Resource recommendation method and device

Abstract

The present application relates to a resource recommendation method, device, computer equipment and storage medium. The method includes: obtaining a first text vector corresponding to each text to be recommended, the first text vector includes a first keyword vector and a first weight vector corresponding to the first keyword; the first keyword vector is used to characterize the semantic features of the first keyword; obtaining a second text vector corresponding to the text to be matched, the second text vector includes a second keyword vector and a second weight vector corresponding to the second keyword; the second keyword vector is used to characterize the semantic features of the second keyword; according to the cosine similarity between the first text vector and the second text vector, each first text vector and the second text vector are matched, and from each first text vector, the first text vector with the highest cosine similarity to the second text vector is determined as the matching target text vector. This method can improve the recommendation accuracy of the recommended consulting doctors.

Description

Technical Field

[0001] The present application relates to the field of computer application technology, and in particular to a resource recommendation method and device. Background Art

[0002] With the continuous development of Internet technology, more and more users choose to share resources through Internet technology, such as remote consultation, remote education and training, and remote monitoring, which provides users with more convenience in real-time resource sharing. Taking remote consultation as an example, when grassroots doctors have questions about the diagnosis and treatment of the disease and need to contact experts in higher-level hospitals, experts in higher-level hospitals can provide grassroots doctors with diagnostic opinions and treatment suggestions through remote consultation. However, grassroots doctors may not necessarily understand the treatment subdivisions of experts in higher-level hospitals. Even in the same department, experts may have different areas of expertise. Therefore, when grassroots doctors apply for expert consultations from higher-level hospitals, it is necessary to recommend suitable experts from higher-level hospitals to grassroots doctors.

[0003] In traditional technology, the recommendation method mainly based on content is used to recommend suitable experts from higher-level hospitals to primary care doctors. This recommendation method first extracts keywords from the patient text to be matched and the expert text to be recommended provided by the primary care doctors, and then calculates the weights corresponding to each keyword in the patient text to be matched and the expert text to be recommended, respectively. Finally, the similarity between the vector composed of the weights corresponding to each keyword in the patient text to be matched and the vector composed of the weights corresponding to each keyword in the expert text to be recommended is calculated, and expert resources are recommended based on the calculated similarity. For example, the keyword in the patient text to be matched is "toothache". According to the above content-based recommendation method, only the experts corresponding to the expert text to be recommended with the highest similarity to "toothache" will be recommended.

[0004] However, traditional content-based resource recommendation methods have low accuracy. Summary of the invention

[0005] Based on this, it is necessary to provide a resource recommendation method, apparatus, computer device and storage medium to address the problems of traditional content-based resource recommendation methods.

[0006] In a first aspect, the present application provides a resource recommendation method, the method comprising:

[0007] Obtaining a first text vector corresponding to each to-be-recommended text, the first text vector including a first keyword vector and a first weight vector corresponding to the first keyword; the first keyword vector is used to represent a semantic feature of the first keyword;

[0008] Obtaining a second text vector corresponding to the text to be matched, the second text vector including a second keyword vector and a second weight vector corresponding to the second keyword; the second keyword vector is used to represent the semantic features of the second keyword;

[0009] According to the cosine similarity between the first text vector and the second text vector, each first text vector is matched with the second text vector, and the first text vector having the highest cosine similarity with the second text vector among the first text vectors is determined as the matching target text vector.

[0010] In a second aspect, the present application provides a resource recommendation device, the device comprising:

[0011] A first text vector acquisition module is used to acquire a first text vector corresponding to each text to be recommended, wherein the first text vector includes a first keyword vector and a first weight vector corresponding to the first keyword; the first keyword vector is used to represent the semantic features of the first keyword;

[0012] A second text vector acquisition module is used to acquire a second text vector corresponding to the text to be matched, the second text vector includes a second keyword vector and a second weight vector corresponding to the second keyword; the second keyword vector is used to represent the semantic features of the second keyword;

[0013] The target recommendation resource determination module is used to match each first text vector with the second text vector according to the cosine similarity between the first text vector and the second text vector, and determine the first text vector with the highest cosine similarity with the second text vector from among the first text vectors as the matching target text vector.

[0014] In a third aspect, the present application provides a computer device, the computer device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the following steps when executing the computer program:

[0015] Obtaining a first text vector corresponding to each to-be-recommended text, the first text vector including a first keyword vector and a first weight vector corresponding to the first keyword; the first keyword vector is used to represent a semantic feature of the first keyword;

[0016] Obtaining a second text vector corresponding to the text to be matched, the second text vector including a second keyword vector and a second weight vector corresponding to the second keyword; the second keyword vector is used to represent the semantic features of the second keyword;

[0017] According to the cosine similarity between the first text vector and the second text vector, each first text vector is matched with the second text vector, and the first text vector having the highest cosine similarity with the second text vector among the first text vectors is determined as the matching target text vector.

[0018] In a fourth aspect, the present application provides a readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the following steps are implemented:

[0019] Obtaining a first text vector corresponding to each to-be-recommended text, the first text vector including a first keyword vector and a first weight vector corresponding to the first keyword; the first keyword vector is used to represent a semantic feature of the first keyword;

[0020] Obtaining a second text vector corresponding to the text to be matched, the second text vector including a second keyword vector and a second weight vector corresponding to the second keyword; the second keyword vector is used to represent the semantic features of the second keyword;

[0021] According to the cosine similarity between the first text vector and the second text vector, each first text vector is matched with the second text vector, and the first text vector having the highest cosine similarity with the second text vector among the first text vectors is determined as the matching target text vector.

