Method, computing device, and storage medium for business recommendation based on identified user intent

Abstract

Embodiments of the present application relate to a method, a computing device and a storage medium for business recommendation based on an identified user intent. The method comprises, for current input information of a user, performing scalar retrieval and vector retrieval in a keyword database respectively; performing ranking based on the scalar retrieval result and the vector retrieval result to obtain a preliminary intent tendency ranking of the user; inputting at least the current input information of the user and the preliminary intent tendency ranking of the user into an intent recognition model to determine a current intent of the user and a business type corresponding to the current intent via the intent recognition model; the intent recognition model is constructed based on a neural network model; and based on the business type corresponding to the current intent of the user, calling a recommendation tool corresponding to the business type to make a business recommendation for the current intent of the user. Thus, the stability and accuracy of effective user intent recognition can be achieved to realize accurate and efficient identification of user intent.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of user intention perception, and more particularly to a method, a computing device and a storage medium for business recommendation based on an identified user intention. BACKGROUND

[0002] Conventional methods for business recommendation based on an identified user intention, such as directly identifying the intention of user input information through a large model (LLM, LFM), or identifying the intention of user input information through a pre-set rule engine, have the following disadvantages: due to the divergence of the large model itself, it is difficult to stably identify the real intention of the user; and the rule engine does not support semantic generalization, making it difficult to accurately identify the intention of the user's fuzzy input.

[0003] In summary, the conventional method for business recommendation based on an identified user intention has the disadvantage that it cannot stably and accurately identify the real intention of the user. SUMMARY

[0004] To solve the above problems, the present application provides a method, a computing device and a storage medium for business recommendation based on an identified user intention, which can effectively improve the stability and accuracy of user intention identification, and achieve accurate and efficient identification of user intention.

[0005] According to a first aspect of the present application, a method for business recommendation based on an identified user intention includes: performing scalar retrieval and vector retrieval in a keyword database for current input information of a user; sorting based on the scalar retrieval result and the vector retrieval result to obtain a preliminary intention tendency ranking of the user; inputting at least the current input information of the user and the preliminary intention tendency ranking of the user into an intention recognition model to determine the current intention of the user and the business type corresponding to the current intention via the intention recognition model; the intention recognition model is constructed based on a neural network model; and based on the business type corresponding to the current intention of the user, calling a recommendation tool corresponding to the business type to make business recommendation for the current intention of the user.

[0006] According to a second aspect of the present application, a computing device is provided, which includes: at least one processing unit; at least one memory coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions, when executed by the at least one processing unit, enabling the computing device to perform the steps of the method according to the first aspect.

[0007] According to a third aspect of the present application, there is also provided a computer readable storage medium. The computer readable storage medium has stored thereon a computer program which, when executed by a machine, performs the method of the first aspect of the present application.

[0008] According to a fourth aspect of the present application, there is also provided a computer program product comprising a computer program which, when executed by a machine, performs the method of the first aspect of the present application.

[0009] In some embodiments, the business types include: commodity recommendation, application recommendation, event recommendation, question inquiry, and / or chat.

[0010] In some embodiments, the keyword database includes: a scalar database constructed based on scalar data about commodity data, event data, application data, and / or question data corresponding to the business types; and a vector database constructed based on vectorization of the commodity data, the event data, the application data, and / or the question data corresponding to the business types.

[0011] In some embodiments, performing the scalar retrieval and the vector retrieval in the keyword database respectively includes: vectorizing the current input information of the user to obtain vectorized current input information; performing similarity retrieval in the vector database for the vectorized current input information; and identifying at least one business type and at least one preliminary intent tendency corresponding to the current input information of the user based on the level of semantic similarity.

[0012] In some embodiments, the sorting based on the scalar retrieval result and the vector retrieval result to obtain the preliminary intent tendency sorting of the user includes: obtaining multiple preliminary intent tendencies of the user based on the scalar retrieval result to respectively assign corresponding weights to the multiple preliminary intent tendencies; obtaining multiple preliminary intent tendencies of the user for which the semantic similarity is higher than a predetermined similarity threshold based on the vector retrieval result to respectively assign corresponding weights; and merging the preliminary intent tendencies of the user and the corresponding weights obtained based on the scalar retrieval result and the vector retrieval result respectively to sort the obtained preliminary intent tendencies of the user from high to low based on the weights.

[0013] In some embodiments, the merging of the preliminary intent tendencies of the user and the corresponding weights obtained based on the scalar retrieval result and the vector retrieval result includes: for the current input information of the user, in response to confirming that the scalar retrieval result and the vector retrieval result exist simultaneously for the same preliminary intent tendency, excluding the vector retrieval result for the same preliminary intent tendency.

[0014] In some embodiments, the inputting, into the intent recognition model, at least the current input information of the user and the preliminary intent tendency of the user to determine the current intent of the user and the business type corresponding to the current intent via the intent recognition model comprises: obtaining the current input information of the user, the historical input information of the user, the historical intent information of the user and the preliminary intent tendency of the user to input into the intent recognition model; identifying, via the intent recognition model, a predicted intent tendency of the user, the predicted intent tendency comprising at least one intent tendency belonging to a predetermined intent tendency set and / or at least one preliminary intent tendency; and sorting based on probabilities of the predicted intent tendency of the user to determine the current intent of the user and the business type corresponding to the current intent.

