Search prompt method, device, computer equipment and storage medium

Through clustering and deduplication technology, the problem of poor diversity of search prompt words is solved, and the search efficiency and user experience are improved.

CN114328905BActive Publication Date: 2024-12-03TENCENT TECHNOLOGY (SHENZHEN) CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202111056368.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-09-09
Publication Date
2024-12-03
Estimated Expiration
2041-09-09

AI Technical Summary

Technical Problem

There are a large number of vocabulary in the search prompt word list with the same meaning but different text expression forms, which leads to the poor diversity of search prompt words, occupying valuable exposure positions, and affecting search efficiency.

Method used

By clustering the search prompt text, a similarity map is generated, and dividing it into subclusters, deduplication of the text of the same subcluster during searching, improving the diversity of search terms.

Benefits of technology

It effectively improves the diversity of search prompt texts hitting search terms during the search process, saves user input time, and improves search efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114328905B_ABST
    Figure CN114328905B_ABST
Patent Text Reader

Abstract

The present application relates to a search prompt method, device, computer equipment and storage medium. The method includes: clustering a set of search prompt texts according to the search entity dimension to obtain an intermediate cluster; abstracting each search prompt text in the intermediate cluster into a node, and generating a similarity graph corresponding to each intermediate cluster according to the similarity between the node and the search prompt text corresponding to each node; clustering the nodes in the similarity graph to divide the intermediate cluster into subclusters; when performing search prompts on search terms, deduplicating the search prompt texts that are hit by the search term and belong to the same subcluster. The use of this method can improve the diversity of search prompt terms.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of text information processing, and in particular to a search prompt method, device, computer equipment and storage medium. Background Art

[0002] Search hints are one of the essential functions of search engines. Search hints can give search suggestions based on the search terms entered by users. That is, search hints related to the search terms entered by users are listed on the search page. Users can quickly find the target search terms from the search hints to quickly search for the target object corresponding to the target search terms, thereby saving users' search input time and improving search efficiency.

[0003] However, there are often a large number of search prompt words with the same meaning but different textual expressions in the search prompt word list. These search prompt words with the same meaning occupy a large number of valuable exposure positions, thereby reducing the diversity of the search prompt words. Summary of the invention

[0004] Based on this, it is necessary to provide a search prompt method, device, computer equipment and storage medium that can improve the diversity of search prompt words in response to the above technical problems.

[0005] A search prompt method, the method comprising:

[0006] Cluster the collection of search prompt texts according to the search entity dimension to obtain the intermediate cluster;

[0007] Abstracting each of the search prompt texts in the intermediate cluster into a node, and generating a similarity graph corresponding to each of the intermediate clusters according to the similarity between the node and the search prompt text corresponding to each of the nodes;

[0008] Clustering the nodes in the similarity graph to divide the intermediate cluster into subclusters;

[0009] When performing search prompts for search terms, duplicate search prompt texts that are matched by the search terms and belong to the same sub-cluster are removed.

[0010] A search prompt device, the device comprising:

[0011] A clustering module is used to cluster the set of search prompt texts according to the search entity dimension to obtain intermediate clusters;

[0012] A generating module, used for abstracting each of the search prompt texts in the intermediate cluster into a node, and generating a similarity graph corresponding to each of the intermediate clusters according to the similarity between the node and the search prompt text corresponding to each of the nodes;

[0013] The clustering module is also used to cluster the nodes in the similarity graph to divide the intermediate cluster into sub-clusters;

[0014] The deduplication module is used to deduplicate the search prompt texts that are hit by the search term and belong to the same sub-cluster when performing search prompts for the search term.

[0015] In one embodiment, the search entity dimension includes a first high-frequency search entity dimension and a second high-frequency search entity dimension; the clustering module is also used to cluster the set of search prompt texts according to the first high-frequency search entity dimension to obtain an initial cluster; and cluster each search prompt text in the initial cluster according to the second high-frequency search entity dimension to divide the initial cluster into intermediate clusters.

[0016] In one embodiment, the apparatus further comprises:

[0017] A determination module is used to determine common characters from characters corresponding to non-entity information contained in two search prompt texts corresponding to any two nodes belonging to the same intermediate cluster; determine the similarity between the two search prompt texts based on a similarity coefficient between the two search prompt texts and the sum of importance coefficients corresponding to each of the common characters; the importance coefficient is used to characterize the importance of the common character in the intermediate cluster.

[0018] In one embodiment, the determination module is also used to determine the adjustment coefficient based on the similarity between the attribute information in the two search prompt texts if both of the two search prompt texts include corresponding attribute information under the key prompt attribute; and determine the similarity between the two search prompt texts based on the similarity coefficient between the two search prompt texts, the sum of the importance coefficients corresponding to each of the common characters, and the adjustment coefficient.

[0019] In one embodiment, the clustering module is also used to select the starting node of this round of traversal from the nodes that have not been visited in the similarity graph, and start traversing and visiting the nodes in the similarity graph from the starting node in this round; when the end condition of this round of visit is met, the nodes visited in this round are divided into a sub-cluster corresponding to the intermediate cluster; the next round is taken as this round, and the starting node of this round of traversal is selected from the nodes that have not been visited in the similarity graph to continue iterative execution, until all the nodes in the similarity graph are visited and the iteration is stopped to obtain the sub-clusters into which the intermediate cluster is divided.

[0020] In one embodiment, the clustering module is also used to obtain a node sequence set; the node sequence set is obtained by sorting the identifiers of unvisited nodes in the similarity graph according to the average degree of the nodes; selecting an identifier corresponding to the maximum average degree from the node sequence set, and using the node represented by the selected identifier as the starting node of this round; removing the identifiers of the nodes visited in this round from the node sequence set to obtain an updated node sequence set, and using the next round as this round, selecting the identifier corresponding to the maximum average degree from the updated node sequence set to continue iterative execution.

[0021] In one embodiment, the clustering module is also used to traverse and visit the nodes in the similarity graph layer by layer starting from the starting node in this round; when the number of traversed layers reaches a preset traversal layer threshold, the starting node visited in this round and the nodes connected to the starting node are divided into a sub-cluster corresponding to the intermediate cluster.

[0022] In one embodiment, the deduplication module is also used to use each search prompt text in the same subcluster as a key and the search prompt text of the cluster center belonging to the subcluster as a value, and store each search prompt text in the subcluster in the form of a key-value pair; when performing a search prompt for a search term, use each search prompt text that hits the search term as a key to search for a corresponding value from a storage space; replace the search prompt text that hits the search term and corresponds to the same value with the search prompt text that belongs to the found value to perform deduplication.

[0023] In one embodiment, the apparatus further comprises:

[0024] The preprocessing module is used to perform at least one of the following preprocessing on the initial search prompt text to obtain a set of the search prompt text: removing non-key information from the initial search prompt text; correcting errors from the initial search prompt text; and standardizing the expression of the initial search prompt text.

[0025] In one embodiment, the non-critical information includes stop words; the preprocessing module is also used to obtain a preset stop word list; determine the words that hit the stop words in the stop word list from the initial search prompt text; and remove the words from the initial search prompt text.

[0026] In one embodiment, the preprocessing module is also used to determine the words that hit the synonyms in the synonym table from the initial search prompt text to obtain the words to be standardized; normalize the words to be standardized into corresponding standard words; and / or, for words with the same attribute and different expressions in the initial search prompt text, normalize the words belonging to the first expression to the words of the second expression; the usage frequency of the second expression is higher than the usage frequency of the first expression.

[0027] In one embodiment, the search prompt text is a video search prompt text; the search term is a video search term; the deduplication module is also used to deduplicate the video search prompt text that hits the video search term and belongs to the same subcluster when performing a video search prompt for the video search term.

[0028] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0029] Cluster the collection of search prompt texts according to the search entity dimension to obtain the intermediate cluster;

[0030] Abstracting each of the search prompt texts in the intermediate cluster into a node, and generating a similarity graph corresponding to each of the intermediate clusters according to the similarity between the node and the search prompt text corresponding to each of the nodes;

[0031] Clustering the nodes in the similarity graph to divide the intermediate cluster into subclusters;

[0032] When performing search prompts for search terms, duplicate search prompt texts that are matched by the search terms and belong to the same sub-cluster are removed.

[0033] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the following steps:

[0034] Cluster the collection of search prompt texts according to the search entity dimension to obtain the intermediate cluster;

[0035] Abstracting each of the search prompt texts in the intermediate cluster into a node, and generating a similarity graph corresponding to each of the intermediate clusters according to the similarity between the node and the search prompt text corresponding to each of the nodes;

[0036] Clustering the nodes in the similarity graph to divide the intermediate cluster into subclusters;

[0037] When performing search prompts for search terms, duplicate search prompt texts that are matched by the search terms and belong to the same sub-cluster are removed.

[0038] A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the following steps are implemented:

[0039] Cluster the collection of search prompt texts according to the search entity dimension to obtain the intermediate cluster;

[0040] Abstracting each of the search prompt texts in the intermediate cluster into a node, and generating a similarity graph corresponding to each of the intermediate clusters according to the similarity between the node and the search prompt text corresponding to each of the nodes;

[0041] Clustering the nodes in the similarity graph to divide the intermediate cluster into subclusters;

[0042] When performing search prompts for search terms, duplicate search prompt texts that are matched by the search terms and belong to the same sub-cluster are removed.