[0022] The resource recommendation method, device, computer device and storage medium provided in this embodiment can obtain the first text vector corresponding to each text to be recommended, including the first keyword vector and the first weight vector corresponding to the first keyword, wherein the first keyword vector is used to characterize the semantic features of the first keyword; and can also obtain the second text vector corresponding to the text to be matched, including the second keyword vector and the second weight vector corresponding to the second keyword, wherein the second keyword vector is used to characterize the semantic features of the second keyword; according to the cosine similarity of the first text vector and the second text vector, each first text vector and the second text vector are matched, and from each first text vector, the first text vector with the highest cosine similarity to the second text vector is determined as the matching target text vector. Since the first keyword vector and the second keyword vector can be used to characterize the semantic features of keywords, when the first keyword in the text to be recommended and the second keyword in the text to be matched have more keyword semantics that are the same or close, the calculated cosine similarity between the first text vector and the second text vector will also be relatively high. Therefore, the consulting doctor recommended to the patient using the resource method provided in this embodiment will also be more appropriate, thereby improving the recommendation accuracy of the consulting doctor to be recommended. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A schematic diagram of a computer device structure provided by an embodiment;

[0024] Figure 2 A schematic diagram of a resource recommendation method flow chart provided in an embodiment;

[0025] Figure 3A schematic diagram of a resource recommendation method flow chart provided in another embodiment;

[0026] Figure 4 A schematic diagram of a resource recommendation method flow chart provided in another embodiment;

[0027] Figure 5 A schematic diagram of the structure of a resource recommendation device provided by an embodiment;

[0028] Figure 6 A schematic diagram of the structure of a resource recommendation device provided in another embodiment;

[0029] Figure 7 A schematic diagram of the structure of a resource recommendation device provided in another embodiment. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0031] The resource recommendation method provided in this embodiment can be applied to Figure 1 The computer device shown includes a processor, a memory and a network interface connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. Optionally, the computer device can be a mobile phone, a tablet computer, a personal digital assistant, etc., and this embodiment does not limit the specific form of the computer device.

[0032] It should be noted that the resource recommendation method provided in the embodiment of the present application may be implemented by a resource recommendation device, which may be implemented as part or all of a computer device through software, hardware, or a combination of software and hardware. In the following method embodiments, the implementation subject is a computer device as an example for description.

[0033] Figure 2 A schematic diagram of a resource recommendation method flow chart provided in an embodiment. This embodiment relates to an implementation process in which a computer device determines a matching target text vector based on the cosine similarity between the first text vector corresponding to each to-be-recommended text and the second text vector corresponding to the to-be-matched text. Figure 2 As shown, the method may include:

[0034] S201, obtaining a first text vector corresponding to each text to be recommended, where the first text vector includes a first keyword vector and a first weight vector corresponding to the first keyword; the first keyword vector is used to represent a semantic feature of the first keyword.

[0035] Specifically, when it is necessary to recommend a suitable doctor to a patient, the above-mentioned text to be recommended can be a text of the consulting doctor's expertise in the field of treatment. When it is necessary to recommend a suitable book to a certain user, the above-mentioned text to be recommended can be an introduction text of each book. This embodiment does not limit the type of text to be recommended. Optionally, the number of the above-mentioned texts to be recommended can be one or more. Further, the above-mentioned first keyword is a keyword extracted from the text to be recommended. The above-mentioned first keyword vector can be a vector composed of one or more numerical values, and each dimension in the first keyword vector can be used to characterize the semantic features of the corresponding keyword. It should be noted that the numerical values ​​of keywords with similar semantics will also be relatively close. For example, the word vector corresponding to the keyword "toothache" can be 0.12, and the word vector corresponding to the keyword "swelling gums" with similar semantics to "toothache" can be 0.14. The first weight vector corresponding to the above-mentioned first keyword can be a vector composed of the weights of each keyword in the text to be recommended in the text to be recommended, or it can be a vector composed of the weights of all keywords extracted from the text to be recommended in the text to be recommended.

[0036] S202, obtaining a second text vector corresponding to the text to be matched, where the second text vector includes a second keyword vector and a second weight vector corresponding to the second keyword; the second keyword vector is used to represent the semantic features of the second keyword.

[0037] Specifically, when it is necessary to recommend a suitable doctor to a patient, the above-mentioned text to be matched can be the patient's medical record text. When it is necessary to recommend a suitable book to a user, the above-mentioned text to be matched can be a bibliographic text that the user frequently reads. This embodiment does not limit the type of text to be matched. Furthermore, the above-mentioned second keyword is a keyword extracted from the text to be matched. The above-mentioned second keyword vector can be a vector composed of one or more numerical values, and each dimension in the second keyword vector can be used to characterize the semantic features of the corresponding keyword. The second weight vector corresponding to the above-mentioned second keyword can be a vector composed of the weights of each keyword in the text to be matched in the text to be matched, or it can be a vector composed of the weights of each keyword in the text to be matched in all keywords extracted from the text to be matched.

[0038] S203, matching each first text vector with the second text vector according to the cosine similarity between the first text vector and the second text vector, and determining, from among the first text vectors, the first text vector having the highest cosine similarity with the second text vector as the matching target text vector.

