Content search method and device, storage medium, and computer equipment

Abstract

The present application discloses a content search method and apparatus, a storage medium, and a computer device, the method comprising: receiving search input information; obtaining target keywords that match the search input information from candidate keywords, wherein the candidate keywords are obtained by performing initial keyword analysis on candidate display content and screening the initial keywords based on the semantic relevance between the initial keywords and the candidate display content; obtaining target display content from the candidate display content based on the target keywords, and displaying search results based on the target display content. The present application extracts keywords based on candidate display content, which helps to obtain richer and more comprehensive candidate keywords used to describe candidate search objects, so that when a user searches, these candidate keywords can be used to recall results, which helps to obtain richer search results and improve the accuracy and comprehensiveness of search results.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and in particular to a content search method and device, a storage medium, and a computer device. Background Art

[0002] With the rapid development of Internet technology, more and more people choose to shop on online platforms. Among them, the online store search function is widely favored by users. Users enter some keywords they want to search for on the platform. The system matches the store name and store dishes based on the keywords entered by the user, determines the search results, and displays a list of stores. The accuracy of this result matching method depends on the merchant's store name and the specified dish name. For example, if a user searches for "barbecue", and a store that sells barbecue does not contain the keyword "barbecue" in its store name and dish name, then the user will not be able to search for this store in this way. Searching in this way using existing technology often results in very few search results, and it is difficult for users to get rich content in each search. Summary of the Invention

[0003] In view of this, the present application provides a content search method and apparatus, a storage medium, and a computer device, which help to obtain richer search results and improve the accuracy and comprehensiveness of search results.

[0004] According to one aspect of the present application, a content search method is provided, the method comprising:

[0005] Receive search input information;

[0006] Obtaining target keywords that match the search input information from candidate keywords, wherein the candidate keywords are obtained by performing an initial keyword analysis on candidate display content and screening the initial keywords based on semantic relevance between the initial keywords and the candidate display content;

[0007] Target display content from the candidate display content is acquired according to the target keyword, and search results are displayed based on the target display content.

[0008] Optionally, before receiving the search input information, the method further includes:

[0009] Obtain at least one candidate display content, and perform keyword analysis on each candidate display content to determine initial keywords;

[0010] For any candidate display content, respectively calculate the semantic relevance between each initial keyword corresponding to the candidate display content and the candidate display content, and filter the initial keywords based on the semantic relevance to determine the tag keyword corresponding to the candidate display content;

[0011] Deduplication of the tag keywords of all candidate display contents is performed to obtain the candidate keywords, and the content to be displayed corresponding to each candidate keyword is marked;

[0012] Accordingly, the step of obtaining target display content from the candidate display content based on the target keyword specifically includes:

[0013] Obtain the to-be-displayed content annotation information corresponding to the target keyword as the target display content.

[0014] Optionally, performing keyword analysis on each candidate display content to determine initial keywords specifically includes:

[0015] Perform entity word recognition on each candidate display content to obtain a first keyword; and / or,

[0016] Perform keyword recognition on each candidate display content, and perform keyword screening based on the frequency of occurrence of the recognized keyword in a preset corpus to obtain a second keyword; and / or,

[0017] Identifying the text to be extracted that contains a preset symbol in each candidate display content, and performing character extraction on the text to be extracted to obtain a third keyword; and / or,

[0018] Obtaining a fourth keyword that is hit by each candidate display content according to a preset keyword database;

[0019] Deduplication processing is performed on the first keyword and / or the second keyword and / or the third keyword and / or the fourth keyword corresponding to each candidate display content to determine the initial keyword corresponding to each candidate display content.

[0020] Optionally, for any candidate display content, respectively calculating the semantic relevance between each initial keyword corresponding to the any candidate display content and the any candidate display content specifically includes:

[0021] Calculating a first semantic representation of any candidate display content and a second semantic representation of each initial keyword corresponding to any candidate display content;

[0022] The semantic relevance between the first semantic representation and each of the second semantic representations is calculated respectively through a preset multi-layer perceptron, wherein the multi-layer perceptron is configured as a leaky relu activation function.

[0023] Optionally, after receiving the search input information, the method further includes:

[0024] Obtaining a search keyword corresponding to the search input information, and expanding the search keyword according to a preset keyword expansion word list;

[0025] Accordingly, obtaining a target keyword matching the search input information from the candidate keywords specifically includes:

[0026] Among the candidate keywords, a target keyword matching the expanded search keyword is obtained.

[0027] Optionally, the step of expanding the search keyword according to a preset keyword expansion word list specifically includes:

[0028] Obtaining synonymous keywords corresponding to the search keyword in a preset synonym word list, and screening the synonymous keywords according to a preset original word word list to obtain a first expanded word that hits the preset original word word list;

[0029] Obtaining a second expanded word hit by the search keyword from the preset original word list, and obtaining a third expanded word corresponding to the search keyword from the preset expanded word list;

[0030] The first expanded word, the second expanded word, and the third expanded word are merged and deduplicated to obtain an expanded search keyword.