[0043] The above-mentioned search prompt method, device, computer equipment and storage medium cluster the set of search prompt texts according to the search entity dimension to obtain intermediate clusters to perform preliminary cluster division of the search prompt texts, which can avoid dividing the search prompt texts under different search entity dimensions into the same cluster during subsequent clustering. Each search prompt text in the intermediate cluster is abstracted as a node, and a similarity graph corresponding to each intermediate cluster is generated based on the similarity between the node and the search prompt text corresponding to each node, so as to connect the nodes with higher similarity. The nodes in the similarity graph are clustered to divide the intermediate clusters into sub-clusters, so that the nodes with higher similarity can be divided into the same cluster. When performing search prompts for search terms, the search prompt texts that are hit by the search terms and belong to the same sub-cluster are deduplicated, thereby improving the diversity of the search prompt texts that are hit by the search terms during the search process. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 An application environment diagram of a search prompt method in an embodiment;

[0045] Figure 2 A schematic diagram of a flow chart of a search prompt method in one embodiment;

[0046] Figure 3 is a schematic diagram of a similarity graph in one embodiment;

[0047] Figure 4 A schematic diagram of clustering nodes in a similarity graph to obtain subclusters corresponding to an intermediate cluster in an embodiment;

[0048] Figure 5 A schematic diagram of a flow chart of a search prompt method in another embodiment;

[0049] Figure 6This is a schematic diagram of displaying the search prompt text corresponding to the search word before the search prompt method of the present application is used to remove duplicates of the search prompt word in an embodiment;

[0050] Figure 7 This is a schematic diagram of displaying search prompt texts corresponding to search words after the search prompt method of the present application is used to remove duplicate search prompt words in one embodiment;

[0051] Figure 8 It is a structural block diagram of a search prompt device in one embodiment;

[0052] Fig. 9 It is a structural block diagram of a search prompt device in another embodiment;

[0053] Fig.10 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0054] 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.

[0055] The search prompt method provided by this application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. Among them, the terminal 102 can be but not limited to various personal computers, laptops, smart phones, tablet computers and portable wearable devices, and the server 104 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. The terminal 102 and the server 104 can be directly or indirectly connected through wired or wireless communication, and this application does not limit this.

[0056] The server 104 may obtain a set of search prompt texts, and cluster the set of search prompt texts according to the search entity dimension to obtain an intermediate cluster. The server 104 may abstract each search prompt text in the intermediate cluster into a node, and generate a similarity graph corresponding to each intermediate cluster based on the similarity between the node and the search prompt text corresponding to each node. The server 104 may cluster the nodes in the similarity graph to divide the intermediate cluster into subclusters. When the terminal 102 performs a search prompt for a search term, the server 104 may remove duplicate search prompt texts that are hit by the search term and belong to the same subcluster.

[0057] In one embodiment, the server 104 may pre-acquire a set of search prompt texts stored in the server 104, and pre-divide the set of search prompt texts into subclusters through the above steps. Then, when the terminal 102 sends a search term to the server 104 to make a search request, the server 104 may deduplicate the search prompt texts that are hit by the search term and belong to the same subcluster. In another embodiment, the terminal 102 may also send a search term to the server 104 to make a search request, and the server 104 may determine the set of search prompt texts that are hit by the search term in response to the search request, and cluster the set of search prompt texts according to the search entity dimension to obtain an intermediate cluster, and perform subsequent steps to deduplicate the search prompt texts that are hit by the search term and belong to the same subcluster.

[0058] In one embodiment, Figure 2 As shown, a search prompt method is provided, which can be applied to the server or to the interaction process between the terminal and the server. Figure 1 Taking the server 104 in the example as an example, the following steps are included:

[0059] Step 202: cluster the set of search prompt texts according to the search entity dimension to obtain an intermediate cluster.

[0060] The search hint text is the text used for the hint during the search. The set of search hint texts is a set consisting of search hint texts. The search entity dimension is the dimension corresponding to the entity information in the search hint text. The entity information is the information with specific meaning in the search hint text. The intermediate cluster is a cluster in an intermediate state. It can be understood that the server can also perform more fine-grained cluster division on the intermediate cluster.

[0061] Specifically, the server may obtain a set of search prompt texts to be aggregated, and cluster the set of search prompt texts according to the search entity dimension to obtain an intermediate cluster.

[0062] It should be noted that the number of search hint texts included in each intermediate cluster may be different, that is, some intermediate clusters may include multiple search hint texts, and some intermediate clusters may include only one search hint text.

[0063] In one embodiment, the entity information includes at least one of a person's name, a place name, an organization name, and a proper noun, etc. A proper noun is a name used exclusively in a specific field. For example, in the video field, a proper noun may include at least one of a video name, a TV series name, a movie name, and a variety show name, etc.

[0064] It can be understood that if the entity information is a video name, the search entity dimension is the dimension corresponding to the video name. If the entity information is a person's name, the search entity dimension is the dimension corresponding to the person's name.

[0065] In one embodiment, the server may obtain the initial search prompt text from the user's historical search log, and preprocess the initial search prompt text to obtain a set of search prompt texts. It can be understood that the set of search prompt texts has higher accuracy, uniformity, and lower redundancy than the initial search prompt text. The initial search prompt text is the original search prompt text.

[0066] Step 204: abstract each search prompt text in the intermediate cluster into a node, and generate a similarity graph corresponding to each intermediate cluster according to the similarity between the node and the search prompt text corresponding to each node.

[0067] The similarity is the degree of similarity between two search prompt texts. The similarity graph is an image used to represent the similarity between the search prompt texts in the corresponding intermediate cluster.

[0068] Specifically, the server may abstract each search prompt text in the intermediate cluster into a node. Further, the server may generate a similarity graph corresponding to each intermediate cluster according to the similarity between the node and the search prompt text corresponding to each node. It can be understood that one search prompt text may be abstracted into one node.

[0069] In one embodiment, the server may first abstract each search prompt text into a node, and then calculate the similarity between each node. It can be understood that the similarity between each node is the similarity between the search prompt texts corresponding to each node. Furthermore, the server may generate a similarity graph corresponding to each intermediate cluster based on the similarity between the node and the search prompt text corresponding to each node.

[0070] In one embodiment, the server may first calculate the similarity between each search prompt text in the intermediate cluster, then abstract each search prompt text in the intermediate cluster into a node, and generate a similarity graph corresponding to each intermediate cluster based on the similarity between the node and the search prompt text corresponding to each node.

[0071] In one embodiment, for two search prompt texts corresponding to any two nodes belonging to the same intermediate cluster, the server may determine a similarity coefficient between the two search prompt texts. Furthermore, the server may directly use the similarity coefficient between the two search prompt texts as the similarity between the two search prompt texts. The similarity coefficient is a coefficient used to characterize the degree of similarity between the two search prompt texts.

[0072] In one embodiment, for two search hint texts corresponding to any two nodes belonging to the same intermediate cluster, the server may determine from the two search hint texts the key attributes that have a greater impact on the accuracy of the similarity between the two search hint texts, and calculate the impact coefficient corresponding to the key attribute. Furthermore, the server may determine the similarity between the two search hint texts based on the similarity coefficient between the two search hint texts and the impact coefficient corresponding to the key attribute.

[0073] In one embodiment, the server may compare the similarity between the nodes with a preset similarity threshold, and connect the nodes whose similarity is greater than the preset similarity threshold to generate a similarity graph corresponding to each intermediate cluster.

[0074] For example, Figure 3 As shown, the intermediate cluster includes node AL. The similarity between node B and node D is greater than the preset similarity threshold, and the server can connect node B to node D. The similarity between node B and node C is also greater than the preset similarity threshold, and the server can also connect node B to node C. The similarity between node C and node F is also greater than the preset similarity threshold, and the server can also connect node C to node F, and so on, until all nodes in the intermediate cluster whose similarity is greater than the preset similarity threshold are connected, and the similarity graph corresponding to the intermediate cluster is generated.

[0075] Step 206: cluster the nodes in the similarity graph to divide the middle cluster into sub-clusters.

[0076] Among them, the subcluster is a cluster in a final state. It can be understood that the server will not perform more fine-grained cluster division on the subcluster.

[0077] Specifically, the server may perform finer-grained clustering on the nodes in the similarity graph to divide the intermediate cluster into subclusters. It should be noted that the number of search prompt texts included in each subcluster may be different, that is, some subclusters may include multiple search prompt texts, and some subclusters may include only one search prompt text.

[0078] In one embodiment, the server may traverse the similarity graph based on a target traversal algorithm to divide the intermediate cluster into subclusters. The target traversal algorithm is a traversal algorithm as a target. In one embodiment, the target traversal algorithm may include any one of a breadth-first traversal algorithm and a depth-first traversal algorithm.

[0079] In one embodiment, the server may directly use the interconnected nodes in the similarity graph as a sub-cluster corresponding to the intermediate cluster.

[0080] In one embodiment, if the middle cluster contains only one search hint text, the server may directly treat the middle cluster as a sub-cluster.

[0081] Step 208: when performing search prompts for the search terms, de-duplicate the search prompt texts that match the search terms and belong to the same sub-cluster.

[0082] The search term is the text input by the user during the search, and the search hint text that matches the search term is the search hint text that has a mapping relationship with the search term.