[0039] Specifically, after obtaining the first text vector corresponding to each text to be recommended and the second text vector corresponding to the text to be matched, the computer device can determine the cosine similarity between the first text vector and the second text vector according to the first cosine similarity between the first keyword vector in the first text vector and the second keyword vector in the second text vector, and the second cosine similarity between the first weight vector in the first text vector and the second weight vector in the second text vector, and according to the sum of the first cosine similarity and the second cosine similarity; it can also determine the cosine similarity between the first text vector and the second text vector according to the first product value of the cosine similarity between the first keyword vector in the first text vector and the second keyword vector in the second text vector and the preset keyword weight, and the second product value of the cosine similarity between the first weight vector in the first text vector and the second weight vector in the second text vector and the weight of the preset weight vector, and according to the sum of the first product value and the second product value; it can also determine the cosine similarity according to the first text vector composed of the first keyword vector and the first weight vector, and the second text vector composed of the second keyword vector and the second weight vector, and this embodiment does not limit this. Furthermore, the computer device may determine the cosine similarity between each first text vector and the second text vector in the first text vector according to the above method of determining the cosine similarity between the first text vector and the second text vector.

[0040] After determining the cosine similarities of the above-mentioned first text vectors and the second text vectors, the computer device can determine the first text vector with the highest cosine similarity to the second text vector from the first text vectors as the matching target text vector. When it is necessary to recommend a suitable doctor to the patient, the doctor corresponding to the matching target text vector can be recommended to the patient. Optionally, the first text vector whose cosine similarity to the second text vector is greater than a preset threshold value can also be determined as the matching target text vector from the first text vectors. In this case, the target text vector can be one or more. When there are multiple target text vectors, the patient can select the desired doctor from the matching multiple doctors. Optionally, the doctor information corresponding to the first text vector can also be displayed to the user in order of the cosine similarity with the second text vector from large to small, so as to provide the user with more choices and improve the flexibility of user selection. Furthermore, multiple matching doctors can also be displayed to the user according to the preset types and the preset order corresponding to each type. The preset types can include at least one of the doctor's praise rate, the number of cooperation cases of the doctor, the time required to wait for the doctor, and the doctor's current workload. The preset order can be at least one of the order from high to low, from more to less, and from long to short, so as to facilitate users to make choices according to actual needs.

[0041] It should be noted that when the semantics of a keyword in the first text vector are relatively close to that of a keyword in the second text, the word vectors of the two keywords will also be relatively close, so the cosine similarity of the two words calculated will also be relatively high. For example, the semantics of the keyword "toothache" in the first text vector and the keyword "swollen and painful teeth" in the second text are relatively close, and the cosine similarity of the word vectors corresponding to "toothache" and "swollen and painful teeth" will also be relatively high. When the first keyword in the text to be recommended and the second keyword in the text to be matched have more keyword semantics that are the same or relatively close, the cosine similarity of the calculated first text vector and the second text vector will also be relatively high. Therefore, the resource method provided in this embodiment can be used to recommend a suitable consulting doctor to the patient.

[0042] In the resource recommendation method provided in this embodiment, the computer device can obtain the first text vector corresponding to each text to be recommended, including the first keyword vector and the first weight vector corresponding to the first keyword, wherein the first keyword vector is used to characterize the semantic features of the first keyword; and can also obtain the second text vector corresponding to the text to be matched, including the second keyword vector and the second weight vector corresponding to the second keyword, wherein the second keyword vector is used to characterize the semantic features of the second keyword; according to the cosine similarity of the first text vector and the second text vector, each first text vector and the second text vector are matched, and from each first text vector, the first text vector with the highest cosine similarity to the second text vector is determined as the matching target text vector. Since the first keyword vector and the second keyword vector can be used to characterize the semantic features of keywords, when the first keyword in the text to be recommended and the second keyword in the text to be matched have more keyword semantics that are the same or relatively close, the calculated cosine similarity between the first text vector and the second text vector will also be relatively high. Therefore, the consulting doctor recommended to the patient using the resource method provided in this embodiment will also be more appropriate, thereby improving the recommendation accuracy of the consulting doctor to be recommended.

[0043] Figure 3 A flowchart of a resource recommendation method provided for another embodiment. This embodiment relates to the implementation process of a computer device performing a preprocessing operation on each text to be recommended to obtain a first keyword set, determining a first weight vector and a first keyword vector based on the first keyword set, and determining a first text vector based on the first weight vector and the first keyword vector. Figure 3 As shown in the above Figure 2 Based on the illustrated embodiment, before the above S201, the method may further include:

[0044] S301, performing a preprocessing operation on each text to be recommended to obtain a first keyword set of each text to be recommended.

[0045] Specifically, the computer device may perform keyword extraction on each text to be recommended according to a preset preprocessing operation to obtain a keyword set for each text to be recommended, and the keyword set corresponding to all texts to be recommended constitutes a first keyword set. Optionally, the preprocessing operation may include a word segmentation operation and a stop word removal operation. In this case, the above-mentioned preprocessing operation on each text to be recommended to obtain the first keyword set for each text to be recommended may include: performing a word segmentation operation on each text to be recommended according to a preset word segmentation tool to obtain a first basic keyword set; removing stop words from the first basic keyword set to obtain a first initial keyword set; obtaining the weight of each keyword in the first initial keyword set, and selecting keywords with weights greater than a first preset threshold to form the first keyword set according to the weight of each keyword in the first initial keyword set.

[0046] Optionally, the above-mentioned word segmentation tool can be a jieba Chinese word segmentation tool or a pangu word segmentation tool, and the present embodiment does not limit the word segmentation tool. Optionally, the full mode or precise mode or search engine mode in the jieba Chinese word segmentation tool can be used to perform word segmentation operations on each text to be recommended to obtain a first basic keyword set. After the text is segmented using the word segmentation tool, multiple keywords can be obtained to form the first basic keyword set. Since there may be words or words in the first basic keyword set that have little impact on the recommended text, a preset credit word list can be used to remove some keywords in the first basic keyword set to obtain a first initial keyword set with stop words removed; wherein, the stop word list can be the stop word list of Harbin Institute of Technology, the stop word list of Baidu, the stop word list of Sichuan University, or other stop word lists, and multiple stop word lists can also be integrated and deduplicated to obtain a new stop word list.

