Search word association method based on unsupervised learning and related equipment
The unsupervised learning method calculates the similarity between search terms and target words in the open banking field, and solves the problems of high cost of manual word segmentation and inaccurate search results in the prior art, and achieves the effect of reducing costs and improving search accuracy.
Patent Information
- Application Number
- CN202411715252.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-27
- Publication Date
- 2025-03-18
AI Technical Summary
In the field of open banking, due to its professionalism, the existing technology requires a large number of manual word segmentation labels, which leads to high implementation costs and the inaccurate search keywords, resulting in inaccurate search results.
The search term association method based on unsupervised learning is adopted, and the search term vector is obtained using the Word2Vec method. The similarity between the target word vector and the search term vector is calculated through unsupervised learning, and alternative search terms are provided based on the similarity sorting.
No need to manually label participle segmentation, which reduces implementation costs and can accurately provide alternative search terms similar to search terms semantics, improving the accuracy of search results.
Smart Images
Figure CN119646273A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of machine learning, and in particular to a search term association method based on unsupervised learning and related equipment. Background Art
[0002] As an emerging platform commercialization model, open banking has broken through the limitations of traditional bank physical outlets and is directly embedded in people's daily life scenarios. Open banking mainly exports services for third-party use through API, SDK, H5 and other means. While empowering externally, it also builds its own open ecosystem. As the first stop for partners to connect with open banking, the portal website provides announcement information, open banking service introduction, open product information, cooperation cases, industry solutions, open interface description documents, development tool downloads, business applications, marketing advertising and other services for individual and corporate customers. When using the portal website, partner customers will inevitably search for information such as open banking connection process, interface classification, and implementation effects through the front-end search box. However, open banking is a highly professional field, and there is some information that partners do not understand and cannot summarize. For this information, it is also impossible to accurately extract search keywords, resulting in inaccurate search results and failure to meet user expectations.
[0003] In the prior art, supervised methods are used to train neural network models, and words with similar semantics to the search words are obtained for users to choose from based on the search words entered by users. Supervised methods for training neural network models require a large number of training sets of manually segmented words. In the field of open banking, due to its professionalism, it is time-consuming and laborious to label the segmented words, resulting in high implementation costs. Summary of the invention
[0004] In view of this, an embodiment of the present invention provides a search term association method based on unsupervised learning and related devices, so as to achieve the purpose of reducing costs without manually labeling segmented words.
[0005] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions:
[0006] A first aspect of an embodiment of the present invention discloses a search term association method based on unsupervised learning, the method comprising:
[0007] When receiving the search word input by the user, the search word vector is obtained using the Word2Vec method;
[0008] Calculate the similarity between each target word vector obtained by unsupervised learning and the search word vector; each target word vector corresponds to a target word;
[0009] Sort the target word vectors based on the similarity from large to small;
[0010] The target word corresponding to the target word vector sorted before the preset order is used as an alternative search word.
[0011] Preferably, the process of obtaining each target word vector through unsupervised learning includes:
[0012] Get target text related to Open Banking;
[0013] Performing word segmentation processing on the target text to obtain multiple target words;
[0014] For each target word, the Word2Vec method is used to obtain the corresponding skip-gram vector;
[0015] For each of the target words, external knowledge corresponding to the target word is obtained, and the external knowledge is supplemented into the skip-gram vector corresponding to the target word to obtain the target word vector.
[0016] Preferably, for each of the target words, obtaining external knowledge corresponding to the target word, and supplementing the external knowledge to the skip-gram vector corresponding to the target word to obtain the target word vector includes:
[0017] For each of the target words, obtain a BERT vector corresponding to the target word obtained through the BERT model;
[0018] Perform vector concatenation processing on the BERT vector and the skip-gram vector corresponding to the target word to obtain a target word vector.
[0019] Preferably, the calculation of obtaining the similarity between each target word vector and the search word vector through unsupervised learning includes:
[0020] For each target word vector obtained through unsupervised learning, the cosine function is used to calculate the similarity between the target word vector and the search word vector.
[0021] Preferably, the method further comprises:
[0022] The alternative search terms are displayed.