[0083] Specifically, the terminal may obtain the search term input by the user based on the search page of the terminal, and send the search term to the server. The server may receive the search term sent by the terminal. When performing a search prompt for the search term, the server may remove duplicate search prompt texts that are hit by the search term and belong to the same subcluster. It can be understood that all the search prompt texts after removal of duplicates correspond to different subclusters.

[0084] In one embodiment, the server may determine the cluster center of each sub-cluster, and deduplicate the search hint text that matches the search term based on the search hint text corresponding to the cluster center of each sub-cluster.

[0085] In one embodiment, the server may search for search hint texts belonging to the same subcluster from among the search hint texts hit by the search term. Further, the server may remove the search hint texts belonging to the same subcluster, so that only a preset number of search hint texts are retained for the same subcluster, for example, only one search hint text is retained for the same subcluster, so as to achieve deduplication of the search hint texts hit by the search term and belonging to the same subcluster.

[0086] In one embodiment, the search hint text for a search term hit can be obtained specifically through the following steps: the server can calculate the similarity between the search term and each search hint text, and use the search hint text corresponding to the similarity greater than a preset threshold as the search hint text for the search term hit.

[0087] In one embodiment, the server may sort the search prompt texts that have been deduplicated and hit the search terms. The server may send the deduplicated and sorted search prompt texts that have hit the search terms to the terminal. The terminal may display the search prompt texts that have hit the search terms on the search page.

[0088] In one embodiment, the server may sort the search prompt texts that are hit by the search term according to the similarity between the search term and each search prompt text. It should be noted that for multiple search prompt texts belonging to the same sub-cluster, the server may only retain the search prompt text with the highest sorting for deduplication. The server may send the deduplicated and sorted search prompt texts and a preset number of search prompt texts that are hit by the search term to the terminal. Furthermore, the terminal may display the received search prompt text on the search page.

[0089] In the above search prompt method, the set of search prompt texts is clustered according to the search entity dimension to obtain the intermediate clusters, so as to perform preliminary cluster division on the search prompt texts, which can avoid the search prompt texts under different search entity dimensions from being divided into the same cluster in subsequent clustering. Each search prompt text in the intermediate cluster is abstracted as a node, and a similarity graph corresponding to each intermediate cluster is generated based on the similarity between the node and the search prompt text corresponding to each node, so as to connect the nodes with higher similarity. The nodes in the similarity graph are clustered to divide the intermediate clusters into subclusters, so that the nodes with higher similarity can be divided into the same cluster. When performing search prompts on search terms, the search prompt texts that are hit by the search terms and belong to the same subcluster are deduplicated, which improves the diversity of the search prompt texts that are hit by the search terms during the search process.

[0090] In one embodiment, the search entity dimension includes a high-frequency search entity dimension; the high-frequency search entity dimension includes a first high-frequency search entity dimension and a second high-frequency search entity dimension. Step 202, i.e., clustering the set of search prompt texts according to the search entity dimension to obtain intermediate clusters, specifically includes: clustering the set of search prompt texts according to the first high-frequency search entity dimension to obtain an initial cluster; clustering each search prompt text in the initial cluster according to the second high-frequency search entity dimension to divide the initial cluster into intermediate clusters.

[0091] The high-frequency search entity dimension is a search entity dimension that is used frequently during the search process. The first high-frequency search entity dimension and the second high-frequency search entity dimension are two of the high-frequency search entity dimensions. The initial cluster is a cluster obtained by clustering the set of search prompt texts according to the first high-frequency search entity dimension.

[0092] Specifically, the server may cluster the set of search prompt texts at the first level according to the first high-frequency search entity dimension to obtain initial clusters. Furthermore, the server may cluster the search prompt texts in each initial cluster at the second level according to the second high-frequency search entity dimension. After the above two levels of clustering, each initial cluster may be divided into intermediate clusters.

[0093] In one embodiment, during the search process, the frequency used by the first high-frequency search entity dimension may be higher, lower, or equal to the frequency used by the second high-frequency search entity dimension.

[0094] In the above embodiment, by clustering the set of search prompt texts in two levels according to the first high-frequency search entity dimension and the second high-frequency search entity dimension, the search prompt texts can be clustered in a more fine-grained manner, thereby improving the clustering accuracy of the search prompt texts. At the same time, for each level of clustering, the number of prompt texts in each cluster is continuously reduced, thereby reducing the time complexity of subsequent clustering and improving the clustering efficiency of the search prompt texts.

[0095] In one embodiment, step 202, that is, clustering the set of search prompt texts according to the search entity dimension to obtain an intermediate cluster, specifically includes: for the first search prompt text containing entity information under the search entity dimension, clustering the first search prompt texts containing the same entity information into one category; for the second search prompt text not containing entity information under the search entity dimension, clustering the second search prompt texts not containing entity information under the search entity dimension into one category to obtain an intermediate cluster.

[0096] The first search hint text is a search hint text containing entity information under the search entity dimension, and the second search hint text is a search hint text not containing entity information under the search entity dimension.

[0097] Specifically, among the search prompt texts, some contain entity information under the search entity dimension, and some contain entity information under the search entity dimension. For the first search prompt text containing entity information under the search entity dimension, the server may cluster the first search prompt texts containing the same entity information into one category. For the second search prompt texts not containing entity information under the search entity dimension, the server may cluster the second search prompt texts not containing entity information under the search entity dimension into one category to obtain an intermediate cluster.

[0098] In one embodiment, if the first search hint text contains multiple entity information under the search entity dimension, the server may ignore the order in which the multiple entity information appear in the first search hint text, and uniformly group the first search hint text containing the multiple entity information into one category.

[0099] For example, the set of search prompt texts is: {I want to watch TV series A, TV series A, I want to watch TV series A variety show B, variety show B TV series A}. TV series A and variety show B are entity information, so the server can cluster "I want to watch TV series A" and "TV series A" into one category, and cluster "I want to watch TV series A variety show B" and "variety show B TV series A" into one category to obtain the middle cluster.

[0100] In the above embodiment, the first search prompt texts containing the same entity information are clustered into one category, and the second search prompt texts not containing entity information under the search entity dimension are clustered into one category, so as to perform more fine-grained clustering of the search prompt texts, thereby further improving the clustering accuracy of the search prompt texts.

[0101] In one embodiment, the above-mentioned search prompt method specifically also includes: for two search prompt texts corresponding to any two nodes belonging to the same intermediate cluster, determining the common characters from the characters corresponding to the non-entity information contained in the two search prompt texts; determining the similarity between the two search prompt texts based on the similarity coefficient between the two search prompt texts and the sum of the importance coefficients corresponding to each common character; the importance coefficient is used to characterize the importance of the common character in the intermediate cluster.

[0102] Among them, non-entity information is information in the search prompt text that does not have a specific meaning. For example, if the search prompt text is "I want to watch movie Q", then "movie Q" is entity information, and "I want to watch" is non-entity information. The characters corresponding to non-entity information are the characters that make up the non-entity information. For example, the characters corresponding to "I want to watch" include the three characters "I", "want" and "watch". Common characters are characters that are common to the non-entity information contained in the two search prompt texts.

[0103] Specifically, if the intermediate cluster includes multiple search prompt texts, then for the two search prompt texts corresponding to any two nodes belonging to the same intermediate cluster, the server can extract non-entity information from the two search prompt texts, and perform character segmentation processing on the non-entity information to obtain the characters corresponding to the non-entity information. The server can determine the common characters from the characters corresponding to the included non-entity information. The server can determine the similarity coefficient between the two search prompt texts, and determine the importance coefficient corresponding to each common character. The server can add the importance coefficient corresponding to each common character to obtain the sum of the importance coefficients corresponding to each common character. The server can determine the similarity between the two search prompt texts based on the similarity coefficient between the two search prompt texts and the sum of the importance coefficients corresponding to each common character.

[0104] In one embodiment, the server may calculate the Jaccard coefficient between the two search prompt texts according to the Jaccard function, and use the Jaccard coefficient between the two search prompt texts as the similarity coefficient between the two search prompt texts. It should be noted that this embodiment is only used for illustration, and the calculation method of the similarity coefficient is not specifically limited. The similarity coefficient may be any coefficient other than the Jaccard coefficient that can characterize the degree of similarity between the two search prompt texts.

[0105] In one embodiment, the server may calculate the tf-idf value (termfrequency-inverse document frequency) of each common character in the corresponding intermediate cluster, and use the tf-idf value of each common character in the corresponding intermediate cluster as the importance coefficient corresponding to each common character. The larger the tf-idf value of the common character in the corresponding intermediate cluster, the more important the common character is to the corresponding intermediate cluster. It should be noted that this embodiment is only used for illustration, and the calculation method of the importance coefficient is not specifically limited. The importance coefficient can be any coefficient other than the tf-idf value that can characterize the importance of the common character in the intermediate cluster.

[0106] In one embodiment, the similarity between two search hint texts may be calculated using the following formula:

[0107] similarity(q1,q2)=Jaccard(q1,q2)-w*tf_idf_coef

[0108] Among them, q1 and q2 are search prompt texts, similarity(q1,q2) is the similarity between the two search prompt texts, Jaccard(q1,q2) is the similarity coefficient between the two search prompt texts q1 and q2, tf_idf_coef is the sum of the importance coefficients corresponding to the common characters of the two search prompt texts q1 and q2, and w is the preset weight. It can be understood that the item w*tf_idf_coef can limit the influence of low-frequency characters on the similarity calculation, making the similarity calculation more accurate, avoiding some low-frequency search prompt texts from being clustered with high-frequency search prompt texts due to their high similarity coefficients, and further improving the accuracy of clustering.