[0047] Optionally, all keywords in the first initial keyword set may be selected to form the first keyword set, or the weight of each keyword in the first initial keyword set may be obtained, and based on the weight of each keyword in the first initial keyword set, keywords with weights greater than a first preset threshold may be selected to form the first keyword set. i,j The weight tfidf(t i,j ) can be calculated according to the formula or contains The formula is determined by, where i represents the number of keywords, j represents the total number of keywords in the text to be recommended, and f(t i ,d m ) indicates the keyword t i In the recommended text m The number of times it appears in the text, m represents the number of the text to be recommended, N represents the total number of keywords in all the texts to be recommended and the texts to be matched, cfti Indicates the keyword t i The number of times it appears in all texts to be recommended and texts to be matched. It should be noted that the weight of each keyword in the first keyword set is determined based on the number of times the first keyword to be calculated appears in all texts to be recommended and texts to be matched, and the total number of keywords in all texts to be recommended and texts to be matched. It is calculated directly based on the keywords. Compared with the method of calculating the first weight vector based on the text in the prior art, the statistical granularity is finer. Therefore, the calculated weight of each keyword is more accurate.

[0048] S302, determining a first weight vector according to the number of times the first keyword to be calculated in the first keyword set appears in the text to be recommended, the number of times the first keyword to be calculated appears in each text to be recommended and the text to be matched, and the number of all keywords in each text to be recommended and the text to be matched.

[0049] Specifically, the weight of each keyword in the first keyword set may be determined in the same manner as that of determining the weight of the keywords in the first initial keyword set in S301 to form the first weight vector.

[0050] Optionally, the computer device may determine the first weight vector according to the following steps: determine the first term frequency (TF) of each keyword in the text to be recommended according to the number of times the first keyword to be calculated appears in the text to be recommended and the sum of the number of times all keywords appear in the text to be recommended; determine the first inverse document frequency index (IDF) of each keyword in the text to be recommended according to the number of times the first keyword to be calculated appears in each text to be recommended and the text to be matched and the number of all keywords in each text to be recommended and the text to be matched; determine the first weight vector according to the product of the first term frequency value and the first inverse document frequency index value. The first term frequency value is the formula in S301 In The first inverse text frequency index is In part, where f(t i ,d m ) indicates the keyword t i In the recommended text m The number of times it appears in , m represents the label of the text to be recommended, N represents the total number of keywords in all the texts to be recommended and the texts to be matched, Indicates the keyword t i The number of times it appears in all the texts to be recommended and matched.

[0051] S303, determining a first keyword vector according to the first keyword set and a preset target neural network model; the preset target neural network model is a model obtained by training an initial neural network according to a preset corpus.

[0052] Specifically, the above-mentioned initial neural network model can be a word2vec model, and each keyword in the first keyword set is input into the target neural network model to obtain a vector corresponding to each keyword, and the vectors corresponding to all keywords constitute the above-mentioned first keyword vector. Optionally, the computer device can also determine the first keyword vector according to the following steps: when training the initial neural network model, the initial neural network model is trained according to a preset corpus to obtain a target neural network model; the first keyword set is input into the target neural network model to obtain a first initial keyword vector; the first keyword vector is determined according to the product of the first initial keyword vector and the first weight vector.

[0053] Optionally, a corpus related to the text to be recommended and the text to be matched can be used for training to further improve the accuracy of resource recommendation. Among them, the training model can be the Skip-gram model or the continuous bag-of-words model (CBOW) in the word2vec model to train keywords to obtain the above-mentioned target neural network model. Furthermore, each keyword in the first keyword set can be input into the trained target neural model to obtain the vector corresponding to each keyword. The vectors corresponding to all keywords constitute the above-mentioned first initial keyword vector, and the above-mentioned first keyword vector is determined according to the product of the first initial keyword vector and the first weight vector. Optionally, the first keyword vector can be determined according to the product of the first initial keyword vector corresponding to each text to be recommended and the corresponding first weight vector. Optionally, the first keyword vector can be determined according to the formula v(c m )=∑ m v(d m )*tfidf(d m ) or contains ∑ m v(d m )*tfidf(d m ) is used to determine the first keyword vector, where v(d m ) represents the first initial keyword vector, v(c m ) indicates the text to be recommended d m The first keyword vector, tfidf(d m ) indicates the text to be recommended d m The first weight vector of , m represents the label of the text to be recommended.

[0054] S304: Determine a first text vector according to the first weight vector and the first keyword vector.

[0055] Specifically, the computer device may use a new vector formed by expanding the first weight vector and the first keyword vector as the first text vector, or may determine the first text vector according to the sum of the first weight vector and the first keyword vector, or may determine the first text vector according to the sum of the first weight vector and the first keyword vector. Figure 4 The first text vector is determined in the manner of the illustrated embodiment, and this embodiment is not limited to this.