[0015] Thus, the present application can combine the advantages of both scalar and vector, construct a scalar database for scalar retrieval on the current input data of the user, and construct a vector database for vector retrieval on the current input data of the user; fuse the high accuracy and high efficiency of scalar retrieval and the ability of supporting generalization and similarity search of vector retrieval; and then fuse the retrieval results of both scalar retrieval and vector retrieval, so as to provide a more accurate input for the intent recognition model to guide it to more accurately and efficiently recognize the intent of the user and the business type corresponding to the intent.

[0016] It should be understood that the content described in this part is not intended to identify key or important features of the embodiments of the present application, nor to limit the scope of the present application. Other features of the present application will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0017] The above and other features, advantages and aspects of embodiments of the present application will become more apparent by describing in detail preferred embodiments thereof with reference to the attached drawings in which:

[0018] Figure 1 A schematic diagram of a system for implementing a method for business recommendation based on an identified user intent according to an embodiment of the present application is shown.

[0019] Figure 2 A flowchart of a method for business recommendation based on an identified user intent according to an embodiment of the present application is shown.

[0020] Figure 3 A flowchart of a method for vector retrieval in a keyword database according to an embodiment of the present application is shown.

[0021] Figure 4 A flowchart of a method for obtaining a preliminary intent tendency ranking of a user according to an embodiment of the present application is shown.

[0022] Figure 5 A flowchart for determining a current intention of a user and a service type corresponding to the current intention according to an embodiment of the present application is shown.

[0023] Figure 6 A block diagram of an electronic device according to an embodiment of the present application is shown.

[0024] Figure 7 A dataflow block diagram of an intention recognition method according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0025] Exemplary embodiments of the present application are described herein with reference to the accompanying drawings, in which various details are set forth to facilitate an understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that various changes in the embodiments of the present application can be made without departing from the scope and spirit of the present application. Also, descriptions of well-known functions and constructions are omitted for clarity and conciseness.

[0026] The term "include" and its variations are intended to cover a non-exclusive inclusion, such that a process, method, system, product, or apparatus that includes a list of elements is not necessarily limited to those elements, but can include other elements not expressly listed or inherent to such process, method, system, product, or apparatus. Unless otherwise defined, the term "or" means "and / or". The term "based on" means "based, at least in part, on". The terms "one example embodiment" and "an embodiment" mean "at least one example embodiment". The term "another embodiment" means "at least one additional embodiment". The terms "a first", "a second", etc. can refer to different or the same objects. Other explicit or implicit definitions can also be included below.

[0027] As described above, the conventional method for service recommendation based on the identified user intention, such as directly identifying the intention of the user input information through a large model (LLM, LFM), or identifying the intention of the user input information through a pre-set rule engine, has the following disadvantages: due to the divergence of the large model itself, it is difficult to stably identify the real intention of the user; and the rule engine does not support semantic generalization, and it is difficult to accurately identify the intention of the user's fuzzy input.

[0028] In summary, the conventional method for service recommendation based on the identified user intention has the disadvantage of being unable to stably and accurately identify the real intention of the user.

[0029] To at least partially solve one or more of the above problems and other potential problems, example embodiments of the present application propose a scheme for business recommendation based on an identified user intent, in which, first, scalar retrieval and vector retrieval are respectively performed in a keyword database for current input information of a user; ranking is performed based on the scalar retrieval result and the vector retrieval result to obtain a preliminary intent tendency ranking of the user; then, at least the current input information of the user and the preliminary intent tendency ranking of the user are input into an intent recognition model to determine a current intent of the user and a business type corresponding to the current intent via the intent recognition model, wherein the intent recognition model is constructed based on a neural network model. The present application further calls a recommendation tool corresponding to the business type based on the business type corresponding to the current intent of the user, so as to make business recommendation for the current intent of the user.

[0030] Therefore, the present application can identify the user intent in combination with scalar retrieval and vector retrieval, and fuse the high accuracy and high efficiency of scalar retrieval and the semantic generalization capability supported by vector retrieval, so as to perform generalization similarity calculation for scenes and keywords not covered by scalar retrieval in the case of high accuracy and high efficiency; and fuse the preliminary intent tendency of the user obtained by scalar retrieval and vector retrieval, and input the preliminary intent tendency of the user into an intent recognition model, so as to obtain the thinking range of the large model constrained by the preliminary intent tendency of the user obtained by retrieval, and prevent the unstable results caused by excessive divergence, thus being able to give more accurate user intent recognition results on the basis of scalar retrieval and vector retrieval in combination with the intent recognition model; both the instability of pure large model intent recognition is avoided and no additional model training is required, thus reducing the model customization cost; and the corresponding recommendation tool can be called based on the identified current intent of the user in combination with the business scene, so as to make more accurate recommendation.