[0109] In the above embodiment, for two search hint texts corresponding to any two nodes belonging to the same intermediate cluster, the similarity between the two search hint texts can be calculated in a more fine-grained manner based on the similarity coefficient between the two search hint texts and the sum of the importance coefficients corresponding to each common character, thereby improving the accuracy of the similarity between the two search hint texts.

[0110] In one embodiment, the search prompt method further includes: if both search prompt texts include corresponding attribute information under the key prompt attribute, an adjustment coefficient is determined based on the similarity between the attribute information in the two search prompt texts. The step of determining the similarity between the two search prompt texts according to the similarity coefficient between the two search prompt texts and the sum of the importance coefficients corresponding to the common characters specifically includes: determining the similarity between the two search prompt texts according to the similarity coefficient between the two search prompt texts, the sum of the importance coefficients corresponding to the common characters, and the adjustment coefficient.

[0111] Among them, the key prompt attribute is an attribute that has a key prompt function in the search process, such as a number. The corresponding attribute information under the key prompt attribute is the specific text information under the key prompt attribute. For example, if the key prompt attribute is a number, the corresponding attribute information under the key prompt attribute can be any specific number, such as 20. The adjustment coefficient is a coefficient used to adjust the similarity between two search prompt texts.

[0112] Specifically, the search prompt text may include corresponding attribute information under the key prompt attribute, or may not include corresponding attribute information under the key prompt attribute. If both search prompt texts include corresponding attribute information under the key prompt attribute, the server may determine the adjustment coefficient based on the similarity between the attribute information in the two search prompt texts. The server may calculate the similarity between the two search prompt texts based on the similarity coefficient between the two search prompt texts, the sum of the importance coefficients corresponding to the common characters, and the adjustment coefficient.

[0113] In one embodiment, the key prompt attribute may also include at least one of letters, words, and preset keywords.

[0114] In one embodiment, if both search prompt texts include corresponding attribute information under the key prompt attribute, the server can calculate the similarity between the attribute information in the two search prompt texts. If the similarity between the attribute information in the two search prompt texts is greater than the preset similarity, the adjustment coefficient is 1, otherwise the adjustment coefficient is 0.

[0115] For example, if the search prompt text A includes the number 21 and the search prompt text B includes the number 22, the numbers 21 and 22 are different, and the adjustment coefficient is 1. If the search prompt text A and the search prompt text B both include the number 25, the adjustment coefficient is 0.

[0116] In one embodiment, if the key prompt attribute is a number, the similarity between two search prompt texts can be calculated using the following formula:

[0117] similarity(q1,q2)=Jaccard(q1,q2)-w1*tf_idf_coef-w2number_coef

[0118] Among them, number_coef is the adjustment coefficient, and w1 and w2 are preset weights. It can be understood that the item w2number_coef adjusts the similarity between the two search prompt texts according to the similarity between the numbers in the two search prompt texts, so that the similarity calculation is more accurate and avoids clustering the search prompt texts with low similarity between the numbers in the search prompt texts into one category.

[0119] In the above embodiment, by determining the similarity between the two search hint texts based on the similarity coefficient between the two search hint texts, the sum of the importance coefficients corresponding to each common character, and the adjustment coefficient, the similarity between the two search hint texts can be calculated in a more fine-grained manner, thereby improving the accuracy of the similarity between the two search hint texts.

[0120] In one embodiment, step 206, i.e., clustering the nodes in the similarity graph to divide the intermediate cluster into sub-clusters, specifically includes: selecting a starting node for this round of traversal from the nodes that have not been visited in the similarity graph, and starting from the starting node, traversing and visiting the nodes in the similarity graph in this round; when the end condition of this round of visit is met, the nodes visited in this round are divided into a sub-cluster corresponding to the intermediate cluster; taking the next round as the current round, returning to select the starting node for this round of traversal from the nodes that have not been visited in the similarity graph to continue iterative execution, until all the nodes in the similarity graph are visited and the iteration is stopped, and the intermediate cluster is divided into each sub-cluster.

[0121] The starting node is the first node visited in each round of traversal. The end condition of this round of visit is the condition for the end of each round of traversal.

[0122] Specifically, the server may need to perform multiple rounds of traversal of the similarity graph before the intermediate cluster can be divided into various sub-clusters. For each round of traversal, the server can select the starting node of this round of traversal from the nodes that have not been visited in the similarity graph, and start traversing the nodes in the access similarity graph from the starting node in this round. The server can monitor whether this round of traversal has reached the end condition of this round of access. When the end condition of this round of access is met, the server can divide the nodes visited in this round into a sub-cluster corresponding to the intermediate cluster. The server can use the next round as this round, return to select the starting node of this round of traversal from the nodes that have not been visited in the similarity graph to continue iterative execution, until all nodes in the similarity graph are visited and the iteration is stopped, and the intermediate cluster is divided into various sub-clusters.

[0123] In one embodiment, the condition for ending this round of visits may specifically be that the number of traversed layers reaches a preset traversal layer threshold, or that the number of nodes visited during the traversal process reaches a preset number.

[0124] In one embodiment, the server may select a starting node for this round of traversal from unvisited nodes in the similarity graph according to the average degree of each node.

[0125] In one embodiment, the server may determine the connected nodes in the similarity graph, and randomly select an unvisited node from the connected nodes as the starting node of this round of traversal.

[0126] In one embodiment, the server may randomly select an unvisited node from each node in the similarity graph as the starting node of this round of traversal.

[0127] In the above embodiment, by traversing the similarity graph corresponding to the intermediate cluster to achieve a finer-grained division of the intermediate cluster and obtain the sub-clusters corresponding to the intermediate cluster, the aggregation accuracy of the search prompt text can be improved.

[0128] In one embodiment, the step of selecting the starting node of this round of traversal from the unvisited nodes in the similarity graph specifically includes: obtaining a node sequence set; the node sequence set is obtained by sorting the identifiers of the unvisited nodes in the similarity graph according to the average degree of the nodes; selecting the identifier corresponding to the maximum average degree from the node sequence set, and using the node represented by the selected identifier as the starting node of this round. The step of taking the next round as the current round, returning to select the starting node of this round of traversal from the unvisited nodes in the similarity graph to continue iterative execution specifically includes: removing the identifiers of the nodes visited in this round from the node sequence set, obtaining an updated node sequence set, and taking the next round as the current round, selecting the identifier corresponding to the maximum average degree from the updated node sequence set to continue iterative execution.

[0129] The node identifier is a string that uniquely identifies the node. The average degree of a node is the average of the similarities between the node and other nodes directly connected to it.

[0130] Specifically, the server may sort the nodes corresponding to the similarity graph based on the similarity of the nodes, and generate a node sequence set based on the identifiers of the sorted nodes. The server may determine the node with the largest average degree, and select the identifier corresponding to the maximum average degree from the node sequence set. The server may use the node represented by the selected identifier as the starting node of this round, and start traversing and accessing the nodes in the similarity graph from the starting node in this round. When the end condition of this round of access is met, the server may divide the nodes accessed in this round into a sub-cluster corresponding to the intermediate cluster, and remove the identifiers of the nodes accessed in this round from the node sequence set to obtain an updated node sequence set. The server may use the next round as this round, and select the identifier corresponding to the maximum average degree from the updated node sequence set to continue iterative execution.

[0131] In the above embodiment, for each round of traversal, the similarity graph corresponding to the intermediate cluster is traversed starting from the node with the largest average degree to ensure that when the end condition of this round of access is met, nodes with relatively close similarities are clustered into one cluster, further improving the aggregation accuracy of the search prompt text.

[0132] In one embodiment, starting from the starting node, the step of traversing and accessing the nodes in the similarity graph in this round specifically includes: starting from the starting node, traversing and accessing the nodes in the similarity graph layer by layer in this round. When the end condition of this round of access is met, the step of dividing the nodes accessed in this round into a sub-cluster corresponding to the intermediate cluster specifically includes: when the number of traversed layers reaches a preset traversal layer threshold, the starting node accessed in this round and the nodes connected to the starting node are divided into a sub-cluster corresponding to the intermediate cluster.

[0133] Specifically, the server may traverse and access the nodes in the similarity graph layer by layer starting from the starting node in this round based on the breadth-first traversal algorithm. The server may compare the number of traversed layers with a preset traversal layer threshold. When the number of traversed layers reaches the preset traversal layer threshold, the server may divide the starting node accessed in this round and the nodes connected to the starting node into a sub-cluster corresponding to the intermediate cluster.

[0134] In the above embodiment, the aggregation accuracy of the search prompt text is improved by traversing the similarity graph layer by layer. Also, by presetting the number of traversal layers, when the number of traversal layers reaches the preset traversal layer threshold, the starting node visited in this round and the nodes connected to the starting node are divided into a sub-cluster corresponding to the intermediate cluster, thereby improving the aggregation efficiency of the search prompt text.