[0056] In the resource recommendation method provided in this embodiment, the computer device can perform preprocessing operations on each text to be recommended to obtain the first keyword set of each text to be recommended; and determine the first weight vector according to the number of times the first keyword to be calculated in the first keyword set appears in the text to be recommended, the number of times the first keyword to be calculated appears in each text to be recommended and the text to be matched, and the number of all keywords in each text to be recommended and the text to be matched; and determine the first keyword vector according to the first keyword set and the preset target neural network model; and then determine the first text vector according to the first weight vector and the first keyword vector. Since the weights of each in the first weight vector are determined according to the number of times the first keyword to be calculated appears in all texts to be recommended and the text to be matched, and the total number of keywords in all texts to be recommended and the text to be matched, they are all calculated directly according to the keywords. Compared with the method of calculating the first weight vector according to the text in the prior art, the statistical granularity is finer. Therefore, the calculated first weight vector can better reflect the weights of each keyword in the text to be recommended, and then the first text vector determined according to the first weight vector and the first keyword vector can also more accurately express the text to be recommended, thereby improving the accuracy of describing the text to be recommended using the first text vector.

[0057] In the resource recommendation method provided in another embodiment, this embodiment relates to an implementation process in which a computer device determines a first text vector based on a first weight vector and a first keyword vector. Based on the above embodiment, optionally, the above S304 may include: determining the first text vector based on the product of the first weight vector and a preset first weight coefficient, and the product of the first keyword vector and a preset first word vector coefficient.

[0058] Specifically, the computer device may determine the first text vector by using the sum of the first product vector obtained by the first weight vector and the preset first weight coefficient and the second product vector obtained by the first keyword vector and the preset first vector coefficient. Alternatively, the first text vector may be determined in the following manner: according to the formula Determine the first text vector U(d m ), where α 1 represents the first word vector coefficient, v(c m ) indicates the text to be recommended dm The first keyword vector, tfidf(d m ) indicates the text to be recommended d m The first weight vector, β 1 represents the first weight coefficient, m represents the label of the text to be recommended, and Respectively represent the normalization of the first keyword vector and the first weight vector. Optionally, the above α 1 +β 1 =1,α 1 and β 1 The value of can be determined based on a preset number of experiments, and the α that can better describe the first text vector 1 and β 1 The value of is determined as the final value, optional, α 1 and β 1 The values ​​of can be 0.7 and 0.3 respectively. Furthermore, the above formula A new vector expanded from the first product vector and the second product vector is determined as the first text vector. When the cosine similarity between the first text vector and the second text vector is subsequently calculated, there is no need to separately calculate the cosine similarity between the first keyword vector in the first text vector and the second keyword vector in the second text vector, and the cosine similarity between the second weight vector in the first text vector and the second weight vector in the second text vector, thereby reducing the amount of calculation and improving the calculation speed.

[0059] In the resource recommendation method provided in this embodiment, a computer device can expand a new vector based on the product of a first weight vector and a preset first weight coefficient, and a first keyword vector and a preset first word vector coefficient, and use the expanded new vector as the first text vector, without separately calculating the cosine similarity between the first keyword vector in the first text vector and the second keyword vector in the second text vector, and the cosine similarity between the second weight vector in the first text vector and the second weight vector in the second text vector, thereby reducing the amount of calculation and improving the calculation speed.

[0060] Figure 4 A schematic diagram of a resource recommendation method flow chart provided for another embodiment. Based on the above embodiment, optionally, the above S202 may include:

[0061] S401, performing a preprocessing operation on the text to be matched to obtain a second keyword set of the text to be matched.

[0062] Among them, the preprocessing operation may include a word segmentation operation and a stop word removal operation. At this time, the above-mentioned preprocessing operation is performed on the matching text to obtain the first keyword set of the matching text, which may include: performing a word segmentation operation on the matching text according to a preset word segmentation tool to obtain a second basic keyword set; removing stop words from the second basic keyword set to obtain a second initial keyword set; obtaining the weight of each keyword in the second initial keyword set, and selecting keywords with weights greater than a second preset threshold to form a second keyword set based on the weights of each keyword in the second initial keyword set.

[0063] S402, determining a second weight vector according to the number of times the second keyword to be calculated in the second keyword set appears in the text to be matched, the number of times the second keyword to be calculated appears in each text to be recommended and the text to be matched, and the number of all keywords in each text to be recommended and the text to be matched.

[0064] Among them, the computer device can determine the second weight vector according to the following steps: determine the second word frequency value of each keyword in the text to be matched according to the number of times the second keyword to be calculated appears in the text to be matched, and the sum of the number of times all the keywords appear in the text to be matched; determine the second inverse text frequency index value of each keyword in the text to be recommended according to the number of times the second keyword to be calculated appears in each text to be recommended and the text to be matched, and the number of all keywords in each text to be recommended and the text to be matched; determine the second weight vector according to the product of the second word frequency value and the second inverse text frequency index value.

[0065] S403, determining a second keyword vector according to the second keyword set and a preset target neural network model; the preset target neural network model is a model obtained by training an initial neural network according to a preset corpus.

[0066] Among them, the computer device can train the initial neural network model according to the preset corpus to obtain the target neural network model; input the second keyword set into the target neural network model to obtain the second keyword vector.

[0067] S404: Determine a second text vector according to the second weight vector and the second keyword vector.

[0068] The computer device may determine the second text vector according to the product of the second weight vector and the preset second weight coefficient, and the product of the second keyword vector and the preset second word vector coefficient. Further, the computer device may determine the second text vector according to the formula Determine the second text vector U(d 匹配 ), where α 2 represents the second word vector coefficient, v(c 匹配) represents the first keyword vector of the text to be matched, tfidf(d 匹配 ) represents the first weight vector of the text to be matched, β 2 represents the first weight coefficient. Optional, α 2 Compared with α in the above embodiment 1 Can be equal or unequal, β 2 Compared with the β in the above embodiment 1 They may be equal or unequal, and this embodiment does not limit this. Optionally, the above α 2 +β 2 =1.