[0031] Figure 1 A schematic diagram of a system 100 (referred to as system 100 for short) for implementing a method for business recommendation based on an identified user intent according to an embodiment of the present application is shown. As shown in Figure 1 , the system 100 includes a computing device 110, a server 130, a network 140 and user terminals (e.g. Figure 1 user terminals Y1, user terminals Y2, …, user terminals Yn in ). The computing device 110, the server 130 and the user terminals can perform data interaction through the network 140 (e.g. the Internet, a local area network or a wide area network, etc.).

[0032] Regarding the user terminals (e.g. Figure 1user terminals Y1, user terminal Y2, …, user terminal Yn) that are, for example, mobile devices, personal terminals, desktop computers, smart watches, tablet computers, interactive devices, etc., on the display interfaces of which users can perform interactive operations (such as browsing information, inputting information, querying products, etc.), the user terminals perform data interaction with the server 130 through the network, and the user inputs information through the user terminal.

[0033] The server 130 receives the input information of the user from the server 130, performs user intent recognition based on the input information of the user, and performs business recommendation based on the user intent. The server 130 is deployed with a keyword database, or the keyword database is deployed on a cloud storage device that interacts with the server 130; the server 130 is also deployed with an intent recognition model, or the intent recognition model is deployed on a computing device 110 that interacts with the server 130.

[0034] Regarding the server 130, it can be, for example, a standalone physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud database, cloud storage, network service, cloud communication, middleware service, domain name service, security service, and big data and artificial intelligence platform.

[0035] Regarding the server 130, the server 130 is deployed with the method for performing business recommendation based on the identified user intent provided by the embodiments of the present application, which can, for example, obtain the input information of the user through the network 140 and send the input information of the user, or the keywords extracted based on the input information of the user, to the computing device 110 for training the network, model and system, etc. of the method for performing business recommendation based on the identified user intent provided by the embodiments of the present application.

[0036] Regarding the computing device 110, it is, for example, used for training the network model of the method for performing business recommendation based on the identified user intent provided by the embodiments of the present application, for training and / or deploying the intent recognition model, for deploying the keyword database, etc. The computing device 110 can have one or more processing units, including special-purpose processing units such as GPUs, FPGAs and ASICs, and general-purpose processing units such as CPUs. The computing device 110 can be an integration of multiple physical servers, an integration of multiple processing units, etc. In addition, one or more virtual machines can also be running on each computing device 110. In some embodiments, the computing device 110 and the server 130 can be integrated together, or can be separately arranged from each other. In some embodiments, the computing device 110, for example, includes an intent retrieval module 112, a preliminary intent acquisition module 114, an intent recognition module 116, and a business recommendation module 118.

[0037] Regarding the intent retrieval module 112, scalar retrieval and vector retrieval are respectively performed in the keyword database for the current input information of the user.

[0038] Regarding the preliminary intent acquisition module 114, sorting is performed based on the scalar retrieval result and the vector retrieval result to acquire a preliminary intent tendency ranking of the user.

[0039] Regarding the intent recognition module 116, at least the current input information of the user and the preliminary intent tendency ranking of the user are input into an intent recognition model to determine the current intent of the user and a service type corresponding to the current intent via the intent recognition model; the intent recognition model is constructed based on a neural network model.

[0040] Regarding the service recommendation module 118, a recommendation tool corresponding to the service type is called based on the service type corresponding to the current intent of the user, so as to perform service recommendation for the current intent of the user.

[0041] Figure 2 A flowchart of a method 200 for performing service recommendation based on an identified user intent according to an embodiment of the present application is shown. The method 200 for performing service recommendation based on an identified user intent can be executed by the computing device 110 as shown, and can also be executed at the electronic device 600 as shown. It should be understood that the method 200 for performing service recommendation based on an identified user intent can also include additional steps not shown and / or can omit the steps shown, and the scope of the present application is not limited in this respect. Figure 1 Figure 6

[0042] At step 202, the computing device 110 performs scalar retrieval and vector retrieval in the keyword database for the current input information of the user.

[0043] In some embodiments, the keyword database includes a scalar database and a vector database.

[0044] In some embodiments, the scalar database is constructed based on scalar data about commodity data, activity data, application data and / or question data corresponding to the service type.

[0045] In some embodiments, the service type includes commodity recommendation, application recommendation, activity recommendation, question inquiry and / or chat.

[0046] Regarding the scalar database, a large number of keywords are stored in the scalar database, which are directly stored in the form of strings. For example, in the present scheme, a plurality of keywords under the same service type are stored in the form of scalar based on the service type.

[0047] ​​For example, the business type is commodity recommendation, and the stored keywords are "commodity ID (unique identifier), commodity name (e.g., milk tea, watermelon, mobile phone, tennis shoes), commodity category / subcategory (e.g., beverage, fruit, electronic device, sports shoes), brand (e.g., Tea Color, Huawei, Xiaomi, Apple), model / specification (e.g., Shengsheng Oolong, Yulan Latte; 8424, Kirin; iPhone 15Pro Max 256GB, Xiaomi 14 white 512GB; Nike Vapor pro white 42 size)".

