Information search method and device, electronic equipment, storage medium and program product

Abstract

The invention discloses an information searching method and device, electronic equipment, a storage medium and a program product. According to the embodiment of the invention, a to-be-recommended search result associated with to-be-searched information can be obtained; recalling the to-be-recommended search results based on a plurality of preset recall strategies to obtain recalled recommended search results corresponding to the preset recall strategies; calculating the relevancy between the recall recommendation search result and the to-be-searched information; and according to the relevancy and a relevancy range specified by a preset relevancy level, determining a relevancy level corresponding to the recall recommendation search results, and then based on relevancy level information corresponding to each preset search scene, determining a target recall recommendation search result corresponding to each preset search scene from the recall recommendation search results. In the embodiment of the invention, targeted and comprehensive search experience is ensured to be provided under different situations, and search results of corresponding relevancy levels can be provided for different search scenes. Therefore, the scheme can improve the search experience of the user.

Description

Technical Field

[0001] The present application relates to the field of computers, and in particular to an information search method, device, electronic device, storage medium and program product. Background Art

[0002] Today's information search has a wide range of information coverage, and can index and provide content in various forms including web pages, pictures, texts, short videos, etc. These contents are huge in volume and quantity, providing users with rich and diverse resources to meet their needs for obtaining the information they need in different fields, topics and interests. Whether it is academic research, news information, entertainment media or practical skills, users can obtain comprehensive and multi-dimensional knowledge and answers through information search.

[0003] However, the current information search model mainly uses literal text matching as the key basis, which has certain limitations when dealing with semantically similar but not identical situations. This means that some relevant results may be excluded, which affects the accuracy and completeness of the search results. Summary of the invention

[0004] The embodiments of the present application provide an information search method, device, electronic device, storage medium and program product, which can improve the user's search experience.

[0005] The present application provides an information search method, including:

[0006] Obtaining information to be searched and search results to be recommended associated with the information to be searched;

[0007] Based on multiple preset recall strategies, multi-way recall processing is performed on the recommended search results to obtain the recalled recommended search results corresponding to each preset recall strategy;

[0008] Calculate the relevance between the recalled recommended search results and the information to be searched;

[0009] Determining the relevance level corresponding to the recalled recommended search results according to the relevance and the relevance range specified by the preset relevance level;

[0010] Based on the relevance level corresponding to the recall recommendation search results and the relevance level information corresponding to each preset search scenario, a target recall recommendation search result corresponding to each preset search scenario is determined from the recall recommendation search results.

[0011] The present application also provides an information search device, including:

[0012] An information acquisition unit, used to acquire information to be searched and search results to be recommended associated with the information to be searched;

[0013] A recall unit, used to perform multi-way recall processing on the recommended search results based on multiple preset recall strategies, and obtain the recalled recommended search results corresponding to each preset recall strategy;

[0014] A calculation unit, used to calculate the relevance between the recalled recommended search results and the information to be searched;

[0015] A level determination unit, used to determine the relevance level corresponding to the recalled recommendation search result according to the relevance and the relevance range specified by the preset relevance level;

[0016] The scenario recommendation unit is used to determine the target recall recommendation search results corresponding to each preset search scenario from the recall recommendation search results based on the relevance level corresponding to the recall recommendation search results and the relevance level information corresponding to each preset search scenario.

[0017] In some embodiments, the information acquisition unit includes an instruction receiving unit;

[0018] An instruction receiving unit, used to receive a service search instruction and obtain information to be searched indicated by the service search instruction, wherein the service search instruction corresponds to a target search scenario;

[0019] After the scene recommendation unit, a search return unit is also included;

[0020] The search return unit is used to return the target recall recommendation search results corresponding to the target search scenario.

[0021] In some embodiments, the recall unit includes a candidate search unit, a feature acquisition unit, a search ranking unit, and a recall sub-unit;

[0022] A candidate search unit, configured to determine, based on the keywords in the information to be searched, candidate recommended search results that match the keywords from the search results to be recommended;

[0023] A feature acquisition unit, used to acquire attribute features of candidate recommendation search results;

[0024] A search ranking unit, used to sort the candidate recommendation search results based on the attribute characteristics of the candidate recommendation search results to obtain a sorted candidate recommendation search result list;

[0025] The recall subunit is used to perform multi-way recall processing based on multiple preset recall strategies to obtain recall recommendation search results corresponding to each preset recall strategy.

[0026] In some embodiments, an index building unit is also included;

[0027] An index building unit, used to build an index list based on the attribute features of the candidate recommendation search results and the candidate recommendation search results;

[0028] The search ranking unit includes an index ranking unit;

[0029] An index sorting unit, used to sort the candidate recommended search results in the index list based on the attribute features of the candidate recommended search results in the index list to obtain a sorted index list;

[0030] The recall subunit includes an index recall unit;

[0031] The index recall unit is used to perform multi-way recall processing on the candidate recommended search results in the sorted index list based on multiple preset recall strategies to obtain the recalled recommended search results corresponding to each preset recall strategy.

[0032] In some embodiments, the computing unit includes a merging subunit, a deduplication subunit, and a computing subunit;

[0033] A merging subunit is used to merge the recall recommendation search results corresponding to each preset recall strategy to obtain a recall recommendation search result set;

[0034] A deduplication subunit is used to perform deduplication processing on the recall recommendation search result set to obtain a deduplicated recall recommendation search result set;

[0035] The calculation subunit is used to calculate the relevance between the recalled recommendation search results in the recalled recommendation search result set after deduplication and the information to be searched.

[0036] In some embodiments, the calculation subunit includes an indicator determination unit and a weighted fusion unit;

[0037] An indicator determination unit, used to determine each quality evaluation indicator of the recall recommendation search results in the recall recommendation search result set after deduplication based on the information to be searched;

[0038] The weighted fusion unit is used to perform weighted fusion processing on various quality evaluation indicators of the recalled recommendation search results to obtain the relevance between the recalled recommendation search results and the information to be searched.

[0039] An embodiment of the present application also provides an electronic device, including a processor and a memory, wherein the memory stores multiple instructions; the processor loads instructions from the memory to execute the steps in any one of the information search methods provided in the embodiments of the present application.