[0031] Optionally, the candidate display content includes user-generated content (UGC) for the candidate store; and receiving search input information specifically includes:

[0032] Receive search input information in the store search page;

[0033] Accordingly, the display of search results based on the target display content specifically includes:

[0034] A target store corresponding to the target display content is determined, and the target store and the target display content are displayed.

[0035] According to another aspect of the present application, a content search device is provided, the device comprising:

[0036] An information receiving module, configured to receive search input information;

[0037] a keyword acquisition module, configured to acquire target keywords matching the search input information from candidate keywords, wherein the candidate keywords are obtained by performing an initial keyword analysis on candidate display content and screening the initial keywords based on semantic relevance between the initial keywords and the candidate display content;

[0038] The result display module is used to obtain target display content from the candidate display content according to the target keyword, and display search results based on the target display content.

[0039] Optionally, the candidate display content includes user-generated content (UGC) for the candidate store;

[0040] The information receiving module is specifically used to: receive search input information in the store search page;

[0041] Correspondingly, the result display module is specifically configured to: determine a target store corresponding to the target display content, and display the target store and the target display content.

[0042] Optionally, the device further includes: a keyword construction module, configured to:

[0043] Obtain at least one candidate display content, and perform keyword analysis on each candidate display content to determine initial keywords;

[0044] For any candidate display content, respectively calculate the semantic relevance between each initial keyword corresponding to the candidate display content and the candidate display content, and filter the initial keywords based on the semantic relevance to determine the tag keyword corresponding to the candidate display content;

[0045] Deduplication of the tag keywords of all candidate display contents is performed to obtain the candidate keywords, and the content to be displayed corresponding to each candidate keyword is marked;

[0046] Correspondingly, the result display module is specifically configured to obtain the annotation information of the content to be displayed corresponding to the target keyword as the target display content.

[0047] Optionally, the keyword building module is further used to:

[0048] Perform entity word recognition on each candidate display content to obtain a first keyword; and / or,

[0049] Perform keyword recognition on each candidate display content, and perform keyword screening based on the frequency of occurrence of the recognized keyword in a preset corpus to obtain a second keyword; and / or,

[0050] Identifying the text to be extracted that contains a preset symbol in each candidate display content, and performing character extraction on the text to be extracted to obtain a third keyword; and / or,

[0051] Obtaining a fourth keyword that is hit by each candidate display content according to a preset keyword database;

[0052] Deduplication processing is performed on the first keyword and / or the second keyword and / or the third keyword and / or the fourth keyword corresponding to each candidate display content to determine the initial keyword corresponding to each candidate display content.

[0053] Optionally, the keyword building module is further used to:

[0054] Calculating a first semantic representation of any candidate display content and a second semantic representation of each initial keyword corresponding to any candidate display content;

[0055] The semantic relevance between the first semantic representation and each of the second semantic representations is calculated respectively through a preset multi-layer perceptron, wherein the multi-layer perceptron is configured as a leaky relu activation function.

[0056] Optionally, the device further comprises:

[0057] A search information expansion module is configured to, after receiving the search input information, obtain a search keyword corresponding to the search input information and expand the search keyword according to a preset keyword expansion word list;

[0058] Correspondingly, the keyword acquisition module is further configured to: acquire, from the candidate keywords, a target keyword that matches the expanded search keyword.

[0059] Optionally, the search information expansion module is specifically configured to:

[0060] Obtaining synonymous keywords corresponding to the search keyword in a preset synonym word list, and screening the synonymous keywords according to a preset original word word list to obtain a first expanded word that hits the preset original word word list;

[0061] Obtaining a second expanded word hit by the search keyword from the preset original word list, and obtaining a third expanded word corresponding to the search keyword from the preset expanded word list;

[0062] The first expanded word, the second expanded word, and the third expanded word are merged and deduplicated to obtain an expanded search keyword.

[0063] According to another aspect of the present application, a storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the above-mentioned content search method is implemented.

[0064] According to another aspect of the present application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor implements the above-mentioned content search method when executing the program.

[0065] By means of the above technical solution, the present application provides a content search method and apparatus, storage medium, and computer equipment, which performs initial keyword analysis on candidate display content in advance, and filters the initial keywords based on the semantic relevance between the initial keywords and the candidate display content to obtain candidate keywords, so that when a user performs a content search, the target keywords that match the search input information are obtained from the candidate keywords, and then the search results are determined and displayed based on the target display content corresponding to the target keywords and the objects such as the stores described. The embodiment of the present application performs keyword extraction based on the candidate display content, which helps to obtain richer and more comprehensive candidate keywords used to describe the candidate search objects, so that when the user searches, these candidate keywords are used to recall results, which helps to obtain richer search results and improve the accuracy and comprehensiveness of the search results.