[0023] A second aspect of an embodiment of the present invention discloses a search term association device based on unsupervised learning, the device comprising:
[0024] The receiving unit, when receiving the search word input by the user, obtains the search word vector using the Word2Vec method;
[0025] A calculation unit, used to calculate the similarity between each target word vector obtained by unsupervised learning and the search word vector; each target word vector corresponds to a target word;
[0026] A sorting unit, used to sort the target word vectors from large to small based on the similarity;
[0027] A selection unit is used to select the target word corresponding to the target word vector sorted before the preset order as an alternative search word.
[0028] Preferably, the device further comprises:
[0029] A learning unit is used to obtain a target text related to open banking; perform word segmentation on the target text to obtain a plurality of target words; for each of the target words, use the Word2Vec method to obtain a corresponding skip-gram vector; for each of the target words, obtain external knowledge corresponding to the target word, supplement the external knowledge into the skip-gram vector corresponding to the target word, and obtain a target word vector.
[0030] Preferably, for each of the target words, obtaining the external knowledge corresponding to the target word, supplementing the external knowledge to the skip-gram vector corresponding to the target word, and obtaining the learning unit of the target word vector is specifically used for:
[0031] For each of the target words, obtain a BERT vector corresponding to the target word obtained through the BERT model;
[0032] Perform vector concatenation processing on the BERT vector and the skip-gram vector corresponding to the target word to obtain a target word vector.
[0033] A third aspect of an embodiment of the present invention discloses a storage medium for storing a computer program. When the computer program is executed, it is specifically used to implement any search term association method based on unsupervised learning as described in the first aspect of the embodiment of the present invention.
[0034] A fourth aspect of an embodiment of the present invention discloses an electronic device, including: a memory and a processor;
[0035] The memory is used to store computer programs;
[0036] The processor is used to execute the computer program, specifically to implement the search term association method based on unsupervised learning as described in any one of the first aspects of the embodiments of the present invention.
[0037] Based on the above-mentioned embodiment of the present invention, a search word association method based on unsupervised learning and related equipment are provided. When a search word input by a user is received, a search word vector is obtained using the Word2Vec method; the similarity between each target word vector obtained by unsupervised learning and the search word vector is calculated; each target word vector corresponds to a target word; the target word vectors are sorted from large to small based on the similarity; the target word corresponding to the target word vector sorted before the preset order is used as an alternative search word. In this solution, the search word is compared with the target word obtained by the unsupervised learning method, so as to provide the user with similar alternative search words. Unlike the supervised learning method, it does not require the use of labeled data sets for training, and there is no need to manually label the segmented words, thereby achieving the purpose of reducing costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
[0039] Figure 1 A flowchart of a search term association method based on unsupervised learning disclosed in an embodiment of the present invention;
[0040] Figure 2 A flowchart of obtaining each target word vector disclosed in an embodiment of the present invention;
[0041] Figure 3 A structural diagram of a search term association device based on unsupervised learning disclosed in an embodiment of the present invention;
[0042] Figure 4 The present invention is a structural diagram of an electronic device disclosed in an embodiment of the present invention. DETAILED DESCRIPTION
[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0044] In this application, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element.
[0045] First, the technical features appearing in this application are explained as follows:
[0046] Unsupervised Learning: Unsupervised learning is a machine learning technique in which the model is not supervised using a training dataset. Instead, the model itself finds hidden patterns and insights from the given data. It can be likened to the learning that occurs in the human brain when learning something new. The goal of unsupervised learning is to find the underlying structure of the dataset and cluster the data into groups based on similarities.
[0047] Word2Vec method: It is a method for converting words into vectors. The Word2Vec method mainly includes two vector modeling methods, namely the Continuous Bag-of-Word Model (CBOW) and the Skip-gram model. Two network models based on the same idea in different forms. The former is based on a feedforward neural network model, and its goal is to use the surrounding words within the current window size to predict the current word; the latter is opposite to the CBOW goal, and mainly uses the current word to predict the context representation within the window.