[0069] The working principle and technical effect of the resource recommendation method provided in this embodiment are as described above. Figure 3 The resource recommendation method provided in the illustrated embodiment is similar and will not be described in detail here.

[0070] It should be understood that although Figures 2 to 4 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figures 2 to 4 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0071] Figure 5 FIG. 1 is a schematic diagram of a resource recommendation device structure provided by an embodiment. Figure 5 As shown, the device may include a first text vector acquisition module 501 , a second text vector acquisition module 502 , and a target recommendation resource determination module 503 .

[0072] Specifically, the first text vector acquisition module 501 is used to acquire the first text vector corresponding to each to-be-recommended text, wherein the first text vector includes a first keyword vector and a first weight vector corresponding to the first keyword; the first keyword vector is used to represent the semantic features of the first keyword;

[0073] A second text vector acquisition module 502 is used to acquire a second text vector corresponding to the to-be-matched text, where the second text vector includes a second keyword vector and a second weight vector corresponding to the second keyword; the second keyword vector is used to represent the semantic features of the second keyword;

[0074] The target recommendation resource determination module 503 is used to match each first text vector with the second text vector according to the cosine similarity between the first text vector and the second text vector, and determine the first text vector with the highest cosine similarity with the second text vector from among the first text vectors as the matching target text vector.

[0075] The resource recommendation device provided in this embodiment can execute the above method embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.

[0076] Figure 6 In a resource recommendation device provided in another embodiment, in the above Figure 5 Based on the embodiment shown, Figure 6 As shown, the above device may further include: a first keyword set acquisition module 504 , a first weight vector determination module 505 , a first keyword vector determination module 506 and a first text vector determination module 507 .

[0077] Specifically, the first keyword set acquisition module 504 is used to perform a preprocessing operation on each text to be recommended to obtain a first keyword set for each text to be recommended.

[0078] The first weight vector determination module 505 is used to determine the first weight vector according to the number of times the first keyword to be calculated in the first keyword set appears in the text to be recommended, the number of times the first keyword to be calculated appears in each text to be recommended and the text to be matched, and the number of all keywords in each text to be recommended and the text to be matched.

[0079] The first keyword vector determination module 506 is used to determine the first keyword vector according to the first keyword set and a preset target neural network model; the preset target neural network model is a model obtained by training the initial neural network according to a preset corpus.

[0080] The first text vector determining module 507 is used to determine the first text vector according to the first weight vector and the first keyword vector.

[0081] The resource recommendation device provided in this embodiment can execute the above method embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.

[0082] In one embodiment, based on the above embodiment, the above-mentioned first text vector determination module 507 may include a first text vector determination unit, which is used to determine the first text vector based on the product of the first weight vector and a preset first weight coefficient, and the product of the first keyword vector and a preset first word vector coefficient.

[0083] Optionally, the first text vector determining unit is specifically configured to determine the text vector according to the formula Determine the first text vector U(d m ), where α 1 represents the first word vector coefficient, v(c m ) indicates the text to be recommended d m The first keyword vector, tfidf(d m ) indicates the text to be recommended d m The first weight vector, β 1 represents the first weight coefficient, and m represents the label of the text to be recommended.

[0084] The resource recommendation device provided in this embodiment can execute the above method embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.

[0085] In one embodiment, based on the above embodiment, the above preprocessing operation includes a word segmentation operation and a stop word removal operation, and the above first keyword set acquisition module 504 may include:

[0086] A first basic keyword set determining unit, configured to perform a word segmentation operation on each text to be recommended according to a preset word segmentation tool to obtain a first basic keyword set;

[0087] A first initial keyword set determining unit, configured to remove stop words from the first basic keyword set to obtain a first initial keyword set;

[0088] The first keyword set determination unit is used to obtain the weight of each keyword in the first initial keyword set, and select keywords with weights greater than a first preset threshold to form a first keyword set according to the weights of each keyword in the first initial keyword set.

[0089] The resource recommendation device provided in this embodiment can execute the above method embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.

[0090] In one embodiment, based on the above embodiment, the first weight vector determination module 505 may include:

[0091] A first word frequency value determining unit, used to determine a first word frequency value of each keyword in the text to be recommended according to the number of times the first keyword to be calculated appears in the text to be recommended and the sum of the number of times all keywords appear in the text to be recommended;

[0092] A first inverse text frequency index value determining unit, used to determine a first inverse text frequency index value of each keyword in the text to be recommended according to the number of times the first keyword to be calculated appears in each text to be recommended and the text to be matched, and the number of all keywords in each text to be recommended and the text to be matched;

[0093] The first weight vector determining unit is used to determine the first weight vector according to the product of the first word frequency value and the first inverse text frequency index value.

[0094] The resource recommendation device provided in this embodiment can execute the above method embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.

[0095] In one embodiment, based on the above embodiment, the first keyword vector determination module 506 may include:

[0096] A target neural network model determination unit is used to train the initial neural network model according to a preset corpus to obtain a target neural network model;

[0097] The first initial keyword vector determining unit is used to input the first keyword set into the target neural network model to obtain the first initial keyword vector.

[0098] The first keyword vector determining unit is used to determine the first keyword vector according to the product of the first initial keyword vector and the first weight vector.

[0099] The resource recommendation device provided in this embodiment can execute the above method embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.