[0066] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0068] Figure 1 A flowchart of a content search method provided by an embodiment of the present application is shown;

[0069] Figure 2 A schematic diagram showing a flow chart of another content search method provided in an embodiment of the present application is shown;

[0070] Figure 3 A flowchart of a search keyword expansion method provided in an embodiment of the present application is shown;

[0071] Figure 4 A schematic structural diagram of a content search device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0072] The present application will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that, unless there is a conflict, the embodiments and features in the embodiments of the present application can be combined with each other.

[0073] In this embodiment, a content search method is provided, such as Figure 1 As shown, the method includes:

[0074] Step 101, receiving search input information;

[0075] Step 102: obtaining target keywords matching the search input information from candidate keywords, wherein the candidate keywords are obtained by performing an initial keyword analysis on candidate display content and screening the initial keywords based on semantic relevance between the initial keywords and the candidate display content;

[0076] Step 103 : acquiring target display content from the candidate display content according to the target keyword, and displaying search results based on the target display content.

[0077] In an embodiment of the present application, the candidate display content may include store introduction content provided by merchants and store introduction content created by Internet users, such as the store exploration experience shared by users on the platform after visiting the store. Of course, the objects described by the candidate display content are not limited to stores, but may also include offline scene objects such as parks, scenic spots, and libraries, or may also include online scene objects such as VR. This application only uses stores as an example for illustration. In order to enrich search results and improve search accuracy, in an embodiment of the present application, the candidate display content of each store is pre-processed in advance, and keywords that can be used to describe the store, i.e., candidate keywords, are extracted based on the candidate display content. When the user searches for content, these candidate keywords can be used to match the search information entered by the user, thereby avoiding the store names and dish names provided by the merchant not being in line with the consumer's search habits, resulting in the store not being searchable.

[0078] In an embodiment of the present application, optionally, the candidate display content includes user-generated content (UGC) for candidate objects, and the candidate objects include candidate stores.

[0079] Taking the candidate display content as the original content of Internet users as an example, when refining candidate keywords, we first conduct an initial keyword analysis on the candidate display content. Specifically, we can preliminarily extract the initial keywords from the candidate display content through entity word recognition, keyword extraction, vocabulary matching, etc., and then calculate the semantic relevance between each initial keyword and the candidate display content. In this way, we can screen out candidate keywords with high semantic relevance to the candidate display content from the initial keywords. Since the candidate display content is also content created by ordinary consumers, these extracted candidate keywords can be used as keywords that are closer to consumers' expression habits and can describe the store.

[0080] After extracting candidate keywords, when a user performs a content search, the matching degree between each candidate keyword and the search input information can be calculated, and the target keyword with the highest matching degree with the search input information can be obtained. Furthermore, based on the target keyword, the display content described by the target keyword can be determined. In other words, based on the correspondence between the candidate keywords and the candidate display content, the target display content corresponding to the target keyword can be determined. The target display content can then be directly displayed as a search result, or the store described by the target display content can be displayed as a search result. This results in richer and more accurate search results.

[0081] By applying the technical solution of this embodiment, by pre-analyzing the candidate display content with initial keywords, and screening the initial keywords based on the semantic relevance between the initial keywords and the candidate display content to obtain candidate keywords, when the user searches for content, the target keywords that match the search input information are obtained from the candidate keywords, and then the search results are determined and displayed based on the target display content corresponding to the target keywords and the objects such as stores described. The embodiment of the present application performs keyword extraction based on the candidate display content, which helps to obtain richer and more comprehensive candidate keywords used to describe the candidate search objects, so that when the user searches, these candidate keywords are used to recall results, which helps to obtain richer search results and improve the accuracy and comprehensiveness of the search results.

[0082] In the embodiment of the present application, optionally, the candidate keywords are obtained by the following method:

[0083] S1, obtaining at least one candidate display content, and performing keyword analysis on each candidate display content to determine initial keywords;

[0084] S2, for any candidate display content, respectively calculating the semantic relevance between each initial keyword corresponding to the candidate display content and the candidate display content, and filtering the initial keywords based on the semantic relevance to determine a tag keyword corresponding to the candidate display content;

[0085] S3, performing deduplication processing on the tag keywords of all candidate display contents to obtain the candidate keywords, and marking the content to be displayed corresponding to each candidate keyword.