[0048] Word vector: Words or phrases from the vocabulary are mapped to vectors of real numbers. Conceptually, it involves mathematical embedding from a one-dimensional space for each word to a continuous vector space of other dimensions. In short, word vector technology converts words into dense vectors, and for similar words, their corresponding word vectors are also similar. Word vectors can be regarded as a specific representation of the semantic information of words.
[0049] BERT (Bidirectional Encoder Representations from Transformers): is a pre-trained language model based on the Transformer model, which learns the deep semantic representation of language through an unsupervised pre-training process.
[0050] BERT vector: refers to the vector representation obtained by encoding text using the BERT model. Specifically, the BERT vector is a high-dimensional numerical representation that can capture rich semantic information in the input text. In essence, it is to learn a good vector representation for words based on a massive corpus. In this patent, the feature representation of BERT can be directly used as the word vector for this task.
[0051] Chinese Word Segmentation: Segment a continuous sequence of Chinese characters into separate words according to certain specifications. In English text, words are separated by spaces as natural delimiters, while in Chinese, only characters, sentences, and paragraphs can be separated by obvious delimiters, but words do not have a formal delimiter. In layman's terms, Chinese word segmentation is to have a machine add marks between words in Chinese text.
[0052] Supervised Learning: Supervised learning is a type of machine learning in which the machine is trained using “labeled” training data and based on that data, the machine predicts the output. Labeled data means that some input data has been labeled with the correct output. In supervised learning, the training data given to the machine acts as a supervisor, teaching the machine to predict the output correctly. It applies the same concept as a student learning under the supervision of a teacher.
[0053] Neural network: Neural network is a computational model that simulates the human brain nervous system. It is composed of a large number of artificial neurons and realizes complex information processing and learning by simulating the connection and information transmission between neurons. Neural network has the characteristics of adaptability, nonlinearity and parallel processing, and is widely used in machine learning, pattern recognition, data mining and other fields.
[0054] Deep Learning: Deep learning is the process of learning the inherent laws and representation levels of sample data. The information obtained in the learning process is very helpful for interpreting data such as text, images, and sounds. Its ultimate goal is to enable machines to have analytical learning capabilities like humans and to recognize data such as text, images, and sounds. Deep learning is a complex machine learning algorithm that has achieved results in speech and image recognition that far exceed previous related technologies.
[0055] As can be seen from the background technology, in the prior art, supervised methods are used to train neural network models, and words with semantic similarity to the search words are obtained for users to choose from according to the search words entered by users. Supervised methods for training neural network models require a large number of training sets of manually segmented words. In the field of open banking, due to its professionalism, it is time-consuming and laborious to label the segmented words, resulting in high implementation costs.
[0056] Therefore, the embodiment of the present invention discloses a search word association method based on unsupervised learning and related equipment. When receiving a search word input by a user, a search word vector is obtained by using the Word2Vec method; the similarity between each target word vector obtained by unsupervised learning and the search word vector is calculated; each target word vector corresponds to a target word; the target word vectors are sorted from large to small based on the similarity; and the target word corresponding to the target word vector sorted before the preset order is used as an alternative search word. In this solution, the search word is compared with the target word obtained by the unsupervised learning method, so as to provide the user with similar alternative search words. Unlike the supervised learning method, it does not require the use of a labeled data set for training, and there is no need to manually label the segmented words, thereby achieving the purpose of reducing costs.
[0057] like Figure 1 FIG. 1 is a flowchart of a search term association method based on unsupervised learning disclosed in an embodiment of the present invention, which mainly includes the following steps:
[0058] Step S101: When receiving a search word input by a user, the search word vector is obtained using the Word2Vec method.
[0059] In step S101, a search box in the portal website of the open bank is used in combination with the search term input by the search user.
[0060] In the specific implementation process of step S101, the skip-gram model obtained by the vector modeling method in the Word2Vec method is used to obtain the skip-gram vector of the search word, and the skip-gram vector is used as the search word vector.
[0061] It should be noted that the specific process of converting the search word into the search word vector is the same as the process of converting the target word into the target word vector described below, and they can be referenced to each other.