[0100] Figure 7 A schematic diagram of the structure of a resource recommendation device provided in another embodiment. Based on the above embodiment, the above device may further include:

[0101] A second keyword set determination module 508 is used to perform a preprocessing operation on the text to be matched to obtain a second keyword set of the text to be matched;

[0102] A second weight vector determination module 509 is used to determine a second weight vector according to the number of times the second to-be-calculated keyword in the second keyword set appears in the to-be-matched text, the number of times the second to-be-calculated keyword appears in each to-be-recommended text and the to-be-matched text, and the number of all keywords in each to-be-recommended text and the to-be-matched text;

[0103] A second keyword vector determination module 510 is used to determine a second keyword vector according to a second keyword set and a preset target neural network model; the preset target neural network model is a model obtained by training an initial neural network according to a preset corpus;

[0104] The second text vector determining module 11 is used to determine the second text vector according to the second weight vector and the second keyword vector.

[0105] The resource recommendation device provided in this embodiment can execute the above method embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.

[0106] In one embodiment, based on the above embodiment, the above-mentioned second text vector determination module 11 includes a second text vector determination unit, which is used to determine the second text vector based on the product of the second weight vector and a preset second weight coefficient, and the product of the second keyword vector and a preset second word vector coefficient.

[0107] Optionally, the second text vector determination unit is specifically configured to determine the text vector according to the formula Determine the second text vector U(d 匹配 ), where α 2 represents the second word vector coefficient, v(c 匹配 ) represents the first keyword vector of the text to be matched, tfidf(d 匹配 ) represents the first weight vector of the text to be matched, β 2 Represents the first weight coefficient.

[0108] The resource recommendation device provided in this embodiment can execute the above method embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.

[0109] In one embodiment, based on the above embodiment, the second keyword set determination module 508 may include:

[0110] A second basic keyword set determination unit, configured to perform a word segmentation operation on the to-be-matched text according to a preset word segmentation tool to obtain a second basic keyword set;

[0111] A second initial keyword set determining unit, configured to remove stop words from the second basic keyword set to obtain a second initial keyword set;

[0112] The second keyword set determination unit is used to obtain the weight of each keyword in the second initial keyword set, and select keywords with weights greater than a second preset threshold to form a second keyword set according to the weights of each keyword in the second initial keyword set.

[0113] The resource recommendation device provided in this embodiment can execute the above method embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.

[0114] In one embodiment, based on the above embodiment, the second weight vector determination module 509 may include:

[0115] A second word frequency value determination unit, used to determine the second word frequency value of each keyword in the to-be-matched text according to the number of times the second to-be-calculated keyword appears in the to-be-matched text and the sum of the number of times all the keywords appear in the to-be-matched text;

[0116] A second inverse text frequency index value determining unit, used to determine the second inverse text frequency index value of each keyword in the text to be recommended according to the number of times the second keyword to be calculated appears in each text to be recommended and the text to be matched, and the number of all keywords in each text to be recommended and the text to be matched;

[0117] The second weight vector determining unit is used to determine the second weight vector according to the product of the second word frequency value and the second inverse text frequency index value.

[0118] The resource recommendation device provided in this embodiment can execute the above method embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.

[0119] In one embodiment, based on the above embodiment, the second keyword vector 510 may include:

[0120] A target neural network model determination unit is used to train the initial neural network model according to a preset corpus to obtain a target neural network model;

[0121] The second keyword vector determination unit is used to input the second keyword set into the target neural network model to obtain a second initial keyword vector.

[0122] The second keyword vector determining unit is configured to determine a second keyword vector according to the product of the second initial keyword vector and the second weight vector.

[0123] The resource recommendation device provided in this embodiment can execute the above method embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.

[0124] For the specific definition of the resource recommendation device, please refer to the definition of the resource recommendation method above, which will not be repeated here. Each module in the above resource recommendation device can be implemented in whole or in part by software, hardware and a combination thereof. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0125] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 1As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a resource recommendation method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0126] Those skilled in the art will understand that Figure 1 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0127] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0128] Obtaining a first text vector corresponding to each to-be-recommended text, the first text vector including a first keyword vector and a first weight vector corresponding to the first keyword; the first keyword vector is used to represent a semantic feature of the first keyword;

[0129] Obtaining a second text vector corresponding to the text to be matched, the second text vector including a second keyword vector and a second weight vector corresponding to the second keyword; the second keyword vector is used to represent the semantic features of the second keyword;

[0130] According to the cosine similarity between the first text vector and the second text vector, each first text vector is matched with the second text vector, and the first text vector having the highest cosine similarity with the second text vector among the first text vectors is determined as the matching target text vector.

[0131] The implementation principle and technical effect of the computer device provided in the above embodiment are similar to those of the above method embodiment, and will not be repeated here.

[0132] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0133] Obtaining a first text vector corresponding to each to-be-recommended text, the first text vector including a first keyword vector and a first weight vector corresponding to the first keyword; the first keyword vector is used to represent a semantic feature of the first keyword;

[0134] Obtaining a second text vector corresponding to the text to be matched, the second text vector including a second keyword vector and a second weight vector corresponding to the second keyword; the second keyword vector is used to represent the semantic features of the second keyword;

[0135] According to the cosine similarity between the first text vector and the second text vector, each first text vector is matched with the second text vector, and the first text vector with the highest cosine similarity with the second text vector is determined as the matching target text vector from among the first text vectors.

[0136] The computer-readable storage medium provided in the above embodiment has similar implementation principles and technical effects as those of the above method embodiment, and will not be described in detail here.