[0040] An embodiment of the present application also provides a computer-readable storage medium, which stores a plurality of instructions, and the instructions are suitable for a processor to load to execute the steps in any one of the information search methods provided in the embodiments of the present application.

[0041] An embodiment of the present application also provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of any one of the information search methods provided in the embodiments of the present application.

[0042] The embodiments of the present application can obtain information to be searched and search results to be recommended associated with the information to be searched; based on multiple preset recall strategies, perform multi-way recall processing on the search results to be recommended to obtain recall recommendation search results corresponding to each preset recall strategy; calculate the correlation between the recall recommendation search results and the information to be searched; determine the correlation level corresponding to the recall recommendation search results based on the correlation and the correlation range specified by the preset correlation level; based on the correlation level corresponding to the recall recommendation search results and the correlation level information corresponding to each preset search scenario, determine the target recall recommendation search results corresponding to each preset search scenario from the recall recommendation search results.

[0043] In the present application, by retrieving the information to be searched, the search results to be recommended associated with it can be obtained, and then, by using a variety of preset recall strategies, the recall recommendation search results that meet each preset recall strategy can be selected from the search results to be recommended, thereby providing more comprehensive and diverse recall recommendation search results. The relevance between the recall recommendation search results and the information to be searched is calculated, and according to the relevance range specified by the preset relevance level, the relevance level corresponding to the recall recommendation search results can be obtained. Using the relevance level information corresponding to each preset search scenario, the target recall recommendation search results corresponding to each preset search scenario can be obtained. In this way, the target recall recommendation search results corresponding to each preset search scenario can be planned, thereby ensuring that a targeted and comprehensive search experience can be provided in different scenarios, and search results with corresponding relevance levels can be provided for different search scenarios, which can improve the user's search experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0045] Figure 1ais a scenario diagram of the information search method provided in an embodiment of the present application;

[0046] Figure 1b It is a flowchart of the information search method provided by the embodiment of the present application;

[0047] Figure 2a It is a flowchart of the information search method provided in the embodiment of the present application applied in an instant messaging application scenario;

[0048] Figure 2b It is a search schematic diagram of the information search method provided in the embodiment of the present application applied in a background scenario;

[0049] Figure 2c It is a schematic diagram of constructing a dual candidate pool of the information search method provided in an embodiment of the present application;

[0050] Figure 3 is a schematic diagram of the structure of an information search device provided in an embodiment of the present application;

[0051] Figure 4 It is a schematic diagram of the structure of the server provided in the embodiment of the present application. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0053] Embodiments of the present application provide an information search method, device, electronic device, storage medium, and program product.

[0054] It is understandable that in the specific implementation of the present application, related data such as search results to be recommended are involved. When the embodiments of the present application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0055] The information search device can be integrated into an electronic device, which can be a terminal, a server, or other devices. The terminal can be a mobile phone, a tablet computer, a smart Bluetooth device, a laptop, or a personal computer (PC), etc. The server can be a single server or a server cluster composed of multiple servers.

[0056] In some embodiments, the information search device may also be integrated into multiple electronic devices. For example, the information search device may be integrated into multiple servers, and the information search method of the present application may be implemented by multiple servers.

[0057] In some embodiments, the server may also be implemented in the form of a terminal.

[0058] For example, refer to Figure 1a The electronic device S01 can obtain information to be searched and search results to be recommended associated with the information to be searched; based on multiple preset recall strategies, perform multi-way recall processing on the search results to be recommended to obtain recall recommendation search results corresponding to each preset recall strategy; calculate the relevance between the recall recommendation search results and the information to be searched; determine the relevance level corresponding to the recall recommendation search results according to the relevance and the relevance range specified by the preset relevance level; based on the relevance level corresponding to the recall recommendation search results and the relevance level information corresponding to each preset search scenario, determine the target recall recommendation search results corresponding to each preset search scenario from the recall recommendation search results.

[0059] Among them, by retrieving the information to be searched, the search results to be recommended associated with it can be obtained, and then, using multiple preset recall strategies, the recall recommendation search results that meet each preset recall strategy can be selected from the search results to be recommended, thereby providing more comprehensive and diverse recall recommendation search results. The relevance between the recall recommendation search results and the information to be searched is calculated, and according to the relevance range specified by the preset relevance level, the relevance level corresponding to the recall recommendation search results can be obtained. Using the relevance level information corresponding to each preset search scenario, the target recall recommendation search results corresponding to each preset search scenario can be obtained. In this way, the target recall recommendation search results corresponding to each preset search scenario can be planned, thereby ensuring that a targeted and comprehensive search experience can be provided in different scenarios, and search results with corresponding relevance levels can be provided for different search scenarios, which can improve the user's search experience.

[0060] It should be noted that the serial numbers of the following embodiments are not intended to limit the preferred order of the embodiments.

[0061] Artificial Intelligence (AI) is a technology that uses digital computers to simulate human perception of the environment, acquire knowledge, and use knowledge. This technology can enable machines to have functions similar to human perception, reasoning, and decision-making. The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, and other technologies. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, as well as machine learning / deep learning, autonomous driving, smart transportation, and other major directions.

[0062] Computer Vision (CV) is a technology that uses computers to replace human eyes to identify, measure and further process target images. Computer vision technology usually includes image processing, image recognition, image semantic understanding, image retrieval, virtual reality, augmented reality, simultaneous positioning and map construction, autonomous driving, smart transportation and other technologies, as well as common biometric recognition technologies such as face recognition and fingerprint recognition. For example, image processing technologies such as image coloring and image stroke extraction.

[0063] The key technologies of speech technology include automatic speech recognition technology, speech synthesis technology and voiceprint recognition technology. Enabling computers to listen, see, speak and feel is the development direction of human-computer interaction in the future, among which speech has become one of the most promising human-computer interaction methods in the future.

[0064] Natural language processing (NLP) is an important direction in the fields of computer science and artificial intelligence. It studies various theories and methods that can achieve effective communication between people and computers using natural language. Natural language processing is a science that integrates linguistics, computer science, and mathematics. Therefore, research in this field will involve natural language, that is, the language people use in daily life, so it is closely related to the study of linguistics. Natural language processing technology usually includes text processing, semantic understanding, machine translation, robot question answering, knowledge graph and other technologies.