[0048] For example, the business type is application recommendation, and the stored keywords are "application ID (e.g., App Store ID, GooglePlay Package Name), application name (exact name, e.g., Honkai Impact, King Glory, QQ, PDF Reader, Scallop, PS), application category / subcategory (e.g., game, social, tool, education, photography), developer / publisher (e.g., Miha, Tencent, Adobe), platform (e.g., iOS, Android, Windows, macOS), version number (current version), file size (xx GB), download / installation volume (interval or specific value), rating (average value, e.g., 4.5 stars), price (e.g., free, paid, in-app purchase), supported languages (e.g., Japanese, Simplified Chinese), age rating (e.g., 4+, 12+, 17+), update time (specific latest update time), keyword label (core words describing application functions, e.g., weather forecast, takeout, express delivery, shopping)".

[0049] In some embodiments, the vector database is constructed based on the vectorization of the commodity data, activity data, application data, and / or question data corresponding to the business type.

[0050] Regarding embedding, for example, text / image / structured data is converted into a fixed-dimensional floating-point number vector using an embedding model (e.g., BERT, CLIP, Text2Vec); for example, for commodity data, activity data, application data, and / or question data corresponding to the business type, semantic and / or visual features are extracted through a semantic model and / or an image model, and then the extracted keywords related to the business type are vectorized and stored in the vector database to facilitate vector retrieval of semantic similarity, image similarity of commodity data, activity data, application data, and / or question data corresponding to the business type.

[0051] For example, we can vectorize product data such as "Apple, Fuji, 10 yuan per jin", activity information such as "Apple, mobile phone", and application data such as "moji, weather forecast", "Meituan, food delivery", "JD.com, shopping / food delivery", and "PUBG, game, FPS". It should be understood that data for various business types can be adjusted according to needs. They will not be listed one by one here, but only a few examples are given.

[0052] For example, in different business scenarios, the same keyword may contain one or more meanings. The meaning of the keyword in the user's intent is determined based on the related meanings of the keyword and the user's input. Regarding the same keyword containing one or more meanings, the meaning of the same keyword corresponding to the user's current input is determined by the difference in business type, the meanings of other related keywords, the user's current input information, and / or the user's initial intent. For example, does the user's input of "apple" refer to "apple among fruits" or "Apple electronic devices"?

[0053] It is worth noting that the same keyword may represent different meanings and correspond to different data content in different business scenarios. For example, the keyword "apple" in the product recommendation business type and the keyword "apple" in the application recommendation business type may have different meanings and data content. The former may correspond to "apple fruit" or "Apple mobile phone", while the latter may correspond to "Apple Store".

[0054] In step 204, the computing device 110 sorts the results based on the scalar retrieval results and the vector retrieval results to obtain the user's preliminary intention preference ranking.

[0055] For example, based on scalar retrieval results and vector retrieval results, three preliminary intent tendencies of the user are obtained: "preliminary intent" and "vector intent". Figure 1 70%; preliminary intention Figure 2 90%; preliminary intention Figure 3 "60%; " Based on the weights (or probabilities) of these initial intentions, the users' initial intention tendency ranking is obtained. Figure 2 90%; preliminary intention Figure 1 70%; preliminary intention Figure 3 60%

[0056] Therefore, the application can combine scalar data and vector data, perform scalar retrieval on the current input data of the user to obtain accurate retrieval results, and perform vector retrieval to obtain retrieval results with high similarity, thereby combining the high accuracy and efficiency of scalar retrieval and the semantic generalization capability of vector retrieval, and fusing the retrieval results of scalar retrieval and vector retrieval to obtain a preliminary intention tendency ranking of the user, thereby providing a more accurate input for the intention recognition model to guide more accurate and efficient recognition of the user intention.

[0057] The following will be combined Figure 3 The method of performing vector retrieval in the keyword database will be described in detail; and the method of fusing the retrieval results of scalar retrieval and vector retrieval will be combined Figure 4 The method of obtaining the preliminary intention tendency ranking of the user will be described in detail; here, no further description is given.

[0058] In step 206, the computing device 110 inputs the current input information of the user and the preliminary intention tendency ranking of the user into the intention recognition model to determine the current intention of the user and the business type corresponding to the current intention via the intention recognition model; the intention recognition model is constructed based on a neural network model.

[0059] For example, it is determined that the current intention of the user is to purchase fruit apples, and the corresponding business type is commodity recommendation; it is determined that the current intention of the user is to purchase a mobile phone, and the corresponding business type is commodity recommendation; it is determined that the current intention of the user is to order takeout, and the corresponding business type is application recommendation and / or commodity recommendation, such as recommending a takeout application, recommending a takeout commodity link, etc.; it is determined that the current intention of the user is to consult a question, and the corresponding business type is question consultation, etc.

[0060] The following will be combined Figure 5 The method of determining the current intention of the user and the business type corresponding to the current intention will be described in detail; here, no further description is given.