[0086] In the above embodiment, taking the user-generated content (UGC) of the candidate store as an example, for each candidate display content, keyword analysis is first performed to obtain the initial keywords describing the candidate display content, and then the semantic relevance of each initial keyword with the candidate display content is calculated, and a group of tag keywords with higher semantic relevance to the candidate display content are screened based on the semantic relevance. These tag keywords can be used as tags to describe the candidate display content. Finally, all the tag keywords extracted from each candidate display content are summarized and counted, and repeated tag keywords are removed. For example, the tag keyword "barbecue" appears 3 times, corresponding to candidate display content 1, candidate display content 2, and candidate display content 3 respectively. After deduplication, the 3 tag keywords "barbecue" are merged into 1 candidate keyword "barbecue", and the candidate keyword "barbecue" is marked with the corresponding selected display content 1, candidate display content 2, and candidate display content 3, or the candidate keyword "barbecue" is marked with the store logo corresponding to selected display content 1, candidate display content 2, and candidate display content 3.

[0087] In the embodiment of the present application, optionally, S1 specifically includes:

[0088] S1-1, performing entity word recognition on each candidate display content to obtain a first keyword; and / or, performing keyword recognition on each candidate display content, and performing keyword screening based on the frequency of occurrence of the recognized keyword in a preset corpus to obtain a second keyword; and / or, identifying a text to be extracted containing a preset symbol in each candidate display content, and performing character extraction on the text to be extracted to obtain a third keyword; and / or, obtaining a fourth keyword that is hit by each candidate display content based on a preset keyword lexicon;

[0089] S1-2, respectively perform deduplication processing on the first keyword and / or the second keyword and / or the third keyword and / or the fourth keyword corresponding to each candidate display content to determine the initial keyword corresponding to each candidate display content.

[0090] In this embodiment, the initial keywords of the candidate display content can be obtained through various methods, such as entity word recognition, keyword extraction, pattern matching, and vocabulary matching. After obtaining the keywords (i.e., the first keyword, the second keyword, the third keyword, and the fourth keyword) through different methods, the keywords corresponding to the candidate display content obtained through the different methods are deduplicated to obtain the initial keywords corresponding to the candidate display content.

[0091] Entity word recognition method: Entity word recognition mainly identifies category words and category modifiers in the text, such as entity words such as pizza and Western food. The specific steps include: (1) segmenting the candidate display content and removing stop words; (2) identifying entity sub-words for the phrases obtained in the previous step. For example, for solutions applied in e-commerce platform scenarios, call the entity word recognition plug-in in the e-commerce mode; (3) filter out unnecessary entity word types through custom feature UDF quality inspection and obtain the first keyword.

[0092] Keyword extraction method: Keyword extraction is to identify the more "important" words in the text. The importance of a word increases proportionally with the number of times it appears in the text, but at the same time decreases inversely with the frequency of its appearance in the corpus. The specific steps include: (1) segmenting the candidate display content and removing the stop words; (2) In this embodiment, the candidate display content is first identified by keywords, and then the identified keywords are filtered according to the frequency of the keywords in the corpus, for example, by using tf-idf and textrank to identify keywords, and then calling the entity word recognition plug-in to identify the noun-type keywords, and finally calculating the frequency of the keywords in the preset corpus to filter out the keywords with higher frequencies; (3) UDF quality inspection and filtering: Use udf to filter out keywords with higher scores to obtain the second keyword.

[0093] Pattern matching method: mainly to identify patterns that are prone to product keywords, such as " " [ ]. The specific steps are: (1) Match the words in " ". For example, "crab roe tofu" matches crab roe tofu, and [signature shrimp dumplings] matches signature shrimp dumplings. (2) UDF quality inspection and filtering: remove words such as "city", "recommended dishes", "store environment", and "must-order signature dishes" that often appear in this pattern, because these words are irrelevant to the topic, and remove English + numbers to obtain the third keyword.

[0094] Thesaurus matching method: Through thesaurus matching, find the product words and category words of the preset categories in the candidate display content, such as catering and life service categories. The specific steps are: (1) Establish a thesaurus for the preset category (i.e., a preset keyword thesaurus), such as the catering + life service thesaurus: construct the words with more than 3 characters in the catering thesaurus that have been simply filtered into the catering thesaurus as the catering thesaurus, and construct the life service words of the top category + content into the life service thesaurus. (2) Use the cleaned candidate display content to directly match the thesaurus, and obtain the keywords that hit the candidate display content as the fourth keyword.

[0095] In the embodiment of the present application, optionally, the step of “respectively calculating the semantic relevance between each initial keyword corresponding to any candidate display content and any candidate display content” in S2 specifically includes:

[0096] S2-1, calculating a first semantic representation of any candidate display content and a second semantic representation of each initial keyword corresponding to any candidate display content;

[0097] S2-2, respectively calculating the semantic relevance between the first semantic representation and each of the second semantic representations through a preset multilayer perceptron, wherein the multilayer perceptron is configured as a leaky relu activation function.