[0065] With the research and advancement of artificial intelligence technology, artificial intelligence technology has been studied and applied in many fields, such as common smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, unmanned driving, automatic driving, drones, robots, smart medical care, smart customer service, Internet of Vehicles, automatic driving, smart transportation, etc. I believe that with the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role.

[0066] In this embodiment, an information search method involving artificial intelligence is provided, such as Figure 1b As shown, the specific process of the information search method can be as follows:

[0067] 101. Obtain information to be searched and search results to be recommended that are associated with the information to be searched.

[0068] The information to be searched refers to information waiting to be searched. For example, the information to be searched may be a keyword or query content input by a user waiting to be searched, which may be text, voice, picture, video, etc., or a feature vector corresponding to the keyword or query content input by a user waiting to be searched.

[0069] The search results to be recommended are various types of information resources related to the information to be searched. For example, the search results to be recommended can be various types of information resources such as web pages, pictures, videos, news articles, forum posts, etc. related to the information to be searched.

[0070] For example, if the information to be searched is text to be searched, the text feature vector corresponding to the text to be searched can be obtained through natural language processing in artificial intelligence. If the information to be searched is speech to be searched, the speech feature vector corresponding to the speech to be searched can be obtained through language technology in artificial intelligence. If the information to be searched is a picture or video to be searched, the visual feature vector corresponding to the picture or video to be searched can be obtained through computer vision in artificial intelligence.

[0071] Web page feature vectors, picture feature vectors, video feature vectors, etc. that match text feature vectors, voice feature vectors, or visual feature vectors can be obtained from the Internet, and the web page corresponding to the web page feature vector that matches the text feature vector, the picture corresponding to the picture feature vector, the video corresponding to the video feature vector, etc. that matches the text feature vector are used as search results to be recommended.

[0072] In some embodiments, the search results to be recommended and associated with the information to be searched are stored offline in a database.

[0073] 102. Based on multiple preset recall strategies, perform multi-way recall processing on the recommended search results to obtain recalled recommended search results corresponding to each preset recall strategy.

[0074] The preset recall strategy is used to filter out search results related to the information to be searched from the retrieved search results to be recommended. For example, the preset recall strategies include at least a reverse recall strategy, a semantic recall strategy, a hotspot recall strategy, a category and tag-based recall strategy, and a collaborative recall strategy.

[0075] The recalled recommended search results refer to the search results to be recommended that are recalled from the search results to be recommended using a preset recall strategy and that meet the information to be searched.

[0076] For example, when using the inverted index recall strategy, the search information is first segmented to obtain at least one word (term), and the recall recommendation search results that match the term are recalled through the inverted index.

[0077] When using the semantic recall strategy, a language model is used to vectorize each search result to be recommended, and to encode the information to be searched. The search results to be recommended that have similar semantics to the information to be searched are recalled through the vector retrieval service to obtain the recalled recommended search results corresponding to the semantic recall strategy.

[0078] When using the hot recall strategy, a batch of highly popular recall recommendation search results are mined, and the recall recommendation search results can be sorted according to the number of likes, comments, and shares.

[0079] When using a recall strategy based on categories and tags, obtain the category to which the information to be searched belongs, and obtain information such as the user's interest categories and interest tags related to the information to be searched, and use this information as a trigger to trigger the recall of recall recommendation search results that match the category, interest category, interest tag, etc.

[0080] When using the collaborative recall strategy, the user interests related to the information to be searched are obtained, and the similarity between users is calculated based on the user interests. Based on the user similarity calculation results, a group of users with similar interests to the user are selected as the reference user set. On the reference user set, the similarity between the search results to be recommended and other search results is calculated, and a group of search results with a high similarity to the search results to be recommended are selected as the recalled recommended search results corresponding to the collaborative recall strategy.

[0081] In some embodiments, in order to facilitate the recall of the search results to be recommended, based on multiple preset recall strategies, the search results to be recommended are subjected to multi-channel recall processing to obtain the recalled recommended search results corresponding to each preset recall strategy, including:

[0082] Based on the keywords in the information to be searched, determining candidate recommended search results that match the keywords from the search results to be recommended;

[0083] Obtaining attribute features of candidate recommendation search results;

[0084] Based on the attribute characteristics of the candidate recommendation search results, the candidate recommendation search results are sorted to obtain a sorted candidate recommendation search result list;

[0085] Based on multiple preset recall strategies, a multi-way recall process is performed on the candidate recommendation search result list to obtain the recall recommendation search results corresponding to each preset recall strategy.

[0086] Among them, keywords are used to describe the content that users need to query, which can reflect the user's query intention and needs. For example, if the information to be searched is "travel guide", the keywords can be "travel", "guide", "travel guide", etc. If the information to be searched is "hot search", the keywords can be "hot", "search", "hot search", etc.

[0087] The candidate recommended search results refer to the search results to be recommended that match the keywords.

[0088] For example, if the information to be searched is "hot search", the search results to be recommended may be "hot videos", "hot topics", "hot products", "hot searches on websites", "hot searches on social media platforms", "hot searches on e-commerce platforms", "hot searches on video platforms", etc. If the keyword is "hot", the candidate recommended search results may be "hot videos", "hot topics", "hot products", etc. If the keyword is "hot search", the candidate recommended search results may be "hot searches on websites", "hot searches on social media platforms", "hot searches on e-commerce platforms", "hot searches on video platforms", etc.

[0089] The attribute feature may express the inherent attributes of the candidate recommendation search results and be used to sort the candidate recommendation search results. For example, the attribute feature may include at least one of the authority, freshness, interaction value, and quality value of the candidate recommendation search results.

[0090] The authority of the candidate recommended search results can be the authority of the candidate recommended search results evaluated by authoritative websites or institutions, or it can be the authority of the candidate recommended search results evaluated by analyzing user feedback and comments to provide certain clues, or it can be the authority corresponding to media reports and citations of the candidate recommended search results.