[0135] In one embodiment, the server may record the similarity graph as G=<V,E>, where G represents the similarity graph, V represents the set of nodes in the similarity graph, and E represents the set of edges between the nodes in the similarity graph. The server may record the set of nodes in the similarity graph as V={v1,v2,...,v m}, where v represents a node in the similarity graph and m represents the number of nodes in the similarity graph. The server can record the node sequence set as V'={v1',v2',...,v m '}, where V' represents a node sequence set, and v' represents a node in the sorted similarity graph. The server can record the set of sub-clusters after division as C = {c1, c2, ..., c N}, where C represents the set of subclusters, c represents each subcluster, and N represents the number of subclusters into which the intermediate cluster is divided. The server can record the set of nodes in each subcluster as c = {n1, n2, ..., n l}, where n represents the nodes in the subcluster and l represents the number of nodes in the subcluster. The set of nodes in the similarity graph and the set of subclusters satisfy V = c1∪c2∪...∪c N .

[0136] The server may start traversal from the starting node v1' with the largest average degree in this round. At this time, c = {v1'}, and the server may remove the nodes connected to the starting node v1' of this round from the node sequence set V' in sequence, and update the node sequence set V' until the number of traversed layers reaches the preset traversal layer threshold, then stop traversal in this round, and add the obtained subcluster c to the subcluster set C. The server may use the next round as this round, and select the starting node with the largest average degree from the updated node sequence set V' to continue iterative execution until there are no elements in V', and stop traversing the similarity graph to obtain each subcluster.

[0137] In one embodiment, Figure 4 As shown, the middle cluster includes node AL. If the average degree of node F is the largest among the nodes that have not been visited in this round of traversal, node F is used as the starting node of this round of traversal. If the preset number of traversal layers is 2, the server can use nodes F, D, G, B and C that have been visited in this round of traversal as a subcluster, where node D is the cluster center of the subcluster. Similarly, the server uses nodes E and H as a subcluster, where node E is the cluster center of the subcluster, and nodes I, K and L as a subcluster, where node L is the cluster center of the subcluster. Nodes A and J are two subclusters respectively.

[0138] In one embodiment, the search prompt method further includes: using each search prompt text in the same subcluster as a key, using the search prompt text of the cluster center belonging to the subcluster as a value, and storing each search prompt text in the subcluster in the form of a key-value pair. When performing a search prompt for a search term, the step of removing duplicates from the search prompt texts that are hit by the search term and belong to the same subcluster specifically includes: when performing a search prompt for a search term, using each search prompt text that is hit by the search term as a key, and searching for the corresponding value from the storage space; replacing the search prompt text that is hit by the search term and corresponds to the same value with the search prompt text that belongs to the found value, so as to remove duplicates.

[0139] Specifically, the server may use each search prompt text in the same subcluster as a key, and the search prompt text belonging to the cluster center of the corresponding subcluster as a value, and store each search prompt text in the subcluster in the storage space in the form of a key-value pair. It can be understood that if a subcluster includes multiple search prompt texts, the subcluster has multiple keys, and each key corresponds to only one value, that is, the search prompt text corresponding to the cluster center of the subcluster. When performing search prompts for search terms, the server may use each search prompt text hit by the search term as a key, and search for the value corresponding to the key from the storage space. The server may replace the search prompt text hit by the search term and corresponding to the same value with the search prompt text belonging to the found value to perform deduplication. It can be understood that if multiple search prompt texts hit by the search term come from the same subcluster, the multiple search prompt terms are represented by the cluster center of the subcluster.

[0140] In the above embodiment, each search prompt text in the sub-cluster is stored offline in the form of a key-value pair, which can improve the efficiency of subsequent deduplication of the search prompt text. The diversity of the search prompt text can be improved by replacing the search prompt text that is hit by the search term and corresponds to the same value with the search prompt text that belongs to the found value for deduplication.

[0141] In one embodiment, the above-mentioned search prompt method specifically also includes: performing at least one of the following preprocessing on the initial search prompt text to obtain a set of search prompt texts: removing non-key information from the initial search prompt text; performing error correction on the initial search prompt text; and standardizing the expression of the initial search prompt text.

[0142] Among them, non-key information refers to information in the search prompt text that does not play a key role in the semantic expression. The expression method refers to the description form of the search prompt text.

[0143] Specifically, the server can extract non-critical information from the search prompt text and perform non-critical information removal processing on the initial search prompt text; and / or, the server can extract error information from the search prompt text and perform error correction processing on the initial search prompt text; and / or, the server can perform standardized processing on the expression mode of the initial search prompt text to obtain a set of search prompt texts.

[0144] In the above embodiments, by preprocessing the initial search prompt text and then clustering the set of search prompt texts obtained by the preprocessing, the accuracy of clustering the search prompt texts can be improved.

[0145] In one embodiment, the server can obtain the initial search prompt text from the user search log, and the initial search prompt text may include many identical search prompt texts. The server can remove duplicates from the identical search prompt texts in the initial search prompt text. For example, if the initial search prompt text includes multiple "TV drama A", the server can remove duplicates from these multiple "TV drama A", and only one "TV drama A" is retained in the search prompt text after duplicate removal. In this way, storage space can be saved and the aggregation efficiency can be improved.

[0146] In one embodiment, the non-critical information includes stop words. The step of performing non-critical information removal processing on the initial search prompt text specifically includes: obtaining a preset stop word list; determining the words in the initial search prompt text that match the stop words in the stop word list; and removing the words from the initial search prompt text.

[0147] Among them, the stop words are pre-stopped texts. The stop word list is a list used to record stop words.

[0148] Specifically, the server can obtain a preset stop word list and determine the words in the initial search prompt text that match the stop words in the stop word list. Furthermore, the server can remove the words from the initial search prompt text. It should be noted that the word can specifically be a single character, a phrase, or a sentence.

[0149] In one embodiment, the stop words may include at least one of "de", "le", "zai", "ma", and "zhe", etc.

[0150] In the above embodiments, according to the stop word list, the non-critical information in the initial search prompt text can be accurately and quickly removed. By removing the non-critical information in the initial search prompt text, storage space can be saved and the accuracy of subsequent aggregation of search prompt words can be improved.

[0151] In one embodiment, the step of performing expression standardization processing on the initial search prompt text specifically includes at least one of the following processing: from the initial search prompt text, determining the words that hit the synonyms in the synonym table to obtain the words to be standardized; normalizing the words to be standardized into corresponding standard words; for words with the same attribute and different expressions in the initial search prompt text, normalizing the words belonging to the first expression to the words of the second expression; the usage frequency of the second expression is higher than the usage frequency of the first expression.

[0152] Among them, synonyms are texts with similar semantics. Synonyms is a table used to record synonyms. To-be-standardized words are texts to be standardized in expression. Standard words are texts that meet the preset standardized expression. The first expression and the second expression are two different expressions.

[0153] Specifically, the server may obtain a synonym table, and determine the words that hit the synonyms in the synonym table from the initial search prompt text to obtain the words to be standardized. The server may normalize the words to be standardized into corresponding standard words. And / or, the server may extract words with the same attribute and different expressions from the initial search prompt text. For the words with the same attribute and different expressions in the initial search prompt text, the server may normalize the words with a first expression with a lower frequency of use to words with a second expression with a higher frequency of use.

[0154] For example, in the initial search prompt text, the words with the same attribute and different expressions are as follows, for example, "I want to watch TV series W Part 3" and "I want to watch TV series W Season 3". In this case, "part" and "season" under the attribute "TV series W" are two words with different expressions. If the most frequently used expression for "TV series W" is "part", the server can normalize all "seasons" to "part".

[0155] In one embodiment, the server may update the usage frequencies of the first expression and the second expression at preset time intervals, for example, updating the usage frequencies of the first expression and the second expression respectively on a daily basis to achieve dynamic customization of the expressions.

[0156] In the above embodiment, based on the synonym table, the initial search prompt text can be accurately and quickly standardized in terms of expression. By standardizing the expressions of words with the same attribute but different expressions in the initial search prompt text, the efficiency of subsequent clustering of the search prompt text can be improved.

[0157] In one embodiment, the search prompt text is a video search prompt text; the search term is a video search term. When performing a search prompt for the search term, the step of removing duplicate search prompt texts that are hit by the search term and belong to the same subcluster specifically includes: when performing a video search prompt for the video search term, removing duplicate video search prompt texts that are hit by the video search term and belong to the same subcluster.

[0158] The video search prompt text is the text used for prompting when searching for a video, and the video search term is the text entered by the user when searching for a video.

[0159] Specifically, the terminal may obtain the video search term input by the user based on the video search page of the terminal, and send the video search term to the server. The server may receive the video search term sent by the terminal. When performing a video search prompt for the video search term, the server may remove duplicates of the video search prompt texts that are hit by the video search term and belong to the same subcluster. It can be understood that all the video search prompt texts after removal of duplicates correspond to different subclusters.

[0160] In one embodiment, the server may cluster the set of video search prompt texts according to the video search entity dimension to obtain intermediate clusters. The server may abstract each video search prompt text in the intermediate cluster into a node, and generate a similarity graph corresponding to each intermediate cluster based on the similarity between the node and the video search prompt text corresponding to each node. The server may cluster the nodes in the similarity graph to divide the intermediate cluster into subclusters. When performing video search prompts for video search terms, the server may deduplicate the video search prompt texts that are hit by the video search terms and belong to the same subcluster.