[0098] In this embodiment, for each candidate display content, the BERT model (Bidirectional Encoder Representation from Transformers, a pre-trained language representation model) can be first used to calculate the first semantic representation corresponding to the candidate display content (specifically, the candidate display content + the corresponding store name can be calculated) and the second semantic representation of each initial keyword corresponding to the candidate display content, and then the multi-layer perceptron MLP (Multilayer Perceptron is a feedforward artificial neural network model) is used to calculate the semantic relevance between each second semantic representation and the first semantic representation, so as to determine the relevance of each initial keyword with the candidate display content, so as to subsequently screen out several initial keywords with a high relevance to the candidate display content as candidate keywords for the candidate display content. Among them, in order to increase the difference of each initial keyword and make the calculated correlation value distribution more dispersed, which is convenient for better selection of candidate keywords, the three-layer neural network part of the multilayer perceptron is configured with a leaky relu activation function to increase the difference of the output logits value of the three-layer neural network part, and the temperature parameter temperature parameter of the sigmoid function of the output layer of the multilayer perceptron is set to a preset value. When the temperature parameter is large, that is, when it tends to positive infinity, the activation probabilities corresponding to all activation values tend to be the same (the difference in activation probabilities is small); and when the temperature parameter is very low, that is, when it tends to 0, the difference in activation probabilities corresponding to different activation values is also greater. For example, the temperature parameter is set to 0.5, so that the difference in semantic correlation values obtained after the logits values are input to the output layer is further increased.

[0099] Furthermore, as a refinement and expansion of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, another content search method is provided, which uses the above candidate keywords to search for content, such as Figure 2 As shown, the method includes:

[0100] Step 201: Receive search input information in a store search page.

[0101] Step 202: Acquire a search keyword corresponding to the search input information, and expand the search keyword according to a preset keyword expansion word list.

[0102] In this embodiment, taking the "Store Search" function as an example, the user can enter search information on the store search page, which can be a sentence or a phrase. After receiving the search input information, the search input information is first pre-processed to extract search keywords from the search input information. In order to obtain richer and more comprehensive content that meets the user's search intent, the search keywords can be further expanded to obtain keywords related to the search keywords, so that the expanded keywords that are richer and more in line with the user's search intent can be used to recall results.

[0103] In the embodiment of the present application, optionally, in step 202, “expanding the search keyword according to a preset keyword expansion word list” specifically includes:

[0104] Step 202-1, obtaining synonymous keywords corresponding to the search keyword from a preset synonym word list, and screening the synonymous keywords according to a preset original word list to obtain the first expanded word that hits the preset original word list;

[0105] Step 202-2, obtaining a second expanded word hit by the search keyword from the preset original word list, and obtaining a third expanded word corresponding to the search keyword from the preset expanded word list;

[0106] Step 202-3: merge and remove duplicates of the first expanded word, the second expanded word, and the third expanded word to obtain an expanded search keyword.

[0107] In this embodiment, a preset synonym word list A, a preset original word list B and a preset extended word list C are pre-constructed. Among them, the preset synonym word list A refers to the synonyms corresponding to a word, such as meat crab pot: crab pot, crab meat pot. The original word corresponding to a word in the preset original word list B is the word itself, such as meat crab pot: meat crab pot; crab pot: crab pot. The preset extended word list C refers to the hyponym corresponding to a word, or the recommended word corresponding to the cuisine word, such as meat crab pot: chicken feet and meat crab pot, small meat crab pot, spicy meat crab pot, chicken feet and meat crab pot, chicken crab pot; Beijing cuisine: Beijing cuisine, Beijing roast duck, stir-fried lamb with green onions, stir-fried diced chicken with sauce, Beijing sauce shredded pork, fried sausage. For example Figure 3As shown, taking the search keyword "meat crab pot" as an example, first, the synonymous keywords "crab pot" and "crab meat pot" of "meat crab pot" are obtained in the preset synonym word list A, and the preset original word word list B is queried whether the synonymous keywords "crab pot" and "crab meat pot" are included. If they are included, it is determined as the first expanded word. If not, the synonymous keyword is deleted; secondly, the preset original word word list B is queried whether the search keyword "meat crab pot" is included. If it is included, the search keyword is used as the second expanded word. If not, the search keyword is deleted; then, the third expanded word corresponding to the search keyword is obtained in the preset expanded word word list C; finally, the first expanded word, the second expanded word, and the third expanded word are merged and deduplicated to achieve the expansion of the search keyword.

[0108] Step 203: Obtain target keywords that match the expanded search keywords from the candidate keywords.

[0109] Step 204 : obtaining the tag information of the content to be displayed corresponding to the target keyword as the target display content, determining the target store corresponding to the target display content, and displaying the target store and the target display content.