[0061] In step 208, the computing device 110 calls a recommendation tool corresponding to the business type based on the business type corresponding to the current intention of the user, so as to perform business recommendation for the current intention of the user.

[0062] For example, the business type is commodity recommendation, and then a commodity search tool is used to recommend related commodities to the user; for example, the business type is activity recommendation, and then an activity search tool is used to recommend related commodities to the user; the business type is application recommendation, and then an application search tool is used to recommend related commodities to the user; for example, the business type is chat, and then a chat tool is used to chat with the user; for example, the business type is question consultation, and then a question and answer tool is used to answer related questions for the user.

[0063] Thus, by determining the business type corresponding to the intention, the intention of the user can be determined more accurately, and based on the user intention, an adapted business tool is called to make more matched business recommendations for the user.

[0064] In the above scheme, the advantages of both scalar and vector are combined, a scalar database is constructed for scalar retrieval of the current input data of the user, and a vector database is constructed for vector retrieval of the current input data of the user; the high accuracy and high efficiency of scalar retrieval and the generalization support and similarity search capability of vector retrieval are combined; then the retrieval results of both scalar retrieval and vector retrieval are combined, thereby providing a more accurate input for the intention recognition model to guide more accurate and efficient recognition of the user intention and the business type corresponding to the intention.

[0065] In addition, the above scheme also provides a plurality of business tools corresponding to each business type, so as to call the business tool corresponding to the business type to make more matched business recommendations for the user according to the identified user intention, compared with the traditional business recommendation method which only uses one tool to make a variety of business recommendations, the above scheme can obtain more efficient and more targeted business recommendations, so that the recommendation result is more matched to the current input content of the user and more consistent with the real intention of the user.

[0066] Figure 3 A flowchart of a method 300 for vector retrieval in a keyword database according to an embodiment of the present application is shown. The method 300 for vector retrieval in a keyword database can be executed by the computing device 110 as shown in Figure 1 , and can also be executed at the electronic device 600 as shown in Figure 6 . It should be understood that the method 300 for vector retrieval in a keyword database can also include additional steps not shown and / or can omit the steps shown, and the scope of the present application is not limited in this regard.

[0067] At step 302, the computing device 110 vectorizes the current input information of the user to obtain the vectorized current input information.

[0068] Regarding the vectorization of the current input information of the user, for example, using an embedding model (such as BERT, CLIP, Text2Vec) to convert the current input information data of the user into vector information, so as to facilitate similarity retrieval in the vector database, the current input information of the user can be one or more of text, picture and voice, if the user inputs voice information, the voice is converted into text; supporting multiple types of input information of the user.

[0069] At step 304, the computing device 110 performs a similarity search in the vector database for the vectorized current input information of the user.

[0070] At step 306, the computing device 110 identifies at least one service type and at least one preliminary intent tendency corresponding to the current input information of the user based on the semantic similarity.

[0071] Regarding the scalar search in the keyword database, for example, after receiving the current input information of the user, a scalar search and a vector search are respectively performed for the current input information of the user, and the vectorization operation is not required for the current input information of the user in the scalar search. In the scalar search, only the operation such as the exact search, the fuzzy matching, etc. is required for the current input information of the user; for example, the current input information of the user “What is good to eat for takeout today? I want to order lunch”, in the scalar search, the full-text exact search, the keyword exact search (such as for “takeout” and “lunch”), the fuzzy matching search, etc. are respectively performed for the input information.

[0072] Thus, the current input information of the user is respectively subjected to the scalar search and the vector search, and the vectorization operation is not required for the current input information of the user in the scalar search, and the current input information of the user is directly subjected to the search in the scalar database in the original form and / or the keyword extracted based on the original form, thereby improving the search efficiency; and the vectorization operation is required for the current input information of the user in the vector search, thereby being capable of performing the similarity calculation, so as to search out the result with high similarity to the current input information of the user in the vector database, to determine the preliminary intent of the user, thereby solving the limitation problem of the scalar database, and supplementing the result that cannot be searched in the scalar search.

[0073] Figure 4 A flow chart of a method 400 for obtaining a preliminary intent tendency ranking of a user according to an embodiment of the present application is shown. The method 400 for obtaining the preliminary intent tendency ranking of the user can be executed by the computing device 110 as shown in Figure 1 , and can also be executed at the electronic device 600 as shown in Figure 6 . It should be understood that the method 400 for obtaining the preliminary intent tendency ranking of the user can also include additional steps not shown and / or can omit the steps shown, and the scope of the present application is not limited in this respect.

[0074] At step 402, the computing device 110 obtains a plurality of preliminary intent tendencies of the user based on the scalar search result, so as to respectively assign corresponding weights to the plurality of preliminary intent tendencies

[0075] For example, based on the scalar retrieval result, obtain (preliminary intent tendency 1; preliminary intent tendency 2), assign weights (preliminary intent tendency 1, 0.9; preliminary intent tendency 2, 0.7).

[0076] At step 404, the computing device 110 obtains, based on the vector retrieval result, a plurality of preliminary intent tendencies of the user for which the semantic similarity is higher than a predetermined similarity threshold, respectively assigning corresponding weights.