[0161] In one embodiment, the server may use each video search prompt text in the same subcluster as a key and the video search prompt text of the cluster center of the subcluster as a value to store each video search prompt text in the subcluster in the form of a key-value pair. When searching for a video search term, the server may use each video search prompt text that is hit by the video search term as a key to search for a corresponding value from the storage space. The video search prompt text that is hit by the video search term and corresponds to the same value is replaced with the video search prompt text that belongs to the found value to perform deduplication.

[0162] In the above embodiment, when performing video search prompts for video search words, the diversity of video search prompt texts that are hit by the video search words and belong to the same subcluster can be improved during the video search process.

[0163] In one embodiment, Figure 5As shown, for the video search field, the server can obtain a set of initial search prompt texts from the user's video search log. The server can perform non-key information removal, error correction, entity information recognition, synonym standardization, and low-frequency expression standardization on the set of initial search prompt texts to obtain a set of search prompt texts to be aggregated. The server can perform a first-level clustering process on the search prompt texts according to the video name to obtain an initial cluster. For the search prompt texts in the initial cluster, the server can perform a second-level clustering process on the search prompt texts in the initial cluster according to the name of the person (for example, the name of the role and the name of the actor, etc.) to obtain an intermediate cluster. The server can further divide the intermediate cluster into subclusters by traversing the intermediate cluster. Furthermore, when performing a video search prompt for the video keyword entered by the user, the server can perform a deduplication process on the video search prompt text that is hit by the video search term according to the cluster center of the subcluster to obtain the deduplication video search prompt text.

[0164] In one embodiment, a search prompt method is provided, the method specifically comprising the following steps:

[0165] (1) Perform at least one of the following preprocessing on the initial video search prompt text to obtain a set of video search prompt texts: remove non-key information from the initial video search prompt text; perform error correction on the initial video search prompt text; and perform expression standardization on the initial video search prompt text.

[0166] In one embodiment, non-critical information includes stop words; performing non-critical information removal processing on the initial video search prompt text includes: obtaining a preset stop word list; determining, from the initial video search prompt text, words that hit the stop words in the stop word list; and removing words from the initial video search prompt text.

[0167] In one embodiment, the initial video search prompt text is subjected to expression standardization processing, including at least one of the following processing: from the initial video search prompt text, determining words that hit synonyms in a synonym table to obtain words to be standardized; normalizing the words to be standardized into corresponding standard words; for words with the same attribute in the initial video search prompt text but with different expressions, normalizing words belonging to a first expression into words of a second expression; and the usage frequency of the second expression is higher than the usage frequency of the first expression.

[0168] (2) Cluster the set of video search prompt texts according to the first high-frequency search entity dimension to obtain an initial cluster.

[0169] (3) Clustering each video search prompt text in the initial cluster according to the second high-frequency search entity dimension to divide the initial cluster into intermediate clusters.

[0170] (4) Each video search prompt text in the intermediate cluster is abstracted into a node. For two video search prompt texts corresponding to any two nodes belonging to the same intermediate cluster, the common characters are determined from the characters corresponding to the non-entity information contained in the two video search prompt texts.

[0171] (5) Determine the similarity between the two video search prompt texts based on the similarity coefficient between the two video search prompt texts and the sum of the importance coefficients corresponding to each common character; the importance coefficient is used to characterize the importance of the common character in the middle cluster.

[0172] In one embodiment, if both video search prompt texts include corresponding attribute information under the key prompt attribute, the adjustment coefficient is determined based on the similarity between the attribute information in the two video search prompt texts; the similarity between the two video search prompt texts is determined based on the similarity coefficient between the two video search prompt texts, the sum of the importance coefficients corresponding to each common character, and the adjustment coefficient.

[0173] (6) Generate a similarity graph corresponding to each intermediate cluster based on the similarity between the node and the search prompt text corresponding to each node.

[0174] (7) Obtaining a node sequence set; the node sequence set is obtained by sorting the identifiers of the unvisited nodes in the similarity graph according to the average degree of the nodes.

[0175] (8) Select the identifier corresponding to the maximum average degree from the node sequence set, and use the node represented by the selected identifier as the starting node of this round.

[0176] (9) Starting from the starting node, this round begins to traverse the nodes in the similarity graph layer by layer.

[0177] (10) When the number of traversed layers reaches a preset traversal layer threshold, the starting node visited in this round and the nodes connected to the starting node are divided into a sub-cluster corresponding to the intermediate cluster.

[0178] (11) The identifiers of the nodes visited in this round are removed from the node sequence set to obtain an updated node sequence set, and the next round is used as the current round. The identifier corresponding to the maximum average degree is selected from the updated node sequence set to continue iterating until all nodes in the similarity graph are visited and the iteration is stopped to obtain the sub-clusters into which the intermediate cluster is divided.

[0179] (12) Using each video search prompt text in the same sub-cluster as a key and the video search prompt text at the cluster center of the sub-cluster as a value, the video search prompt text in the sub-cluster is stored in the form of a key-value pair.

[0180] (13) When searching for a video search term, the video search prompt text that matches the video search term is used as a key to find the corresponding value from the storage space.

[0181] (14) When performing a video search prompt for a video search term, the video search prompt text that matches the video search term and corresponds to the same value is replaced with the video search prompt text that belongs to the found value to remove duplication.

[0182] In one embodiment, before the search prompt method of the present application is used to remove duplicates from the search prompt words, Figure 6 As shown in FIG. 1 , among the search prompt texts displayed when the search word hits, there are two search prompt texts with different expressions but the same meaning (i.e., search prompt text 9.1 and search prompt text 9.2), which results in poor diversity of search prompt texts. After the search prompt method of the present application is used to remove duplicate search prompt words, as shown in FIG. Figure 7 As shown, the expression forms and meanings of all search hint texts displayed for the search term hit are different (ie, the expression forms and meanings of search hint texts 1-10 are different), which improves the diversity of the search hint texts.

[0183] In one embodiment, the server may use 300 search terms to perform GSB (Good Similarity Bad) experimental analysis on the diversity of search prompt terms before the search prompt method of the present application is used to remove duplicates, and the server may perform GSB (Good Similarity Bad) experimental analysis on the diversity of search prompt terms after the search prompt method of the present application is used to remove duplicates, and the obtained GSB result is 56:244:0, that is, there are 56 search terms among the 300 search terms, and the diversity of search prompt terms after the search prompt method of the present application is used to remove duplicates is better than the diversity of search prompt terms before the search prompt method of the present application is used to remove duplicates. There are 244 search terms among the 300 search terms, and the diversity of search prompt terms after the search prompt method of the present application is used to remove duplicates is equal to the diversity of search prompt terms before the search prompt method of the present application is used to remove duplicates. And, there are 0 search terms among the 300 search terms, and the diversity of the search prompt terms after the search prompt method of the present application is used to remove duplicate search prompt terms is worse than the diversity of the search prompt terms before the search prompt method of the present application is used to remove duplicate search prompt terms. The GSB result shows that the search prompt method of the present application has a good effect on removing duplicate search prompt terms and can improve the diversity of search prompt terms.

[0184] The present application also provides an application scenario, which applies the above-mentioned search prompt method. Specifically, the search prompt method can be applied to the scenario of video search prompts. The server can obtain the initial video search prompt text, the preset stop word list and the preset synonym list from the user's video search log. The server performs error correction processing on the initial search prompt text, and can determine the words that hit the stop words in the stop word list from the initial video search prompt text, and remove the words from the initial video search prompt text. Furthermore, the server can determine the words that hit the synonyms in the synonym list from the initial video search prompt text, obtain the words to be standardized, and normalize the words to be standardized into corresponding standard words. At the same time, for the words under the same attribute and with different expressions in the initial video search prompt text, the words belonging to the first expression with lower frequency of use are normalized to the words of the second expression with higher frequency of use, so as to obtain a set of video search prompt texts.

[0185] The server may cluster the set of video search prompt texts at the first level according to the video names to obtain initial clusters, and cluster the video search prompt texts in the initial clusters at the second level according to the names to divide the initial clusters into intermediate clusters.

[0186] The server may abstract each video search prompt text in the intermediate cluster into a node. For two video search prompt texts corresponding to any two nodes belonging to the same intermediate cluster, the server may determine the common characters from the characters corresponding to the non-entity information contained in the two video search prompt texts. If both video search prompt texts include numbers, the adjustment coefficient is determined based on the similarity between the numbers in the two video search prompt texts. Furthermore, the server may determine the similarity between the two video search prompt texts based on the Jaccard coefficient between the two video search prompt texts, the sum of the tf-idf values ​​corresponding to the common characters, and the adjustment coefficient.

[0187] The server may generate a similarity graph corresponding to each intermediate cluster according to the similarity between the node and the search prompt text corresponding to each node. The server may sort the identifiers of the unvisited nodes in the similarity graph according to the average degree of the nodes to obtain a node sequence set. The server may select the identifier corresponding to the maximum average degree from the node sequence set, and use the node represented by the selected identifier as the starting node of this round. The server may traverse and access the nodes in the similarity graph layer by layer starting from the starting node in this round based on the breadth-first traversal algorithm. When the number of traversed layers reaches the preset traversal layer threshold, the starting node visited in this round and the nodes connected to the starting node are divided into a sub-cluster corresponding to the intermediate cluster. The server may remove the identifiers of the nodes visited in this round from the node sequence set to obtain an updated node sequence set, and use the next round as this round, select the identifier corresponding to the maximum average degree from the updated node sequence set to continue iterative execution, until all the nodes in the similarity graph are visited and the iteration is stopped, and the intermediate cluster is divided into each sub-cluster.