[0137] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0138] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0139] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A resource recommendation method, characterized in that: The method comprises: A first keyword set is obtained from each text to be recommended, and a first keyword vector is determined according to the first keyword set and a preset target neural network model. Determine the first text vector U(d m ), where α1 represents the first word vector coefficient, v(c m ) represents the text to be recommended d m The first keyword vector, tfidf(d m ) represents the text to be recommended d m The first weight vector corresponding to the first keyword in the text to be recommended is β1, β2 represents the first weight coefficient, and m represents the label of the text to be recommended; the preset target neural network model is a model obtained by training the initial neural network according to the preset corpus; the first keyword vector is used to characterize the semantic features of the first keyword; the first weight vector is determined based on the first inverse text frequency index value of each keyword in the text to be recommended, and the first inverse text frequency index value is determined according to the formula Determine, N represents the number of all keywords in each of the texts to be recommended and the texts to be matched, represents the first keyword to be calculated in the first keyword set t i The number of times the number appears in each of the text to be recommended and the text to be matched; A second keyword set is obtained from the text to be matched, and a second keyword vector is determined according to the second keyword set and a preset target neural network model. Determine the second text vector U(d 匹配 ), where α2 represents the second word vector coefficient, v(c 匹配 ) represents the text to be matched d 匹配 The second keyword vector, tfidf(d 匹配 ) represents the second weight vector corresponding to the second keyword in the text to be matched, β2 represents the second weight coefficient; the second keyword vector is used to characterize the semantic features of the second keyword; the second weight vector is determined based on the second inverse text frequency index value of each keyword in the text to be recommended, and the second inverse text frequency index value is determined according to the formula Sure, represents the second keyword to be calculated in the second keyword set t j The number of times the number appears in each of the text to be recommended and the text to be matched; According to the cosine similarity between the first text vector and the second text vector, each first text vector is matched with the second text vector, and the first text vector having the highest cosine similarity with the second text vector among the first text vectors is determined as the matching target text vector.

2. The method according to claim 1, characterized in that The process of constructing the first weight vector includes: The first weight vector is determined according to the number of times the first to-be-calculated keyword in the first keyword set appears in the to-be-recommended text, the sum of the number of times all keywords appear in the to-be-recommended text, and the first inverse text frequency index value.

3. The method according to claim 2, characterized in that The determining the first weight vector according to the number of times the first to-be-calculated keyword in the first keyword set appears in the to-be-recommended text, the sum of the number of times all keywords appear in the to-be-recommended text, and the first inverse text frequency index value includes: Determine a first word frequency value of each keyword in the text to be recommended according to the number of times the first keyword to be calculated appears in the text to be recommended and the sum of the number of times all keywords appear in the text to be recommended; The first weight vector is determined according to the product of the first word frequency value and the first inverse text frequency index value.

4. The method according to claim 1, characterized in that: The construction process of the second weight vector includes: The second weight vector is determined according to the number of times the second to-be-calculated keyword in the second keyword set appears in the to-be-matched text, the sum of the number of times all keywords appear in the to-be-matched text, and the second inverse text frequency index value.

5. The method according to claim 4, characterized in that The determining the second weight vector according to the number of times the second to-be-calculated keyword in the second keyword set appears in the to-be-matched text, the sum of the number of times all keywords appear in the to-be-matched text, and the second inverse text frequency index value comprises: Determine a second word frequency value of each keyword in the text to be matched according to the number of times the second keyword to be calculated appears in the text to be matched and the sum of the number of times all keywords appear in the text to be matched; The second weight vector is determined according to the product of the second word frequency value and the second inverse text frequency index value.

6. A resource recommendation device, characterized in that: The device comprises: The first text vector acquisition module is used to obtain a first keyword set from each text to be recommended, determine a first keyword vector according to the first keyword set and a preset target neural network model, and Determine the first text vector U(d m ), where α1 represents the first word vector coefficient, v(c m ) represents the text to be recommended d m The first keyword vector, tfidf(d m ) represents the text to be recommended d m The first weight vector corresponding to the first keyword in the text to be recommended is β1, β2 represents the first weight coefficient, and m represents the label of the text to be recommended; the preset target neural network model is a model obtained by training the initial neural network according to the preset corpus; the first keyword vector is used to characterize the semantic features of the first keyword; the first weight vector is determined based on the first inverse text frequency index value of each keyword in the text to be recommended, and the first inverse text frequency index value is determined according to the formula Determine, N represents the number of all keywords in each of the texts to be recommended and the texts to be matched, represents the first keyword to be calculated in the first keyword set t i The number of times the number appears in each of the text to be recommended and the text to be matched; The second text vector acquisition module is used to obtain a second keyword set from the to-be-matched text, determine a second keyword vector according to the second keyword set and a preset target neural network model, and Determine the second text vector U(d 匹配 ), where α2 represents the second word vector coefficient, v(c 匹配 ) represents the second keyword vector of the text to be matched, tfidf(d 匹配 ) represents the second weight vector corresponding to the second keyword in the text to be matched, β2 represents the second weight coefficient; the second keyword vector is used to characterize the semantic features of the second keyword; the second weight vector is determined based on the second inverse text frequency index value of each keyword in the text to be recommended, and the second inverse text frequency index value is determined according to the formula OK, cf tj represents the second keyword to be calculated in the second keyword set t j The number of times the number appears in each of the text to be recommended and the text to be matched; The target recommendation resource determination module is used to match each first text vector with the second text vector according to the cosine similarity between the first text vector and the second text vector, and determine the first text vector having the highest cosine similarity with the second text vector as the matching target text vector from among the first text vectors.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images