[0062] Step S102: Calculate the similarity between each target word vector and the search word vector obtained through unsupervised learning.
[0063] Among them, each target word vector corresponds to a target word.
[0064] Specifically, the process of obtaining each target word vector through unsupervised learning can be performed in advance, such as Figure 2 As shown, it is a flowchart of obtaining each target word vector disclosed in an embodiment of the present invention, which mainly includes the following steps:
[0065] Step S201: Acquire target text related to open banking.
[0066] In step S201, a text introducing open banking is obtained on the Internet as a target text.
[0067] Step S202: performing word segmentation processing on the target text to obtain multiple target words.
[0068] In step S202, Chinese word segmentation is performed on the target text to obtain a word set consisting of multiple target words.
[0069] Step S203: For each target word, use the Word2Vec method to obtain the corresponding skip-gram vector.
[0070] In step S203, the skip-gram model obtained by the vector modeling method in the Word2Vec method is used to obtain the skip-gram vector of the target word, and the skip-gram vector is used as the target word vector.
[0071] Use X i ∈R k represents the i-th word in the text, and k is the vector dimension. Correspondingly, the text word vector with a length of n can be expressed as:
[0072] X={X1,X2,X3...X n}
[0073] Among them, the training process of the skip-gram model is to predict the possible context information through the current word.
[0074] The skip-gram model uses a sliding window approach to select context information for the current word, and maximizes the prediction probability of possible context words by restricting the current word.
[0075] Assuming that the current word is the target word mentioned above, for a target word sequence:
[0076] W={W1,W2,W3...Wn}
[0077] The skip-gram model defines the likelihood function as follows:
[0078]
[0079] Among them, k is the size of the sliding window, Pr(w i-k ,...,w i+k |w i ) represents the probability of predicting the context with the target word, which is obtained by the following softmax function:
[0080]
[0081] wc and w i and represent the vector representation of context information and target word respectively, A vector representation representing the contextual information of the target word, A vector representation of the context information of any target word.
[0082] In the Skip-gram model, the updating method of each parameter is based on the stochastic gradient descent algorithm.
[0083] Step S204: For each target word, obtain the external knowledge corresponding to the target word, and add the external knowledge to the skip-gram vector corresponding to the target word to obtain the target word vector.
[0084] In step S204, external knowledge is added to the skip-gram vector corresponding to the target word, the semantics of the target word is expanded from a single information source to multiple information sources, and the vector representation of the word is enriched.
[0085] The purpose of supplementing external knowledge is to expand text features and enrich the vector representation of words in the text. By integrating external knowledge with text data information, the expressive power of word vectors is enhanced, thus obtaining higher quality word vectors.
[0086] In the specific implementation process of step S204, first, for each target word, a BERT vector corresponding to the target word obtained by the BERT model is obtained.
[0087] Then, the BERT vector and the skip-gram vector corresponding to the target word are concatenated to obtain the target word vector.
[0088] It should be noted that the open banking field is a professional field. If the unsupervised method is used directly for modeling, the results obtained are not accurate enough. Therefore, the BERT vector of the word needs to be added as external knowledge to supplement the semantic information of the word. Then, the word vector with supplementary information can be used to quantitatively calculate the similarity between words, and finally sort them to obtain satisfactory results.
[0089] In the specific implementation process of step S102, for each target word vector obtained through unsupervised learning, the cosine function is used to calculate the similarity between the target word vector and the search word vector.
[0090] Specifically, the obtained search word vector and target word vector are used as inputs for word similarity calculation, and the similarity between the two is calculated using the cosine function, and this is used to represent the similarity between the words.
[0091] The similarity calculation formula is as follows:
[0092] Sim(X,Y)=cos(X,Y)
[0093] Among them, X represents the search word vector, Y represents the target word vector, cos represents the cosine function, and the cosine calculation is performed on the two vectors to represent the similarity of the words.
[0094] Step S103: sort the target word vectors from large to small based on similarity.
[0095] Step S104: taking the target word corresponding to the target word vector sorted before the preset order as the candidate search word.