[0091] The freshness of the candidate recommended search results may be calculated by the number of days since the timestamp or release date information of the candidate recommended search results was checked, or may be calculated by the number of days since the latest developments of the candidate recommended search results provided by news reports.

[0092] The interaction value of the candidate recommended search result may be an interaction value calculated based on the click-through rate, number of comments, number of shares, number of favorites, number of saves, etc. of the candidate recommended search result.

[0093] The quality value of the candidate recommended search results may refer to the quality degree obtained based on the degree of match between the candidate recommended search results and the information to be searched, or the quality degree obtained based on the evaluation of the authority of the candidate recommended search results, or the quality degree obtained based on the user's feedback and evaluation of the candidate recommended search results, and so on.

[0094] The candidate recommendation search result list is a list obtained by sorting the candidate recommendation search results according to their attribute characteristics.

[0095] For example, the candidate recommendation search result list may reflect the ranking of the candidate recommendation search results in terms of attribute features, so as to recall the top-ranked candidate recommendation search results first based on multiple preset recall strategies as much as possible.

[0096] In some embodiments, in order to facilitate the sorting of the candidate recommendation search results according to the attribute characteristics of the candidate recommendation search results, the candidate recommendation search results are sorted based on the attribute characteristics of the candidate recommendation search results to obtain a sorted candidate recommendation search result list, including:

[0097] Based on the weights corresponding to the attribute features of the candidate recommended search results, weighted processing is performed on the attribute features of the candidate recommended search results to obtain weighted attribute feature values ​​of the candidate recommended search results;

[0098] Add each weighted attribute feature value of the candidate recommendation search results to obtain a comprehensive attribute feature value;

[0099] Based on the comprehensive attribute feature values, the candidate recommendation search results are sorted to obtain a sorted candidate recommendation search result list.

[0100] Among them, the weight corresponding to the attribute feature can represent the influence of the candidate recommendation search result during recall.

[0101] For example, the attribute characteristics of the candidate recommended search results include authority, freshness, interaction value and quality value. The weight corresponding to the authority of the candidate recommended search results can be w1, the weight corresponding to the freshness can be w2, the weight corresponding to the interaction value can be w3, and the weight corresponding to the quality value can be w4, w1+w2+w3+w4=1.

[0102] The weighted attribute feature value refers to the value obtained by weighting the attribute features of the candidate recommendation search results.

[0103] The comprehensive attribute feature value is the value obtained by adding all weighted attribute feature values ​​of the candidate recommendation search results. The comprehensive attribute feature value facilitates the sorting of the candidate recommendation search results.

[0104] Comprehensive attribute characteristic value = w1*authority+w2*newness+w3*interaction value+w4*quality value.

[0105] In some embodiments, in order to facilitate the recall of the search results to be recommended, after obtaining the attribute features of the candidate recommended search results, the following is further included:

[0106] Building an index list based on the attribute features of the candidate recommendation search results and the candidate recommendation search results;

[0107] Based on the attribute characteristics of the candidate recommendation search results, the candidate recommendation search results are sorted to obtain a sorted candidate recommendation search result list, including:

[0108] Based on the attribute characteristics of the candidate recommended search results in the index list, the candidate recommended search results in the index list are sorted to obtain a sorted index list;

[0109] Based on multiple preset recall strategies, a multi-way recall process is performed on the candidate recommendation search result list to obtain the recall recommendation search results corresponding to each preset recall strategy, including:

[0110] Based on multiple preset recall strategies, multiple recall processes are performed on the candidate recommended search results in the sorted index list to obtain the recalled recommended search results corresponding to each preset recall strategy.

[0111] The index list is a data structure constructed based on the attribute characteristics and content of the candidate recommendation search results, and is used to quickly retrieve and find results that meet specific conditions. For example, in addition to the candidate recommendation search results, the index list also includes the attribute characteristics of the candidate recommendation search results.

[0112] The sorted index list is a list obtained by sorting the candidate recommendation search results in the index list according to their attribute features.

[0113] For example, if the attribute characteristics of the candidate recommended search results include at least one of authority, currency, interaction value and quality value, the sorted index list can be the candidate recommended search results in the index list, a list obtained by sorting according to at least one of authority, currency, interaction value and quality value, or a list obtained by sorting according to the comprehensive attribute characteristic values ​​obtained by authority, currency, interaction value and quality value, and so on.

[0114] The candidate recommended search results in the index list may be sorted from large to small according to the comprehensive attribute feature values ​​to obtain a sorted index list so that the top ranked candidate recommended search results may be recalled.

[0115] 103. Calculate the relevance between the recalled recommended search results and the information to be searched.

[0116] Among them, the relevance can reflect the relevance between the recalled recommended search results and the information to be searched.

[0117] For example, the relevance can be calculated by using a text retrieval algorithm (such as BM25) to calculate a score based on the frequency of co-occurring words between the recalled recommendation search results and the information to be searched. The BM25 score can well balance the importance of keywords in the information to be searched and the frequency of keywords in the recalled recommendation search results.

[0118] Alternatively, a relevance model may be trained, and the recalled recommended search results and the information to be searched may be input into the trained relevance model to obtain a score output by the relevance model.

[0119] In some embodiments, considering that the recall recommendation search results corresponding to each preset recall strategy may have the same recall recommendation search results, in order to avoid repeatedly calculating the relevance between the same recall recommendation search results and the information to be searched, calculating the relevance between the recall recommendation search results and the information to be searched includes:

[0120] Merge the recall recommendation search results corresponding to each preset recall strategy to obtain a recall recommendation search result set;

[0121] Deduplication processing is performed on the recall recommendation search result set to obtain a deduplication recall recommendation search result set;

[0122] The relevance between the recalled recommendation search results in the removed duplicate recalled recommendation search result set and the information to be searched is calculated.

[0123] The recall recommendation search result set is a set obtained by merging the recall recommendation search results corresponding to each preset recall strategy.

[0124] The deduplicated recall recommendation search result set is a set obtained by removing duplicate items in the recall recommendation search result set.

[0125] For example, if there are multiple recall recommendation search results A in the recall recommendation search result set, one recall recommendation search result A is retained, and the redundant recall recommendation search results A are deleted to obtain a deduplicated recall recommendation search result set.