[0077] For example, based on the vector retrieval result, obtain (preliminary intent tendency 2; preliminary intent tendency 3, preliminary intent tendency 4), assign weights (preliminary intent tendency 2, 0.8; preliminary intent tendency 3, 0.7; preliminary intent tendency 4, 0.2).

[0078] At step 406, the computing device 110 merges the preliminary intent tendencies of the user and corresponding weights obtained based on the scalar retrieval result and the vector retrieval result respectively, so as to sort the obtained preliminary intent tendencies of the user based on the weights from high to low.

[0079] In some embodiments, merging the preliminary intent tendencies of the user and corresponding weights obtained based on the scalar retrieval result and the vector retrieval result respectively comprises: for the current input information of the user, in response to confirming that the scalar retrieval result and the vector retrieval result exist simultaneously for the same preliminary intent tendency, excluding the vector retrieval result for the same preliminary intent tendency.

[0080] For example, continuing the above example, the scalar retrieval result (preliminary intent tendency 1, 0.9; preliminary intent tendency 2, 0.7); the vector retrieval result (preliminary intent tendency 2, 0.8; preliminary intent tendency 3, 0.7; preliminary intent tendency 4, 0.2); merging the two, obtaining the preliminary intent tendencies (preliminary intent tendency 1, 0.9; preliminary intent tendency 2, 0.7; preliminary intent tendency 3, 0.7; preliminary intent tendency 4, 0.2), the preliminary intent tendency 2 appears in both the scalar retrieval result and the vector retrieval result, at this time, the scalar retrieval result is given priority, and the repeated vector retrieval result is removed.

[0081] Thus, the characteristics of both the scalar retrieval and the vector retrieval can be combined, the scalar retrieval provides more accurate retrieval results, in the blind area of the scalar retrieval, the vector retrieval provides more similar and closer retrieval results, and the two are fused to obtain the preliminary intent tendency of the user corresponding to the current input information of the user.

[0082] Figure 5 A flowchart of a method 500 for determining a current intent of a user and a service type corresponding to the current intent according to an embodiment of the present application is shown. The method 500 for determining a current intent of a user and a service type corresponding to the current intent can be performed by the computing device 110 as shown, and can also be performed by a server. Figure 1 The method 500 for determining a current intent of a user and a service type corresponding to the current intent can be performed by the computing device 110 as shown, and can also be performed by a server.Figure 6 The method 500 for determining the current intention of the user and the business type corresponding to the current intention can further include additional steps not shown or can omit the steps shown, and the scope of the present application is not limited in this respect.

[0083] At step 502, the computing device 110 inputs the current input information of the user, the historical input information of the user, the historical intention information of the user, and the preliminary intention tendency ranking of the user into the intention recognition model.

[0084] As to the intention recognition model, it is, for example, a Large Language Model (LLM), a Large Foundation Model (LFM), or is built based on the LLM model, the LFM model, or a neural network model.

[0085] For example, the preliminary intention tendency ranking of the user is input into the intention recognition model, and the top n input intention recognition model is ranked, for example, the top two preliminary intention tendencies, or the top three preliminary intention tendencies.

[0086] As to the historical input information of the user, it includes, for example, the input information of the user within t minutes (e.g., t is 10, t is 100), the previous m input information of the user (e.g., m is 5, m is 10), and / or the historical question and answer information of the user.

[0087] At step 504, the computing device 110 identifies the predicted intention tendency of the user via the intention recognition model, and the predicted intention tendency includes at least one intention tendency belonging to a predetermined intention tendency set and / or at least one preliminary intention tendency.

[0088] As to the predetermined intention tendency set, it is, for example, a FAQ intention (FAQ, Frequently Asked Questions), i.e., a common intention in the present application, and further includes a predetermined intention tendency set for collecting common intentions. When the FAQ intention tendency exists in one or more intention tendencies identified by the intention recognition model based on the input information, and the probability of the FAQ intention tendency is the highest, the FAQ intention tendency is identified as the current intention of the user. If the probability of the FAQ intention tendency is not the highest, the identified intention tendencies are ranked.

[0089] For example, the intention recognition model outputs at least one predicted intention tendency, for example, two, so as to determine the current intention of the user and the business type corresponding to the current intention according to the output predicted intention tendency, and further perform business recommendation.

[0090] At step 506, the computing device 110 ranks the probabilities of the predicted intention tendencies of the user in order to determine the current intention of the user and the business type corresponding to the current intention.