[0110] In this embodiment, after the search keyword is expanded, the target keyword hit by the expanded search keyword is obtained from the candidate keywords. Since the candidate keyword is labeled with the corresponding candidate store, after determining the target keyword, the content to be displayed corresponding to the target keyword can be determined based on the labeling information corresponding to the target keyword, and used as the target display content. The target store described by the target display content can also be determined, and the target store and target display content can be displayed. For example, a user-generated content "Delicious Barbecue" describes that "the barbecue at a certain store is very delicious, with a delicate taste, and the Orleans grilled wings are very authentic." The corresponding candidate keywords extracted from this content include "Barbecue" and "Orleans grilled wings." The labeling information corresponding to the candidate keywords includes this user-generated content. If the search keyword hits "Barbecue", the target display content "Delicious Barbecue" can be determined based on the labeling information corresponding to the hit target keyword "Barbecue". The name of the store described in the article "Delicious Barbecue" can also be obtained, and the store and the article "Delicious Barbecue" can be displayed on the result display page. Of course, only the target store or only the target display content can be displayed.

[0111] Further, as Figure 1 The specific implementation of the method, the embodiment of the present application provides a content search device, such as Figure 4 As shown, the device includes:

[0112] An information receiving module, configured to receive search input information;

[0113] a keyword acquisition module, configured to acquire target keywords matching the search input information from candidate keywords, wherein the candidate keywords are obtained by performing an initial keyword analysis on candidate display content and screening the initial keywords based on semantic relevance between the initial keywords and the candidate display content;

[0114] The result display module is used to obtain target display content from the candidate display content according to the target keyword, and display search results based on the target display content.

[0115] Optionally, the device further includes: a keyword construction module, configured to:

[0116] Obtain at least one candidate display content, and perform keyword analysis on each candidate display content to determine initial keywords;

[0117] For any candidate display content, respectively calculate the semantic relevance between each initial keyword corresponding to the candidate display content and the candidate display content, and filter the initial keywords based on the semantic relevance to determine the tag keyword corresponding to the candidate display content;

[0118] Deduplication of the tag keywords of all candidate display contents is performed to obtain the candidate keywords, and the content to be displayed corresponding to each candidate keyword is marked;

[0119] Correspondingly, the result display module is specifically configured to obtain the annotation information of the content to be displayed corresponding to the target keyword as the target display content.

[0120] Optionally, the keyword building module is further used to:

[0121] Perform entity word recognition on each candidate display content to obtain a first keyword; and / or,

[0122] Perform keyword recognition on each candidate display content, and perform keyword screening based on the frequency of occurrence of the recognized keyword in a preset corpus to obtain a second keyword; and / or,

[0123] Identifying the text to be extracted that contains a preset symbol in each candidate display content, and performing character extraction on the text to be extracted to obtain a third keyword; and / or,

[0124] Obtaining a fourth keyword that is hit by each candidate display content according to a preset keyword database;

[0125] Deduplication processing is performed on the first keyword and / or the second keyword and / or the third keyword and / or the fourth keyword corresponding to each candidate display content to determine the initial keyword corresponding to each candidate display content.

[0126] Optionally, the keyword building module is further used to:

[0127] Calculating a first semantic representation of any candidate display content and a second semantic representation of each initial keyword corresponding to any candidate display content;

[0128] The semantic relevance between the first semantic representation and each of the second semantic representations is calculated respectively through a preset multi-layer perceptron, wherein the multi-layer perceptron is configured as a leaky relu activation function.

[0129] Optionally, the device further comprises:

[0130] A search information expansion module is configured to, after receiving the search input information, obtain a search keyword corresponding to the search input information and expand the search keyword according to a preset keyword expansion word list;

[0131] Correspondingly, the keyword acquisition module is further configured to: acquire, from the candidate keywords, a target keyword that matches the expanded search keyword.

[0132] Optionally, the search information expansion module is specifically configured to:

[0133] Obtaining synonymous keywords corresponding to the search keyword in a preset synonym word list, and screening the synonymous keywords according to a preset original word word list to obtain a first expanded word that hits the preset original word word list;

[0134] Obtaining a second expanded word hit by the search keyword from the preset original word list, and obtaining a third expanded word corresponding to the search keyword from the preset expanded word list;

[0135] The first expanded word, the second expanded word, and the third expanded word are merged and deduplicated to obtain an expanded search keyword.

[0136] Optionally, the candidate display content includes user-generated content (UGC) for the candidate store;

[0137] The information receiving module is specifically used to: receive search input information in the store search page;

[0138] Correspondingly, the result display module is specifically configured to: determine a target store corresponding to the target display content, and display the target store and the target display content.