[0096] Preferably, the target words corresponding to the target word vectors ranked in the top ten in similarity are used as candidate search words, and the candidate search words are displayed in the portal website search box for user selection.
[0097] Based on the above-mentioned search term association method based on unsupervised learning disclosed in the embodiment of the present invention, when receiving the search term input by the user, the search term vector is obtained by using the Word2Vec method; the similarity between each target term vector obtained by unsupervised learning and the search term vector is calculated; a target word corresponding to each target term vector; each target term vector is sorted from large to small based on the similarity; the target word corresponding to the target term vector sorted before the preset order is used as an alternative search term. In this scheme, the search term and the open bank text are modeled, and multiple words that are most similar to the search term at the semantic level are calculated, and they are displayed in the search box as alternative search terms. The purpose of external knowledge is to expand text features and enrich the vector representation of words in the text. The external knowledge is integrated with the target word information to enhance the expression ability of the word vector, so as to obtain a higher quality word vector. In this scheme, the unsupervised method is used, which is different from the supervised method. It does not require training with a labeled data set. The corresponding word results can be directly output from the input text, which is very convenient to use and saves manpower and material resources. Moreover, under the guidance and supplement of external knowledge, the proposed unsupervised method can better understand the semantic information of words in the field of open banking, so as to more accurately calculate and obtain alternative search terms.
[0098] Corresponding to the search term association method based on unsupervised learning disclosed in the above embodiment of the present invention, Figure 3 As shown, it is a structural diagram of a search word association device based on unsupervised learning disclosed in an embodiment of the present invention, including: a receiving unit 301, a calculating unit 302, a sorting unit 303 and a selecting unit 304.
[0099] The receiving unit 301, when receiving a search word input by a user, obtains a search word vector using the Word2Vec method.
[0100] The calculation unit 302 is used to calculate the similarity between each target word vector and the search word vector obtained through unsupervised learning; each target word vector corresponds to a target word.
[0101] The sorting unit 303 is used to sort the target word vectors from large to small based on similarity.
[0102] The selection unit 304 is used to select the target word corresponding to the target word vector sorted before the preset order as the candidate search word.
[0103] In one embodiment, the device further comprises:
[0104] The learning unit is used to obtain target text related to open banking; perform word segmentation on the target text to obtain multiple target words; for each target word, use the Word2Vec method to obtain the corresponding skip-gram vector; for each target word, obtain the external knowledge corresponding to the target word, supplement the external knowledge into the skip-gram vector corresponding to the target word, and obtain the target word vector.
[0105] In one embodiment, for each target word, obtaining external knowledge corresponding to the target word, adding the external knowledge to the skip-gram vector corresponding to the target word, and obtaining a learning unit of the target word vector is specifically used for:
[0106] For each target word, obtain the BERT vector corresponding to the target word obtained through the BERT model; perform vector concatenation processing on the BERT vector and the skip-gram vector corresponding to the target word to obtain the target word vector.
[0107] In one embodiment, the calculation unit 302 is specifically configured to:
[0108] For each target word vector obtained through unsupervised learning, the cosine function is used to calculate the similarity between the target word vector and the search word vector.
[0109] In one embodiment, the device further comprises:
[0110] A display unit is used to display alternative search terms.
[0111] Based on the above-mentioned embodiment of the present invention, a search term association device based on unsupervised learning is disclosed. When receiving the search term input by the user, the Word2Vec method is used to obtain the search term vector; the similarity between each target term vector obtained by unsupervised learning and the search term vector is calculated; each target term vector corresponds to a target word; each target term vector is sorted from large to small based on the similarity; the target word corresponding to the target term vector sorted before the preset order is used as an alternative search term. In this scheme, the search term and the open bank text are modeled, and multiple words that are most similar to the search term at the semantic level are calculated, and they are displayed in the search box as alternative search terms. The purpose of external knowledge is to expand the text features and enrich the vector representation of words in the text. The external knowledge is integrated with the target word information to enhance the expression ability of the word vector, so as to obtain a higher quality word vector. In this scheme, the unsupervised method is used, which is different from the supervised method. It does not require training with a labeled data set. The corresponding word results can be directly output from the input text, which is very convenient to use and saves manpower and material resources. Moreover, under the guidance and supplement of external knowledge, the proposed unsupervised method can better understand the semantic information of words in the field of open banking, so that it can more accurately calculate and obtain alternative search terms.