[0091] For example, the intention recognition model identifies one or more intentions according to the input data (the current input information of the user, the historical input information of the user, the historical intention information of the user, and the preliminary intention tendency ranking of the user), and if there is an FAQ intention and the intention score is the highest, the FAQ intention is directly identified as the current intention of the user; if there are other multiple intentions, the top two intention tendencies are selected according to the ranking of the intention tendencies, and then business recommendations are made based on the two intention tendencies. For example, if the business type is 1, the user intention is product recommendation, the intention recognition model calls the product query interface through FunctionCall, and returns a list of product information to the user; if the business type is 2, the user intention is application recommendation, the intention recognition model calls the application query interface through FunctionCall, and returns a list of applications to the user; if the business type is 3, the user intention is discount activity recommendation, the intention recognition model calls the discount activity query interface through FunctionCall, and returns a discount activity to the user; if the business type is 4, the user intention is question inquiry, the intention recognition model retrieves and enhances through the RAG (Retrieval-Augmented Generation) module to generate an answer, and feeds the answer back to the user; if the business type is empty, the large model determines if it is a chatting scenario, and then calls the chat module to chat with the user.

[0092] For example, the input information of the user, the intention prediction information of the intention recognition model are collected based on a predetermined period in order to determine the accuracy of the prediction results of the intention recognition model; the accuracy feedback information of the intention recognition model is provided to the optimization module in order to adjust the related data stored in the scalar database, the vector database, and / or the predetermined intention tendency set.

[0093] Figure 7 A data flow framework diagram of the intention recognition method according to an embodiment of the present application is shown. The intention recognition method provided by the embodiment of the present application is described in combination with the data flow framework diagram. Figure 7 The intention recognition method provided by the embodiment of the present application is further described.

[0094] The current input information is obtained from the user;

[0095] The scalar retrieval result is obtained by performing scalar retrieval on the current input information via the scalar database;

[0096] The vector retrieval result is obtained by performing vector retrieval on the current input information after vectorization via the vector database;

[0097] By merging scalar and vector search results, preliminary intent tendencies can be obtained.

[0098] At least the current input information and the initial intent tendency are input into the intent recognition model (one or more of the historical input information and historical intent information can also be input), and the current intent is obtained through the intent recognition model;

[0099] Based on the current intent and the corresponding business domain, a recommendation tool is invoked to make business recommendations. The recommendation tool includes multiple tools (such as tool a11, tool a21, ..., tool an1). For example, the corresponding recommendation tool is invoked based on the business type. For example, if the business type corresponding to the current intent is business type a11, tool a11 is invoked.

[0100] Figure 6 A schematic step diagram of an example electronic device 600 that can be used to implement embodiments of the contents of this specification is shown. For example, as Figure 1 The computing device 110 shown can be implemented by electronic device 600. As shown, electronic device 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to computer program instructions stored in read-only memory (ROM) 602 or loaded into random access memory (RAM) 603 from storage unit 608. The RAM 603 can also store various programs and data required for the operation of electronic device 600. The CPU 601, ROM 602, and RAM 603 are interconnected via bus 604. Input / output (I / O) interface 605 is also connected to bus 604.

[0101] Multiple components in electronic device 600 are connected to input / output I / O interface 605, including: input unit 606, such as keyboard, mouse, microphone, etc.; output unit 607, such as various types of monitors, speakers, etc.; storage unit 608, such as disk, optical disk, etc.; and communication unit 609, such as network card, modem, wireless transceiver, etc. Communication unit 609 allows electronic device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0102] The various processes and processes described above, such as the method 200 for making a service recommendation based on an identified user intent to the method 500 for determining a current intent of a user and a service type corresponding to the current intent, can be performed by the central processing unit CPU 601. For example, in some embodiments, the method 200 for making a service recommendation based on an identified user intent to the method 500 for determining a current intent of a user and a service type corresponding to the current intent can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 608. In some embodiments, some or all of the computer program can be loaded and / or installed onto the electronic device 600 via the read-only memory ROM 602 and / or the communication unit 609. When the computer program is loaded onto the random access memory RAM 603 and executed by the central processing unit CPU 601, one or more acts of the method 200 for making a service recommendation based on an identified user intent to the method 500 for determining a current intent of a user and a service type corresponding to the current intent described above can be performed.

[0103] The present application is related to methods, apparatus, systems, electronic devices, computer readable storage media, and / or computer program products. The computer program product can include computer readable program instructions for executing various aspects of the present application.

[0104] The computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium can be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium include the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or punched tape, and any suitable combination of the foregoing. A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.

[0105] Computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network can comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge computers. A network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing / processing device.

[0106] Computer readable program instructions for carrying out operations of the present application can be assembler instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer readable program instructions can execute entirely on the user's computing / processing device, partly on the user's computing / processing device, as a stand-alone software package, partly on the user's computing / processing device and partly on a remote computing / processing device or entirely on the remote computing / processing device or server. In the latter scenario, the remote computing / processing device can be connected to the user's computing / processing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computing / processing device, for example, through the Internet using an Internet Service Provider. In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate arrays (FPGA), or programmable logic arrays (PLA) can execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present application.