[0126] In some embodiments, in order to calculate the relevance between the recalled recommendation search results and the information to be searched, calculating the relevance between the recalled recommendation search results in the deduplicated recalled recommendation search results set and the information to be searched includes:

[0127] Based on the information to be searched, determine each quality evaluation index of the recall recommendation search results in the recall recommendation search result set after deduplication;

[0128] The various quality evaluation indicators of the recalled recommendation search results are weighted and fused to obtain the relevance between the recalled recommendation search results and the information to be searched.

[0129] Among them, the quality evaluation index is used to measure whether the recall recommendation search results meet the standards of the information to be searched, that is, to evaluate whether the recall recommendation search results meet the standards of the user's search intention. For example, the quality evaluation index may include relevance index, consumption index, authority index and timeliness index, etc. The relevance index for determining the recall recommendation search results includes but is not limited to:

[0130] a) A text retrieval algorithm (such as BM25) may be used to calculate a score based on the frequency of co-occurring words between the recalled recommended search results and the information to be searched.

[0131] b) Calculate the cosine similarity between the information to be searched and each recalled recommended search result to measure the correlation between them.

[0132] The consumption indicators for determining the recall recommendation search results include but are not limited to:

[0133] Ⅰ) Click-through rate: Click-through rate refers to the proportion of users who click on a recalled recommended search result after searching. It is usually used to measure the attractiveness and relevance of recalled recommended search results. Click-through rate = number of clicks / total number of impressions.

[0134] II) Conversion rate: Conversion rate refers to the proportion of users who have completed a predetermined goal (such as purchasing a product or filling out a form). It is usually used to measure the actual effect and commercial value of recalling recommended search results. Conversion rate = number of users who have completed their goals / total number of visiting users.

[0135] III) Average dwell time: Average dwell time refers to the average length of time a user stays on a recall recommendation search result page. It is usually used to measure the attractiveness and user experience of recall recommendation search results. Average dwell time = total dwell time / number of visits.

[0136] The authoritative indicators for determining the recall recommendation search results include but are not limited to:

[0137] A) An authority index can be assigned to each recalled recommendation search result. This index can be determined based on multiple factors, such as the domain authority of the website, content quality, professional background of the author, etc. The authority index = the sum of the comprehensively considered authority factors.

[0138] B) The authority index is measured by calculating the number of backlinks from other websites pointing to the recalled recommended search results. A large number of backlinks may mean that the recalled recommended search results are widely cited and recognized by other websites.

[0139] C) Evaluate the author’s or publisher’s professional background, qualifications, experience, and reputation to determine the authority of the recalled recommended search results.

[0140] Indicators for determining the timeliness of recall recommendation search results include but are not limited to:

[0141] 1) The freshness index is calculated by recalling the release timestamp of the recommended search results. Usually, the time difference is calculated by subtracting the release time from the current time. The freshness index = current time - release time.

[0142] 2) Consider the time point when the search engine crawls and indexes the recall recommendation search results. If a recall recommendation search result is found in the most recent index, it can be considered to be more up-to-date than the recall recommendation search results found previously.

[0143] 3) Evaluate the update frequency of the website or article. If it is updated frequently, it may be more up-to-date. Up-to-dateness index = update frequency / release time. Relevance = w5*relevance index + w6*consumption index + w7*authority index + w8*up-to-dateness index. 104. Determine the relevance level corresponding to the recalled recommended search results based on the relevance and the relevance range specified by the preset relevance level.

[0144] The relevance range specified by the preset relevance level is used to assign a relevance level to each recall recommendation search result when the recall recommendation search results are sorted according to relevance, and this level is related to the relevance.

[0145] The relevance level refers to a preset relevance level corresponding to a relevance within a relevance range specified by the preset relevance level.

[0146] For example, the preset correlation level corresponding to the correlation range greater than 0.6 may be a strong correlation level result, the preset correlation level corresponding to the correlation range less than 0.3 may be a general correlation level result, and the preset correlation level corresponding to the correlation range 0.6-0.3 may be a weak correlation level result.

[0147] The recalled recommended search results corresponding to the strongly relevant gear results are grouped into list A, the recalled recommended search results corresponding to the weakly relevant gear results are grouped into list B, and the recalled recommended search results corresponding to the broadly relevant gear results are grouped into list C. List A and list B can form a search result candidate pool, list C is a recommended result candidate pool, and the search result candidate pool can be prioritized ahead of the recommended result candidate pool.

[0148] 105. Based on the relevance level corresponding to the recall recommendation search results and the relevance level information corresponding to each preset search scenario, determine the target recall recommendation search results corresponding to each preset search scenario from the recall recommendation search results.

[0149] The preset search scenario is a preset search scenario that matches the user's search requirements. For example, the preset search scenario includes a comprehensive search scenario, a vertical search scenario, a vertical search scenario with no results, and the like.

[0150] The relevance level information is used to indicate the relevance level matching the preset search scenario.

[0151] The target recall recommendation search results are the recall recommendation search results that need to be returned in the preset search scenario.

[0152] For example, if the relevance level includes strongly relevant gear results, weakly relevant gear results and generally relevant gear results, the relevance level information can record the strongly relevant gear results corresponding to the comprehensive search scenario, the strongly relevant gear results and weakly relevant gear results corresponding to the vertical search scenario, and the generally relevant gear results corresponding to the vertical no-result search scenario.

[0153] Based on the strongly correlated gear results corresponding to the recall recommendation search results and the strongly correlated gear results corresponding to the comprehensive search scenario, the target recall recommendation search results corresponding to the comprehensive search scenario can be determined from the recall recommendation search results, that is, the comprehensive search scenario corresponds to list A in the aforementioned search result candidate pool.

[0154] Based on the strongly relevant gear results corresponding to the recalled recommendation search results, the weakly relevant gear results corresponding to the recalled recommendation search results, and the strongly relevant gear results and weakly relevant gear results corresponding to the vertical search scenario, the target recalled recommendation search results corresponding to the vertical search scenario can be determined from the recalled recommendation search results, that is, List A and List B in the aforementioned search result candidate pool corresponding to the vertical search scenario.