[0112] The present invention also provides an electronic device, please refer to Figure 4 , the electronic device includes a memory 401 and a processor 402.
[0113] Wherein, the memory 401 is used to store computer programs;
[0114] The processor 402 is used to execute a computer program, specifically to implement the search term association method based on unsupervised learning provided in any embodiment of the present application.
[0115] The present application also provides a computer storage medium for storing a computer program. When the computer program is executed, it is specifically used to implement the search term association method based on unsupervised learning provided in any embodiment of the present application.
[0116] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without creative work.
[0117] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0118] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A search term association method based on unsupervised learning, characterized in that: The method comprises: When receiving the search word input by the user, the search word vector is obtained using the Word2Vec method; Calculate the similarity between each target word vector obtained by unsupervised learning and the search word vector; each target word vector corresponds to a target word; Sort the target word vectors based on the similarity from large to small; The target word corresponding to the target word vector sorted before the preset order is used as an alternative search word.
2. The method according to claim 1, characterized in that The process of obtaining each target word vector through unsupervised learning includes: Get target text related to Open Banking; Performing word segmentation processing on the target text to obtain multiple target words; For each target word, the Word2Vec method is used to obtain the corresponding skip-gram vector; For each of the target words, external knowledge corresponding to the target word is obtained, and the external knowledge is supplemented into the skip-gram vector corresponding to the target word to obtain the target word vector.
3. The method according to claim 2, characterized in that For each of the target words, acquiring external knowledge corresponding to the target word, and supplementing the external knowledge to the skip-gram vector corresponding to the target word to obtain the target word vector, including: For each of the target words, obtain a BERT vector corresponding to the target word obtained through the BERT model; Perform vector concatenation processing on the BERT vector and the skip-gram vector corresponding to the target word to obtain a target word vector.
4. The method according to claim 1, characterized in that: The calculation obtains the similarity between each target word vector and the search word vector through unsupervised learning, including: For each target word vector obtained through unsupervised learning, the cosine function is used to calculate the similarity between the target word vector and the search word vector.
5. The method according to any one of claims 1 to 4, characterized in that: The method further comprises: The alternative search terms are displayed.
6. A search term association device based on unsupervised learning, characterized in that: The device comprises: The receiving unit, when receiving the search word input by the user, obtains the search word vector using the Word2Vec method; A calculation unit, used to calculate the similarity between each target word vector obtained by unsupervised learning and the search word vector; each target word vector corresponds to a target word; A sorting unit, used to sort the target word vectors from large to small based on the similarity; A selection unit is used to select the target word corresponding to the target word vector sorted before the preset order as an alternative search word.
7. The device according to claim 6, characterized in that The device also includes: A learning unit is used to obtain a target text related to open banking; perform word segmentation on the target text to obtain a plurality of target words; for each of the target words, use the Word2Vec method to obtain a corresponding skip-gram vector; for each of the target words, obtain external knowledge corresponding to the target word, supplement the external knowledge into the skip-gram vector corresponding to the target word, and obtain a target word vector.
8. The device according to claim 7, characterized in that The learning unit for acquiring, for each target word, external knowledge corresponding to the target word, and adding the external knowledge to the skip-gram vector corresponding to the target word to obtain the target word vector is specifically used for: For each of the target words, obtain a BERT vector corresponding to the target word obtained through the BERT model; Perform vector concatenation processing on the BERT vector and the skip-gram vector corresponding to the target word to obtain a target word vector.
9. A storage medium, characterized in that: Used to store a computer program, which, when executed, is specifically used to implement the search term association method based on unsupervised learning as described in any one of claims 1 to 5.
10. An electronic device, characterized in that: include: Memory and processor; The memory is used to store computer programs; The processor is used to execute the computer program, specifically to implement the search term association method based on unsupervised learning as described in any one of claims 1 to 5.