[0107] The computer readable program instructions can also be loaded onto a computing / processing device, other programmable data processing apparatus, or other device to cause a series of operations to be performed on the computing / processing device, other programmable apparatus or other device to produce a computer implemented process, such that the instructions which execute on the computing / processing device, other programmable apparatus, or other device implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0108] These computer readable program instructions can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks. These computer readable program instructions can also be stored in a computer readable storage medium that can include a non-transitory computer readable storage medium that can be a computer- readable storage medium having no data storage cycles that change state. The instructions can be executed by one or more processors of a computer, other programmable data processing apparatus, or other devices to produce a computer-implemented process such that the instructions which execute via the one or more processors of the computer or other programmable data processing apparatus create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks. The computer program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0109] The computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0110] The flow and block diagrams in the drawings show the architectural, functional, and operational views of possible implementations of systems, methods, and computer program products according to the embodiments of the present application. In this regard, each block in the flow and block diagrams can represent a module, a segment, or a portion of code which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently or the blocks may

[0111] Embodiments of the application have been described above, and the description is intended to be illustrative of the embodiments of the application and not exhaustive or limiting. Numerous modifications and adaptations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The scope of the embodiments of the application is defined by the appended claims, rather than the description preceding it, which is provided primarily for purposes of patentability, and wherein features providing for patentability are as broad as possible within the scope of the patent laws. The choice of words in the description is not intended to limit the scope of the embodiments of the application but is intended to highlight the principles of the embodiments of the application, practical applications of the embodiments of the application, or improvements in the art made by the embodiments of the application, or to enable others skilled in the art to understand the embodiments of the application.

Claims

1. A method for business recommendation based on an identified user intent, characterized by, The method comprises: performing scalar retrieval and vector retrieval in a keyword database based on current input information of a user; in the scalar retrieval, performing full-text accurate retrieval, keyword accurate retrieval and fuzzy matching retrieval based on the current input information; sorting based on the scalar retrieval result and the vector retrieval result to obtain a preliminary intention tendency ranking of the user; inputting at least the current input information of the user and the preliminary intention tendency ranking of the user into an intention recognition model to determine, via the intention recognition model, a current intention of the user and a business type corresponding to the current intention; the intention recognition model is constructed based on a neural network model; the business type includes: commodity recommendation, application recommendation, activity recommendation, question inquiry and / or chat; based on the business type corresponding to the current intention of the user, calling a recommendation tool corresponding to the business type to perform business recommendation for the current intention of the user; and merging the preliminary intention tendency of the user obtained from the scalar retrieval result and the vector retrieval result and corresponding weights, including: in response to the existence of the same preliminary intention tendency in the scalar retrieval result and the vector retrieval result, excluding the vector retrieval result of the same preliminary intention tendency for the current input information of the user; In different business scenarios, a same keyword contains one or more meanings, and based on the associated keyword meaning and the input content of the user, the corresponding meaning of the keyword in the current user intention is determined.

2. The method of claim 1, wherein, The keyword database comprises: a scalar database constructed based on scalar data about commodity data, activity data, application data and / or question data corresponding to the business type; and a vector database constructed based on vectorization of the commodity data, the activity data, the application data and / or the question data corresponding to the business type.

3. The method of claim 1, wherein, The scalar retrieval and the vector retrieval in the keyword database comprise: vectorizing the current input information of the user to obtain vectorized current input information; performing similarity retrieval in the vector database based on the vectorized current input information; and based on the high and low of the semantic similarity, identifying at least one business type and at least one preliminary intention tendency corresponding to the current input information of the user.

4. The method of claim 1, wherein, The sorting based on the scalar retrieval result and the vector retrieval result to obtain the preliminary intention tendency ranking of the user comprises: based on the scalar retrieval result, obtaining multiple preliminary intention tendencies of the user to respectively assign corresponding weights to the multiple preliminary intention tendencies; based on the vector retrieval result, obtaining multiple preliminary intention tendencies of the user for which the semantic similarity is higher than a predetermined similarity threshold, and respectively assigning corresponding weights; and merging the preliminary intention tendencies of the user and the corresponding weights obtained based on the scalar retrieval result and the vector retrieval result respectively to sort the obtained preliminary intention tendencies of the user based on the weights from high to low.

5. The method of claim 1, wherein, The inputting at least the current input information of the user and the preliminary intention tendency ranking of the user into the intention recognition model to determine, via the intention recognition model, the current intention of the user and the business type corresponding to the current intention comprises: obtaining current input information of a user, historical input information of the user, historical intent information of the user, and a preliminary intent tendency ranking of the user, to input an intent recognition model; identifying, via the intent recognition model, a predicted intent tendency of the user, the predicted intent tendency comprising at least one intent tendency belonging to a predetermined intent tendency set, and / or at least one preliminary intent tendency; and ranking based on probabilities of the predicted intent tendency of the user so as to determine a current intent of the user and a business type corresponding to the current intent.

6. A computing device, comprising: comprising: at least one processing unit; at least one memory coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions, when executed by the at least one processing unit, cause the device to perform steps of the method according to any one of claims 1 to 5.

7. A computer-readable storage medium having stored thereon a computer program, characterized in that The computer program, when executed by a machine, implements the method according to any one of claims 1 to 5.

8. A computer program product comprising a computer program, characterized in that, The computer program, when executed by a machine, performs the method according to any one of claims 1 to 5.

Images