[0155] Based on the pan-relevant gear results corresponding to the recalled recommendation search results and the pan-relevant gear results corresponding to the vertical no-result search scenario, the target recalled recommendation search results corresponding to the vertical no-result search scenario can be determined from the recalled recommendation search results, that is, list C in the aforementioned recommendation result candidate pool corresponding to the vertical no-result search scenario.

[0156] In some embodiments, in order to return search results matching the business search scenario to the user, obtaining the information to be searched includes:

[0157] Receiving a business search instruction, obtaining information to be searched indicated by the business search instruction, where the business search instruction corresponds to a target search scenario;

[0158] After determining the target recall recommendation search results corresponding to each preset search scenario from the recall recommendation search results based on the relevance level corresponding to the recall recommendation search results and the relevance level information corresponding to each preset search scenario, the method further includes:

[0159] Returns the target recall recommendation search results corresponding to the target search scenario.

[0160] The business search instruction expresses the user's search needs and expectations. The business search instruction is targeted at a specific search scenario, and different search scenarios may correspond to different business search instructions.

[0161] The target search scenario refers to the search scenario corresponding to the business search instruction.

[0162] For example, through business search instructions, after the user enters the information to be searched on the search page, if there is a corresponding recall recommendation search result under the strongly relevant gear result, then you can jump from the search page to the comprehensive search scenario page, which includes target recall recommendation search results of various types (web pages, documents, channels, videos, etc.). If there is no corresponding recall recommendation search result under the strongly relevant gear result and the weakly relevant gear result, then jump from the search page to the vertical no-result search scenario page, which includes target recall recommendation search results of various types under the pan-related gear result.

[0163] When the user selects a channel vertical search scene on the comprehensive search scene page, he can receive a business search instruction corresponding to the channel handling search scene. According to the business search instruction, he can jump from the comprehensive search scene page to the channel vertical search scene page. The channel vertical search scene page only includes target recall recommended search results related to the channel.

[0164] From the above, it can be seen that the embodiments of the present application can obtain information to be searched, and search results to be recommended associated with the information to be searched; based on multiple preset recall strategies, perform multi-way recall processing on the search results to be recommended, and obtain recall recommendation search results corresponding to each preset recall strategy; calculate the correlation between the recall recommendation search results and the information to be searched; determine the correlation level corresponding to the recall recommendation search results based on the correlation and the correlation range specified by the preset correlation level; based on the correlation level corresponding to the recall recommendation search results and the correlation level information corresponding to each preset search scenario, determine the target recall recommendation search results corresponding to each preset search scenario from the recall recommendation search results.

[0165] Therefore, this solution can obtain the search results to be recommended associated with the information to be searched by retrieving it. Then, by using a variety of preset recall strategies, the recall recommendation search results that meet each preset recall strategy can be selected from the search results to be recommended, thereby providing more comprehensive and diverse recall recommendation search results. The relevance between the recall recommendation search results and the information to be searched is calculated, and according to the relevance range specified by the preset relevance level, the relevance level corresponding to the recall recommendation search results can be obtained. Using the relevance level information corresponding to each preset search scenario, the target recall recommendation search results corresponding to each preset search scenario can be obtained. In this way, the target recall recommendation search results corresponding to each preset search scenario can be planned, thereby ensuring that a targeted and comprehensive search experience can be provided in different scenarios, and search results with corresponding relevance levels can be provided for different search scenarios, which can improve the user's search experience.

[0166] The method described in the above embodiment will be further described in detail below.

[0167] In this embodiment, the method of the embodiment of the present application will be described in detail by taking searching in an instant messaging application as an example.

[0168] like Figure 2a As shown, a specific process of an information search method is as follows:

[0169] 201. Obtain information to be searched and search results to be recommended that are associated with the information to be searched.

[0170] 202. Offline, store all the search results to be recommended in a database at one time, build an inverted index, and determine candidate recommended search results that match the keyword from the search results to be recommended.

[0171] For example, according to the keywords in the information to be searched, the matching candidate recommendation search results are retrieved from the index. The search recommendation fusion method under the query relevance constraint is particularly effective in search scenarios where the number of search results to be recommended is small (hundreds of thousands).

[0172] 203. After the index list is created, the candidate recommended search results in the index list are sorted based on the attribute characteristics of the candidate recommended search results, that is, the comprehensive scores of the candidate recommended search results in terms of authority, timeliness, interactivity and quality, to obtain a sorted index list.

[0173] L0 index pre-sorting, such as Figure 2b As shown, including:

[0174] Calculate Score L0 =w1*Score 权威 +w2*Score时新 +w3*Score 互动 +w4*Score 质量 , where Score L0 Score is a comprehensive score of the candidate recommended search results in terms of authority, freshness, interaction value and quality value. 权威 is the authority of the candidate recommended search results, Score 时新 is the freshness of the candidate recommended search results, Score 互动 is the interaction value of the candidate recommendation search result, Score 质量 is the quality value of the candidate recommendation search result.

[0175] Authority calculation: It can be calculated based on the publication source of the candidate recommended search results and the author's authority in the relevant field.

[0176] Recentness calculation: It can be calculated based on the number of days since the candidate recommended search results were published.

[0177] Interaction value calculation: It can be calculated through indicators such as the historical views, likes and reposts of the candidate recommended search results.

[0178] Quality value calculation: It can be calculated based on the quality of the candidate recommendation search results, such as the clarity of the video, the completeness of the document content, etc.

[0179] 204. Adopt multiple preset recall strategies, such as reverse recall strategy, semantic recall strategy, hot recall strategy, category and tag based recall strategy, collaborative recall strategy, perform multi-way recall processing on the candidate recommended search results in the sorted index list, and obtain the recalled recommended search results corresponding to each preset recall strategy.

[0180] For example, Figure 2b As shown in the figure, L1 recall: regards search as a recommendation based on (information to be searched (query), user identifier (user id)), integrates search inverse recall and recommendation multi-way recall, and takes into account user interests and search relevance. In addition to the inverse recall and semantic recall technologies in traditional search, the recall technology in the recommendation system (hot recall strategy (Hot recall), category and tag-based recall strategy (CB recall) and collaborative recall strategy (CF recall)) is also used. In the case of no results / few results in the search, the pan-related recommendation results are supplemented to meet the user's consumption needs as much as possible and improve the user experience.