[0139] It should be noted that for other corresponding descriptions of the functional units involved in the content search device provided in the embodiment of the present application, please refer to Figures 1 to 3 The corresponding description in the method will not be repeated here.

[0140] Based on the above Figures 1 to 3 The method shown in FIG. 1 is a method for performing the above-mentioned operation. Accordingly, the embodiment of the present application further provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned operation is performed. Figures 1 to 3 Content search method shown.

[0141] Based on this understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, USB flash drive, mobile hard disk, etc.), including a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in each implementation scenario of the present application.

[0142] Based on the above Figures 1 to 3 The method shown, and Figure 4 In order to achieve the above-mentioned purpose, the embodiment of the present application further provides a computer device, which can be a personal computer, a server, a network device, etc. The computer device includes a storage medium and a processor; the storage medium is used to store a computer program; the processor is used to execute the computer program to achieve the above-mentioned Figures 1 to 3 Content search method shown.

[0143] Optionally, the computer device may further include a user interface, a network interface, a camera, a radio frequency (RF) circuit, a sensor, an audio circuit, a Wi-Fi module, etc. The user interface may include a display, an input unit such as a keyboard, etc., and the optional user interface may also include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface, a wireless interface (such as a Bluetooth interface, a Wi-Fi interface), etc.

[0144] Those skilled in the art will understand that the computer device structure provided in this embodiment does not constitute a limitation on the computer device, and may include more or fewer components, or a combination of certain components, or different component arrangements.

[0145] The storage medium may also include an operating system and a network communication module. An operating system is a program that manages and stores the hardware and software resources of a computer device, supporting the execution of information processing programs and other software and / or programs. The network communication module facilitates communication between components within the storage medium, as well as with other hardware and software within the physical device.

[0146] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform, or can be implemented by hardware by pre-analyzing the candidate display content with initial keywords, and screening the initial keywords based on the semantic relevance between the initial keywords and the candidate display content to obtain candidate keywords, so that when the user searches for content, the target keywords that match the search input information are obtained from the candidate keywords, and then the search results are determined and displayed based on the target display content corresponding to the target keywords and the objects such as the described shops. The embodiment of the present application performs keyword extraction based on the candidate display content, which helps to obtain richer and more comprehensive candidate keywords used to describe the candidate search objects, so that when the user searches, these candidate keywords are used to recall the results, which helps to obtain richer search results and improve the accuracy and comprehensiveness of the search results.

[0147] Those skilled in the art will understand that the accompanying drawings are only schematic diagrams of a preferred implementation scenario, and the modules or processes in the accompanying drawings are not necessarily required to implement the present application. Those skilled in the art will understand that the modules in the devices in the implementation scenario can be distributed in the devices of the implementation scenario according to the implementation scenario description, or can be changed accordingly and located in one or more devices different from the implementation scenario. The modules of the above-mentioned implementation scenario can be combined into one module, or can be further split into multiple sub-modules.

[0148] The serial numbers of the above application are for descriptive purposes only and do not represent the advantages or disadvantages of the implementation scenarios. The above disclosure only discloses several specific implementation scenarios of the present application, but the present application is not limited thereto. Any changes that can be conceived by those skilled in the art should fall within the scope of protection of the present application.

Claims

1. A content search method, characterized in that: The method comprises: Receive search input information, obtain a search keyword corresponding to the search input information; obtain a synonym keyword corresponding to the search keyword from a preset synonym word list, and obtain a first expanded word of the synonym keyword that hits a preset original word list; obtain a second expanded word hit by the search keyword from the preset original word list, and obtain a third expanded word corresponding to the search keyword from a preset expanded word list; merge and remove duplicates from the first expanded word, the second expanded word, and the third expanded word to obtain an expanded search keyword; Obtaining target keywords that match the expanded search keywords from candidate keywords, wherein the candidate keywords are obtained by performing initial keyword analysis on candidate display content and screening the initial keywords based on semantic relevance between the initial keywords and the candidate display content; Target display content from the candidate display content is acquired according to the target keyword, and search results are displayed based on the target display content.

2. The method according to claim 1, characterized in that The candidate display content includes user-generated content (UGC) for candidate objects, and the candidate objects include candidate stores. The receiving of search input information specifically includes: Receive search input information in the store search page; Accordingly, the display of search results based on the target display content specifically includes: A target store corresponding to the target display content is determined, and the target store and the target display content are displayed.

3. The method according to claim 1, characterized in that Before receiving the search input information, the method further includes: Obtain at least one candidate display content, and perform keyword analysis on each candidate display content to determine initial keywords; For any candidate display content, respectively calculate the semantic relevance between each initial keyword corresponding to the candidate display content and the candidate display content, and filter the initial keywords based on the semantic relevance to determine the tag keyword corresponding to the candidate display content; Deduplication of the tag keywords of all candidate display contents is performed to obtain the candidate keywords, and the content to be displayed corresponding to each candidate keyword is marked; Accordingly, the step of obtaining target display content from the candidate display content based on the target keyword specifically includes: Obtain the to-be-displayed content annotation information corresponding to the target keyword as the target display content.