[0188] The server can use each video search prompt text in the same subcluster as a key and the video search prompt text of the cluster center of the subcluster as a value to store each video search prompt text in the subcluster in the form of a key-value pair. When searching for a video search term, the server can use each video search prompt text that is hit by the video search term as a key to find the corresponding value from the storage space, and replace the video search prompt text that is hit by the video search term and corresponds to the same value with the video search prompt text that belongs to the found value to remove duplication, thereby improving the diversity of the video search prompt text.

[0189] The present application also provides an application scenario, which applies the above-mentioned search prompt method. Specifically, the search prompt method can be applied to the scenario of document search prompt. The server can cluster the set of document search prompt texts according to the search entity dimension to obtain intermediate clusters. The server can abstract each document search prompt text in the intermediate cluster into a node. The server can generate a similarity graph corresponding to each intermediate cluster based on the similarity between the node and the document search prompt text corresponding to each node. The server can cluster the nodes in the similarity graph to divide the intermediate cluster into subclusters. When performing document search prompts for document search terms, the document search prompt texts that are hit by the document search terms and belong to the same subcluster are deduplicated, thereby improving the diversity of the document search prompt texts.

[0190] It should be understood that although Figure 2 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, Figure 2 At least part of the steps may include multiple steps or multiple stages. These steps or stages are not necessarily performed at the same time, but can be performed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed in turn or alternately with other steps or at least part of the steps or stages in other steps.

[0191] In one embodiment, Figure 8 As shown, a search prompt device 800 is provided. The device can adopt a software module or a hardware module, or a combination of the two to become a part of a computer device. The device specifically includes: a clustering module 801, a generation module 802 and a deduplication module 803, wherein:

[0192] The clustering module 801 is used to cluster the set of search prompt texts according to the search entity dimension to obtain intermediate clusters.

[0193] The generating module 802 is used to abstract each search prompt text into a node, and generate a similarity graph corresponding to each intermediate cluster according to the similarity and the node.

[0194] The clustering module 801 is also used to cluster the nodes in the similarity graph to divide the middle cluster into sub-clusters.

[0195] The deduplication module 803 is used to deduplicate the search prompt texts that are matched with the search word and belong to the same sub-cluster when performing search prompts for the search word.

[0196] In one embodiment, the search entity dimension includes a first high-frequency search entity dimension and a second high-frequency search entity dimension; the clustering module 801 is also used to cluster the set of search prompt texts according to the first high-frequency search entity dimension to obtain an initial cluster; and cluster each search prompt text in the initial cluster according to the second high-frequency search entity dimension to divide the initial cluster into intermediate clusters.

[0197] In one embodiment, the apparatus further comprises:

[0198] Determination module 804 is used to determine the common characters from the characters corresponding to the non-entity information contained in the two search prompt texts for the two search prompt texts corresponding to any two nodes belonging to the same intermediate cluster; determine the similarity between the two search prompt texts based on the similarity coefficient between the two search prompt texts and the sum of the importance coefficients corresponding to the common characters; the importance coefficient is used to characterize the importance of the common characters in the intermediate cluster.

[0199] In one embodiment, the determination module 804 is also used to determine the adjustment coefficient based on the similarity between the attribute information in the two search prompt texts if both search prompt texts include corresponding attribute information under the key prompt attribute; and determine the similarity between the two search prompt texts based on the similarity coefficient between the two search prompt texts, the sum of the importance coefficients corresponding to each common character, and the adjustment coefficient.

[0200] In one embodiment, the clustering module 801 is also used to select the starting node of this round of traversal from the nodes that have not been visited in the similarity graph, and start traversing the nodes in the similarity graph from the starting node in this round; when the end condition of this round of visit is met, the nodes visited in this round are divided into a sub-cluster corresponding to the intermediate cluster; the next round is taken as this round, and the starting node of this round of traversal is selected from the nodes that have not been visited in the similarity graph to continue iterative execution, until all the nodes in the similarity graph are visited and the iteration is stopped to obtain the sub-clusters into which the intermediate cluster is divided.

[0201] In one embodiment, the clustering module 801 is also used to obtain a node sequence set; the node sequence set is obtained by sorting the identifiers of unvisited nodes in the similarity graph according to the average degree of the nodes; selecting the identifier corresponding to the maximum average degree from the node sequence set, and using the node represented by the selected identifier as the starting node of this round; removing the identifiers of the nodes visited in this round from the node sequence set to obtain an updated node sequence set, and using the next round as this round, selecting the identifier corresponding to the maximum average degree from the updated node sequence set to continue iterative execution.

[0202] In one embodiment, the clustering module 801 is also used to traverse the nodes in the similarity graph layer by layer starting from the starting node in this round; when the number of traversed layers reaches a preset traversal layer threshold, the starting node visited in this round and the nodes connected to the starting node are divided into a sub-cluster corresponding to the intermediate cluster.

[0203] In one embodiment, the deduplication module 803 is also used to use each search prompt text in the same subcluster as a key and the search prompt text of the cluster center belonging to the subcluster as a value, and store each search prompt text in the subcluster in the form of a key-value pair; when performing a search prompt for a search term, use each search prompt text that hits the search term as a key to find the corresponding value from the storage space; replace the search prompt text that hits the search term and corresponds to the same value with the search prompt text belonging to the found value to perform deduplication.

[0204] In one embodiment, the search prompt text is a video search prompt text; the search term is a video search term; and the deduplication module 803 is also used to deduplicate the video search prompt text that hits the video search term and belongs to the same subcluster when performing a video search prompt on the video search term.

[0205] refer to Fig. 9 In one embodiment, the search prompt device 800 further includes: a determination module 804 and a pre-processing module 805, wherein:

[0206] The preprocessing module 805 is used to perform at least one of the following preprocessing on the initial search prompt text to obtain a set of search prompt texts: removing non-key information from the initial search prompt text; correcting errors in the initial search prompt text; and standardizing the expression of the initial search prompt text.

[0207] In one embodiment, the non-critical information includes stop words; the pre-processing module 805 is further used to obtain a preset stop word list; determine the words that hit the stop words in the stop word list from the initial search prompt text; and remove the words from the initial search prompt text.

[0208] In one embodiment, the preprocessing module 805 is also used to determine the words that hit the synonyms in the synonym table from the initial search prompt text to obtain the words to be standardized; normalize the words to be standardized into corresponding standard words; and / or, for words with the same attribute and different expressions in the initial search prompt text, normalize the words belonging to the first expression to the words of the second expression; the usage frequency of the second expression is higher than the usage frequency of the first expression.

[0209] The above-mentioned search prompt device clusters the set of search prompt texts according to the search entity dimension to obtain intermediate clusters to perform preliminary cluster division on the search prompt texts, which can avoid dividing the search prompt texts under different search entity dimensions into the same cluster during subsequent clustering. Each search prompt text in the intermediate cluster is abstracted as a node, and a similarity graph corresponding to each intermediate cluster is generated based on the similarity between the node and the search prompt text corresponding to each node, so as to connect the nodes with higher similarity. The nodes in the similarity graph are clustered to divide the intermediate clusters into sub-clusters, so that the nodes with higher similarity can be divided into the same cluster. When performing search prompts for search terms, the search prompt texts that are hit by the search terms and belong to the same sub-cluster are deduplicated, thereby improving the diversity of the search prompt texts that are hit by the search terms during the search process.

[0210] For the specific definition of the search prompt device, please refer to the definition of the search prompt method above, which will not be repeated here. Each module in the above-mentioned search prompt device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned 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.

[0211] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig.10 As shown. The computer device includes a processor, a memory and a network interface 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, 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 through a network connection. When the computer program is executed by the processor, a search prompt method is implemented.

[0212] Those skilled in the art will understand that Fig.10 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.

[0213] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0214] In one embodiment, a computer-readable storage medium is provided, storing a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0215] In one embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0216] Those of ordinary skill 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 at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0217] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above 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.

[0218] 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 search prompt method, It is characterized in that The method comprises: Cluster the collection of search prompt texts according to the search entity dimension to obtain the intermediate cluster; Abstracting each of the search prompt texts in the intermediate cluster into a node, and generating a similarity graph corresponding to each of the intermediate clusters according to the similarity between the node and the search prompt text corresponding to each of the nodes; Clustering the nodes in the similarity graph to divide the intermediate cluster into subclusters; Using each search prompt text in the same sub-cluster as a key and the search prompt text of the cluster center belonging to the sub-cluster as a value, storing each search prompt text in the sub-cluster in the form of a key-value pair; When performing a search prompt for a search term, each search prompt text that is hit by the search term is used as a key to find the corresponding value; The search hint text that is hit by the search term and corresponds to the same value is replaced with the search hint text belonging to the found value to remove duplication.

2. The method according to claim 1, It is characterized in that The search entity dimension includes a first high-frequency search entity dimension and a second high-frequency search entity dimension; clustering the set of search prompt texts according to the search entity dimension to obtain an intermediate cluster includes: Clustering the set of search prompt texts according to the first high-frequency search entity dimension to obtain initial clusters; The search prompt texts in the initial cluster are clustered according to the second high-frequency search entity dimension to divide the initial cluster into intermediate clusters.