[0181] a. When a user initiates a search request, a multi-way recall operation is performed below the information to be searched:

[0182] 1. Inverted recall: First, split the query into multiple terms, such as "Honor of Kings" into "king" and "glory", and then recall the candidate recommendation search results that match the terms through the inverted index.

[0183] 2. Semantic recall: Use the language model to vector encode all the candidate recommendation search results in the index library (offline one-time) and store them in the candidate recommendation search result ranking service; use the language model to vector encode the search information in real time; recall the candidate recommendation search results that are semantically similar to the search information through the vector retrieval service, where the language model uses the pre-trained language model commonly used in the industry (roberta model).

[0184] 3. Hot recall: Offline mining of a batch of highly popular candidate recommendation search results (sorted by likes / comments / shares), updated daily, as the recall results.

[0185] 4. CB recall: Get the category to which the information to be searched belongs (e.g., Query = Honor of Kings, the category is games), use the Query category as a trigger, and recall the candidate recommended search results that match the Query category; similar triggers also include: user's interest categories and user's interest tags.

[0186] 205. Perform a merging and deduplication processing on the recall recommendation search results corresponding to each preset recall strategy to obtain a deduplication-free recall recommendation search result set.

[0187] 206. Calculate the relevance index, consumption index, authority index, and timeliness index between the recall recommendation search results and the information to be searched in the recall recommendation search results set;

[0188] 207. The relevance index, consumption index, authority index and timeliness index are weighted and integrated to obtain the relevance.

[0189] Calculate Score L2 =w1*Score 相关性 +w2*Score 消费性 +w3*Score 权威性 +w4*Score 时新性 , where Score L2 is the relevance, Score 相关性 It is the correlation index between the recalled recommended search results and the information to be searched. 消费性 It is the consumption index of recalling recommended search results. Score 权威性 It is an authoritative indicator for recalling recommended search results. Score 时新性 It is an indicator of the freshness of the recalled recommended search results.

[0190] For example, relevance calculation can use the BM25 algorithm or train a relevance model to predict the score. Use the xgboost algorithm to train the relevance model, as follows:

[0191] 1) Sample labeling: Randomly sample business data. For each query, randomly sample 6-8 samples for labeling. The total number of samples is about 100,000 (w) query-doc pairs. Relevant is 1, and irrelevant is 0.

[0192] 2) Feature calculation: query-side features (query length / number of words / category), query-doc cross features (BM25 / matching rate / category matching);

[0193] 3) Correlation model: The extreme gradient boosting decision tree (xgboost) training architecture is used to train the correlation discrimination model. The tree model has strong interpretability and good fitting effect on dense class features. GridSearchCV is used for parameter optimization and cross-validation is used for model training.

[0194] 4) Online prediction: When a user initiates a search request, the query and doc text are passed into the xgboost model as input, and the model performs prediction and scoring in the range of [0,1].

[0195] 208. Determine the relevance level corresponding to the recalled recommended search results according to the relevance and the relevance range specified by the preset relevance level.

[0196] 209. Based on the relevance level corresponding to the recall recommendation search results and the relevance level information corresponding to each preset search scenario, determine the target recall recommendation search results corresponding to each preset search scenario from the recall recommendation search results.

[0197] For example, Figure 2c As shown in the figure, the core idea of ​​relevance level is to construct two candidate pools according to the relevance score: the search result candidate pool and the recommendation result candidate pool. First, the recall recommendation search results with a relevance score greater than 0.6 to the information to be searched are considered to be strongly relevant results and are recorded as List A; the recall recommendation search results with a relevance score less than 0.3 are pan-relevant results, recorded as List C; the rest are weakly relevant results, recorded as List B. List A and List B constitute the search result candidate pool, List C is the recommendation result candidate pool, and the candidate recommendation search results in the search result candidate pool are prioritized in front of the candidate recommendation search results in the recommendation result candidate pool.

[0198] Business strategy: This application can flexibly meet the needs of different business scenarios. For example, only strongly relevant results (list A) are displayed in the comprehensive search scenario, strong / weakly relevant results (list A and list B) are displayed in the channel vertical search scenario, and other pan-related results (list C) are displayed for the channel vertical search scenario with no results.

[0199] Compared with the traditional inverted search method, this application has been implemented in the vertical search scenario and achieved significant benefits, among which, channel_search plus channel conversion rate: +5.1188%, channel_search query click rate: +2.3379% (significant), channel_search average channel exposure per person: +10.8793% (significant).

[0200] From the above, we can see that when there are no results or few results for a search, we can supplement the search with general related recommended results to meet the user's consumption needs as much as possible and improve the user experience.

[0201] In order to better implement the above method, the embodiment of the present application also provides an information search device, which can be integrated in an electronic device, and the electronic device can be a terminal, a server, etc. Among them, the terminal can be a mobile phone, a tablet computer, a smart Bluetooth device, a laptop, a personal computer, etc.; the server can be a single server or a server cluster composed of multiple servers.

[0202] For example, in this embodiment, the method of the embodiment of the present application is described in detail by taking the information search device being specifically integrated in the server as an example.

[0203] For example, Figure 3 As shown, the information search device may include an information acquisition unit 301, a recall unit 302, a calculation unit 303, a level determination unit 304, and a scene recommendation unit 305, as follows:

[0204] (i) Information acquisition unit 301.

[0205] The information acquisition unit 301 is used to acquire information to be searched and search results to be recommended that are associated with the information to be searched.

[0206] (ii) Recall unit 302.

[0207] The recall unit 302 is used to perform multi-way recall processing on the recommended search results based on multiple preset recall strategies to obtain the recalled recommended search results corresponding to each preset recall strategy.