4. The method according to claim 3, characterized in that The performing keyword analysis on each candidate display content to determine the initial keyword specifically includes: Perform entity word recognition on each candidate display content to obtain a first keyword; and / or, Perform keyword recognition on each candidate display content, and perform keyword screening based on the frequency of occurrence of the recognized keyword in a preset corpus to obtain a second keyword; and / or, Identifying the text to be extracted that contains a preset symbol in each candidate display content, and performing character extraction on the text to be extracted to obtain a third keyword; and / or, Obtaining a fourth keyword that is hit by each candidate display content according to a preset keyword database; Deduplication processing is performed on the first keyword and / or the second keyword and / or the third keyword and / or the fourth keyword corresponding to each candidate display content to determine the initial keyword corresponding to each candidate display content.

5. The method according to claim 3, characterized in that The calculating, for any candidate display content, respectively the semantic relevance between each initial keyword corresponding to the any candidate display content and the any candidate display content specifically includes: Calculating a first semantic representation of any candidate display content and a second semantic representation of each initial keyword corresponding to any candidate display content; The semantic relevance between the first semantic representation and each of the second semantic representations is calculated respectively through a preset multi-layer perceptron, wherein the multi-layer perceptron is configured as a leaky relu activation function.

6. A content search device, characterized in that: The device comprises: An information receiving module, configured to receive search input information; A search information expansion module is configured to obtain a search keyword corresponding to the search input information, obtain a synonymous keyword corresponding to the search keyword in a preset synonym word list, and obtain a first expanded word among the synonymous keywords that hits a preset original word list; obtain a second expanded word hit by the search keyword in the preset original word list, and obtain a third expanded word corresponding to the search keyword in a preset expanded word list; merge and remove duplicates from the first expanded word, the second expanded word, and the third expanded word to obtain an expanded search keyword; a keyword acquisition module, configured to acquire target keywords matching the expanded search keywords from candidate keywords, wherein the candidate keywords are obtained by performing an initial keyword analysis on candidate display content and screening the initial keywords based on semantic relevance between the initial keywords and the candidate display content; The result display module is used to obtain target display content from the candidate display content according to the target keyword, and display search results based on the target display content.

7. The device according to claim 6, characterized in that The candidate display content includes user-generated content (UGC) for the candidate store; The information receiving module is specifically used to: receive search input information in the store search page; Correspondingly, the result display module is specifically configured to: determine a target store corresponding to the target display content, and display the target store and the target display content.

8. The device according to claim 6, characterized in that The device further includes a keyword building module, configured to: Obtain at least one candidate display content, and perform keyword analysis on each candidate display content to determine initial keywords; For any candidate display content, respectively calculate the semantic relevance between each initial keyword corresponding to the candidate display content and the candidate display content, and filter the initial keywords based on the semantic relevance to determine the tag keyword corresponding to the candidate display content; Deduplication of the tag keywords of all candidate display contents is performed to obtain the candidate keywords, and the content to be displayed corresponding to each candidate keyword is marked; Correspondingly, the result display module is specifically configured to obtain the annotation information of the content to be displayed corresponding to the target keyword as the target display content.

9. The device according to claim 8, characterized in that The keyword building module is also used to: Perform entity word recognition on each candidate display content to obtain a first keyword; and / or, Perform keyword recognition on each candidate display content, and perform keyword screening based on the frequency of occurrence of the recognized keyword in a preset corpus to obtain a second keyword; and / or, Identifying the text to be extracted that contains a preset symbol in each candidate display content, and performing character extraction on the text to be extracted to obtain a third keyword; and / or, Obtaining a fourth keyword that is hit by each candidate display content according to a preset keyword database; Deduplication processing is performed on the first keyword and / or the second keyword and / or the third keyword and / or the fourth keyword corresponding to each candidate display content to determine the initial keyword corresponding to each candidate display content.

10. The device according to claim 8, characterized in that The keyword building module is also used to: Calculating a first semantic representation of any candidate display content and a second semantic representation of each initial keyword corresponding to any candidate display content; The semantic relevance between the first semantic representation and each of the second semantic representations is calculated respectively through a preset multi-layer perceptron, wherein the multi-layer perceptron is configured as a leaky relu activation function.

11. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the content search method according to any one of claims 1 to 5 is implemented.

12. A computer device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, wherein: When the processor executes the computer program, the content search method according to any one of claims 1 to 5 is implemented.

Images