includes: an acquisition unit 41, an identification unit 42, and a retrieval unit 43. Detailed descriptions are as follows:

[0134] Acquisition unit 41 is used to acquire first information for retrieval;

[0135] The recognition unit 42 is used to perform intent recognition on the first information based on the trained intent recognition model to obtain at least one retrieval condition string;

[0136] The retrieval unit 43 is used to perform a retrieval based on at least one retrieval condition string.

[0137] In some embodiments, the first information is text information or voice information.

[0138] In some embodiments, the identification unit 42 is used for:

[0139] If the first information is text information, then the first information is processed based on the intent recognition model to obtain at least one search term; and at least one search condition string is determined based on the at least one search term.

[0140] In some embodiments, the identification unit 42 is used for:

[0141] If the first information is speech information, then speech recognition is performed on the first information to obtain the second information, which is the text information corresponding to the first information; the second information is processed based on the intent recognition model to obtain at least one search term; at least one search condition string is determined based on the at least one search term.

[0142] The housing search device provided in this disclosure can execute the housing search method provided in any embodiment of this disclosure, and has the corresponding functional modules and beneficial effects of executing the method.

[0143] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described device embodiments can be referred to the corresponding process in the method embodiments, and will not be repeated here.

[0144] This disclosure provides an electronic device, which includes: a storage device storing a computer program thereon; and a processing device for executing the computer program in the storage device to implement the steps of any method of this disclosure.

[0145] The following is for reference. Figure 5The diagram illustrates a structural schematic of an electronic device 500 suitable for implementing embodiments of the present disclosure. The terminal devices in the embodiments of the present disclosure may include, but are not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 5 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0146] like Figure 5 As shown, the electronic device 500 may include a processing unit (e.g., a central processing unit, a graphics processor, etc.) 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage device 508 into a random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of the electronic device 500. The processing unit 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0147] Typically, the following devices can be connected to I / O interface 505: input devices 506 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 507 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 508 including, for example, magnetic tapes, hard disks, etc.; and communication devices 509. Communication device 509 allows electronic device 500 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 5 An electronic device 500 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively.

[0148] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a storage device 508, or installed from a ROM 502. When the computer program is executed by the processing device 501, it performs the functions defined in the methods of the embodiments of this disclosure.

[0149] In addition to the methods and devices described above, embodiments of this disclosure can also be computer program products, comprising computer program instructions that, when executed by a processor, cause the processor to perform the methods provided in the embodiments of this disclosure. The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this disclosure. These programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on a user computing device, partially on a user device, as a standalone software package, partially on a user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0150] Furthermore, embodiments of this disclosure may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the housing search method provided in embodiments of this disclosure.

[0151] The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0152] This disclosure also provides a computer program product, including a computer program / instructions, which, when executed by a processor, implements the property search method in this disclosure.

[0153] It is understood that before using the technical solutions disclosed in the various embodiments of this disclosure, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this disclosure in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0154] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the software or hardware, such as the electronic device, application, server, or storage medium performing the operations of this disclosed technical solution, based on the prompt message.

[0155] As an optional but non-limiting implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0156] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this disclosure.

[0157] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0158] The above description is merely a specific embodiment of this disclosure, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for training an intent recognition model for retrieval, characterized in that, include: An intent recognition model is obtained by jointly training a text generation model and a similarity calculation model. The text generation model is used to generate text samples based on at least one first search term; The similarity calculation model is used to calculate the similarity between the at least one first search term and at least one second search term, wherein the at least one second search term is a search term obtained by the intent recognition model from the intent recognition of the text sample.

2. The method according to claim 1, characterized in that, The intent recognition model, jointly trained based on the text generation model and the similarity calculation model, includes: Based on the text generation model, at least one first search term is processed to generate a text sample; Based on the intent recognition model, the intent of the text sample is recognized to obtain at least one second search term; The similarity between the at least one first search term and the at least one second search term is calculated based on the similarity calculation model. The parameters of the intent recognition model are optimized based on the similarity until the similarity converges to 1, thus obtaining the optimized intent recognition model.

3. The method according to claim 2, characterized in that, The process of processing at least one first search term based on the text generation model to generate text samples includes: The text generation model obtains at least one search condition string as the at least one first search term through a knowledge base, and obtains the natural language text corresponding to the at least one first search term; wherein, the knowledge base stores multiple correspondences between the search condition strings and natural language texts; The text generation model generates a text sample of natural language expression based on the natural language text corresponding to at least one first search term.

4. The method according to claim 3, characterized in that, The process of performing intent recognition on the text sample based on the intent recognition model to obtain at least one second search term includes: The intent recognition model performs intent recognition on the text sample to obtain at least one intent recognition result text. Search the knowledge base for at least one of the search condition strings corresponding to the at least one intent recognition result text, and use the search condition strings as the second search terms to obtain at least one second search term.

5. A retrieval method, characterized in that, include: Obtain the first information used for retrieval; The intent recognition model trained based on the method according to any one of claims 1 to 4 performs intent recognition on the first information to obtain at least one search condition string; The search is performed based on at least one search condition string.

6. The method according to claim 5, characterized in that, The first information is either text information or voice information.

7. The method according to claim 6, characterized in that, The step of performing intent recognition on the first information to obtain at least one search condition string includes: If the first information is text information, then the first information is processed based on the intent recognition model to obtain at least one search term; and at least one search condition string is determined based on the at least one search term.

8. The method according to claim 6, characterized in that, The step of performing intent recognition on the first information to obtain at least one search condition string includes: If the first information is voice information, then the first information is subjected to voice recognition to obtain the second information, which is the text information corresponding to the first information; the second information is processed based on the intent recognition model to obtain at least one search term; at least one search condition string is determined based on the at least one search term.

9. An electronic device, characterized in that, The electronic device includes: A storage device on which computer programs are stored; A processing device for executing the computer program in the storage device to implement the steps of the intent recognition model training method for retrieval as described in any one of claims 1-4 or the steps of the retrieval method as described in any one of claims 5-8.

10. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the steps of the intent recognition model training method for retrieval as described in any one of claims 1-4, or implements the retrieval method as described in any one of claims 5-8.

Images