3. The method according to claim 1, It is characterized in that The method further comprises: For two search prompt texts corresponding to any two nodes belonging to the same intermediate cluster, determine a common character from characters corresponding to non-entity information contained in the two search prompt texts; The similarity between the two search prompt texts is determined based on the similarity coefficient between the two search prompt texts and the sum of the importance coefficients corresponding to the common characters; the importance coefficient is used to characterize the importance of the common character in the middle cluster.

4. The method according to claim 3, It is characterized in that The method further comprises: If both of the two search prompt texts include corresponding attribute information under the key prompt attribute, then determining an adjustment coefficient based on the similarity between the attribute information in the two search prompt texts; Determining the similarity between the two search prompt texts according to the similarity coefficient between the two search prompt texts and the sum of the importance coefficients corresponding to the common characters includes: The similarity between the two search prompt texts is determined according to the similarity coefficient between the two search prompt texts, the sum of the importance coefficients corresponding to the common characters, and the adjustment coefficient.

5. The method according to claim 1, It is characterized in that The clustering of the nodes in the similarity graph to divide the intermediate cluster into subclusters includes: Selecting a starting node for this round of traversal from unvisited nodes in the similarity graph, and starting from the starting node, traversing and accessing nodes in the similarity graph in this round; When the end condition of this round of visits is met, the nodes visited in this round are divided into a sub-cluster corresponding to the intermediate cluster; The next round is taken as the current round, and the starting node of the current round traversal is selected from the unvisited nodes in the similarity graph to continue iterative execution until all nodes in the similarity graph are visited and the iteration is stopped to obtain the sub-clusters into which the intermediate cluster is divided.

6. The method according to claim 5, It is characterized in that The step of selecting a starting node for this round of traversal from the unvisited nodes in the similarity graph comprises: Obtaining a node sequence set; the node sequence set is obtained by sorting the identifiers of the unvisited nodes in the similarity graph according to the average degree of the nodes; the average degree of the node is the average value of the similarities between the node and other nodes directly connected to it; Selecting an identifier corresponding to the maximum average degree from the node sequence set, and using the node represented by the selected identifier as the starting node of this round; The step of taking the next round as the current round and returning to selecting a starting node of the current round of traversal from the unvisited nodes in the similarity graph to continue iterative execution includes: The identifiers of the nodes visited in this round are removed from the node sequence set to obtain an updated node sequence set, and the next round is used as this round, and the identifier corresponding to the maximum average degree is selected from the updated node sequence set to continue iterative execution.

7. The method according to claim 5, It is characterized in that Starting from the starting node, traversing and accessing the nodes in the similarity graph in this round includes: Starting from the starting node, traverse and visit the nodes in the similarity graph layer by layer in this round; When the end condition of this round of visit is met, the nodes visited in this round are divided into a sub-cluster corresponding to the intermediate cluster, including: When the number of traversed layers reaches a preset traversal layer threshold, the starting node visited in this round and the nodes connected to the starting node are divided into a sub-cluster corresponding to the intermediate cluster.

8. The method according to claim 1, It is characterized in that The method further comprises: Perform at least one of the following preprocessing on the initial search prompt text to obtain a set of the search prompt text: Removing non-critical information from the initial search prompt text; Performing error correction processing on the initial search prompt text; The initial search prompt text is subjected to expression standardization processing.

9. The method according to claim 8, It is characterized in that The non-key information includes stop words; and the non-key information removal process of the initial search prompt text includes: Get the preset stop word list; Determining, from the initial search prompt text, words that hit stop words in the stop word list; The term is removed from the initial search hint text.

10. The method according to claim 8, It is characterized in that The step of normalizing the expression of the initial search prompt text includes at least one of the following processes: Determine the words that hit the synonyms in the synonym table from the initial search prompt text to obtain the words to be standardized; normalize the words to be standardized into corresponding standard words; For words with the same attribute and different expressions in the initial search prompt text, the words belonging to the first expression are normalized to words of the second expression; the usage frequency of the second expression is higher than the usage frequency of the first expression.

11. The method according to any one of claims 1 to 10, It is characterized in that The search prompt text is a video search prompt text; the search term is a video search term; and the search prompt is a video search prompt.

12. A search prompt device, It is characterized in that The device comprises: A clustering module is used to cluster the set of search prompt texts according to the search entity dimension to obtain intermediate clusters; A generating module, used for abstracting each of the search prompt texts in the intermediate cluster into a node, and generating a similarity graph corresponding to each of the intermediate clusters according to the similarity between the node and the search prompt text corresponding to each of the nodes; The clustering module is also used to cluster the nodes in the similarity graph to divide the intermediate cluster into sub-clusters; A deduplication module is used to store the search prompt texts in the subcluster in the form of key-value pairs, using each search prompt text in the same subcluster as a key and the search prompt text in the cluster center of the subcluster as a value; when performing a search prompt for a search term, each search prompt text that hits the search term is used as a key to find the corresponding value; and the search prompt text that hits the search term and corresponds to the same value is replaced with the search prompt text that belongs to the found value to perform deduplication.

13. The device according to claim 12, It is characterized in that The search entity dimension includes a first high-frequency search entity dimension and a second high-frequency search entity dimension; the clustering module is also used to cluster the set of search prompt texts according to the first high-frequency search entity dimension to obtain an initial cluster; and cluster each search prompt text in the initial cluster according to the second high-frequency search entity dimension to divide the initial cluster into intermediate clusters.

14. The device according to claim 12, It is characterized in that The device also includes: A determination module is used to determine common characters from characters corresponding to non-entity information contained in two search prompt texts corresponding to any two nodes belonging to the same intermediate cluster; determine the similarity between the two search prompt texts based on a similarity coefficient between the two search prompt texts and the sum of importance coefficients corresponding to each of the common characters; the importance coefficient is used to characterize the importance of the common character in the intermediate cluster.

15. The device according to claim 14, It is characterized in that The determination module is also used to determine the adjustment coefficient based on the similarity between the attribute information in the two search prompt texts if both of the two search prompt texts include corresponding attribute information under the key prompt attribute; and determine the similarity between the two search prompt texts based on the similarity coefficient between the two search prompt texts, the sum of the importance coefficients corresponding to each of the common characters, and the adjustment coefficient.

16. The device according to claim 12, It is characterized in that The clustering module is also used to select a starting node for this round of traversal from the nodes that have not been visited in the similarity graph, and start traversing and visiting the nodes in the similarity graph from the starting node in this round; when the end condition of this round of visit is met, the nodes visited in this round are divided into a sub-cluster corresponding to the intermediate cluster; the next round is taken as this round, and the starting node for this round of traversal is selected from the nodes that have not been visited in the similarity graph to continue iterative execution, until all the nodes in the similarity graph are visited and the iteration is stopped to obtain the sub-clusters into which the intermediate cluster is divided.

17. The device according to claim 16, It is characterized in that The clustering module is also used to obtain a node sequence set; the node sequence set is obtained by sorting the identifiers of unvisited nodes in the similarity graph according to the size of the average degree of the nodes; the average degree of the node is the average value of the similarities between the node and other nodes directly connected to it; the identifier corresponding to the maximum average degree is selected from the node sequence set, and the node represented by the selected identifier is used as the starting node of this round; the identifiers of the nodes visited in this round are removed from the node sequence set to obtain an updated node sequence set, and the next round is used as this round, and the identifier corresponding to the maximum average degree is selected from the updated node sequence set to continue iterative execution.

18. The device according to claim 16, It is characterized in that The clustering module is also used to traverse and visit the nodes in the similarity graph layer by layer starting from the starting node in this round; when the number of traversed layers reaches a preset traversal layer threshold, the starting node visited in this round and the nodes connected to the starting node are divided into a sub-cluster corresponding to the intermediate cluster.

19. The device according to claim 12, It is characterized in that The device also includes: The preprocessing module is used to perform at least one of the following preprocessing on the initial search prompt text to obtain a set of the search prompt text: removing non-key information from the initial search prompt text; correcting errors from the initial search prompt text; and standardizing the expression of the initial search prompt text.

20. The device according to claim 19, It is characterized in that The non-key information includes stop words; the non-key information removal processing of the initial search prompt text includes: obtaining a preset stop word list; determining, from the initial search prompt text, words that hit the stop words in the stop word list; and removing the words from the initial search prompt text.

21. The device according to claim 19, It is characterized in that The step of normalizing the expression of the initial search prompt text includes at least one of the following processes: Determine the words that hit the synonyms in the synonym table from the initial search prompt text to obtain the words to be standardized; normalize the words to be standardized into corresponding standard words; For words with the same attribute and different expressions in the initial search prompt text, the words belonging to the first expression are normalized to words of the second expression; the usage frequency of the second expression is higher than the usage frequency of the first expression.

22. The device according to any one of claims 12 to 21, It is characterized in that The search prompt text is a video search prompt text; the search term is a video search term; and the search prompt is a video search prompt.

23. A computer device comprising a memory and a processor, wherein the memory stores a computer program, It is characterized in that When the processor executes the computer program, the steps of the method according to any one of claims 1 to 11 are implemented.

24. A computer-readable storage medium storing a computer program, It is 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 11 are implemented.

25. A computer program product comprising a computer program, It is 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 11 are implemented.

Citation Information

Patent Citations

  • Search suggestion clustering and presentation

    CN102687138A

  • Matched text deduplication method and device and electronic equipment

    CN112287655A