[0208] In some embodiments, the recall unit includes a candidate search unit, a feature acquisition unit, a search ranking unit, and a recall sub-unit;

[0209] A candidate search unit, configured to determine, based on the keywords in the information to be searched, candidate recommended search results that match the keywords from the search results to be recommended;

[0210] A feature acquisition unit, used to acquire attribute features of candidate recommendation search results;

[0211] A search ranking unit, used to sort the candidate recommendation search results based on the attribute characteristics of the candidate recommendation search results to obtain a sorted candidate recommendation search result list;

[0212] The recall subunit is used to perform multi-way recall processing based on multiple preset recall strategies to obtain recall recommendation search results corresponding to each preset recall strategy.

[0213] In some embodiments, an index building unit is also included;

[0214] An index building unit, used to build an index list based on the attribute features of the candidate recommendation search results and the candidate recommendation search results;

[0215] The search ranking unit includes an index ranking unit;

[0216] An index sorting unit, used to sort the candidate recommended search results in the index list based on the attribute features of the candidate recommended search results in the index list to obtain a sorted index list;

[0217] The recall subunit includes an index recall unit;

[0218] The index recall unit is used to perform multi-way recall processing on the candidate recommended search results in the sorted index list based on multiple preset recall strategies to obtain the recalled recommended search results corresponding to each preset recall strategy.

[0219] (iii) Calculation unit 303.

[0220] The calculation unit 303 is used to calculate the relevance between the recalled recommendation search results and the information to be searched.

[0221] In some embodiments, the computing unit includes a merging subunit, a deduplication subunit, and a computing subunit;

[0222] A merging subunit is used to merge the recall recommendation search results corresponding to each preset recall strategy to obtain a recall recommendation search result set;

[0223] A deduplication subunit is used to perform deduplication processing on the recall recommendation search result set to obtain a deduplicated recall recommendation search result set;

[0224] The calculation subunit is used to calculate the relevance between the recalled recommendation search results in the recalled recommendation search result set after deduplication and the information to be searched.

[0225] In some embodiments, the calculation subunit includes an indicator determination unit and a weighted fusion unit;

[0226] An indicator determination unit, used to determine each quality evaluation indicator of the recall recommendation search results in the recall recommendation search result set after deduplication based on the information to be searched;

[0227] The weighted fusion unit is used to perform weighted fusion processing on various quality evaluation indicators of the recalled recommendation search results to obtain the relevance between the recalled recommendation search results and the information to be searched.

[0228] (iv) Level determination unit 304.

[0229] The level determination unit 304 is used to determine the relevance level corresponding to the recalled recommendation search result according to the relevance and the relevance range specified by the preset relevance level.

[0230] (V) Scene recommendation unit 305.

[0231] The scenario recommendation unit 305 is used to determine the target recall recommendation search results corresponding to each preset search scenario from the recall recommendation search results based on the relevance level corresponding to the recall recommendation search results and the relevance level information corresponding to each preset search scenario.

[0232] In some embodiments, the information acquisition unit includes an instruction receiving unit;

[0233] An instruction receiving unit, used to receive a service search instruction, where the service search instruction corresponds to a target search scenario;

[0234] After the scene recommendation unit, a search return unit is also included;

[0235] The search return unit is used to return the target recall recommendation search results corresponding to the target search scenario.

[0236] In specific implementation, the above units can be implemented as independent entities, or can be arbitrarily combined to be implemented as the same or several entities. The specific implementation of the above units can refer to the previous method embodiments, which will not be repeated here.

[0237] As can be seen from the above, the information search device of this embodiment obtains the information to be searched and the search results to be recommended associated with the information to be searched by the information acquisition unit; the recall unit performs multi-way recall processing on the recommended search results based on multiple preset recall strategies to obtain the recall recommendation search results corresponding to each preset recall strategy; the calculation unit calculates the relevance between the recall recommendation search results and the information to be searched; the level determination unit determines the relevance level corresponding to the recall recommendation search results according to the relevance and the relevance range specified by the preset relevance level; the scene recommendation unit determines the target recall recommendation search results corresponding to each preset search scenario from the recall recommendation search results based on the relevance level corresponding to the recall recommendation search results and the relevance level information corresponding to each preset search scenario.

[0238] Therefore, the embodiments of the present application can improve the breadth and coverage of the search and meet the diverse needs of users.

[0239] The embodiment of the present application also provides an electronic device, which can be a terminal, a server, etc. The terminal can be a mobile phone, a tablet computer, a smart Bluetooth device, a laptop, a personal computer, etc. The server can be a single server or a server cluster composed of multiple servers, etc.

[0240] In some embodiments, the information search device may also be integrated into multiple electronic devices. For example, the information search device may be integrated into multiple servers, and the information search method of the present application may be implemented by multiple servers.

[0241] In this embodiment, the electronic device of this embodiment is a server as an example for detailed description, for example, Figure 4 As shown, it shows a schematic diagram of the structure of the server involved in the embodiment of the present application, specifically:

[0242] The server may include one or more processing core processors 401, one or more computer-readable storage media memories 402, a power supply 403, an input module 404, and a communication module 405. Those skilled in the art will appreciate that Figure 4 The server structure shown in the figure does not constitute a limitation on the server, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. Among them:

[0243] The processor 401 is the control center of the server, and uses various interfaces and lines to connect various parts of the entire server. It executes various functions of the server and processes data by running or executing software programs and / or modules stored in the memory 402, and calling data stored in the memory 402. In some embodiments, the processor 401 may include one or more processing cores; in some embodiments, the processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 401.

[0244] The memory 402 can be used to store software programs and modules. The processor 401 executes various functional applications and data processing by running the software programs and modules stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data created according to the use of the server, etc. In addition, the memory 402 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 with access to the memory 402.

[0245] The server also includes a power supply 403 for supplying power to various components. In some embodiments, the power supply 403 may be logically connected to the processor 401 through a power management system, so that the power management system can manage charging, discharging, power consumption, and other functions. The power supply 403 may also include any components such as one or more DC or AC power supplies, recharging systems, power failure detection circuits, power converters or inverters, and power status indicators.

[0246] The server may further include an input module 404, which may be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal input related to user settings and function control.

[0247] The server may further include a communication module 405. In some embodiments, the communication module 405 may include a wireless module. The server may perform short-range wireless transmission through the wireless module of the communication module 405, thereby providing wireless broadband Internet access for the user. For example, the communication module 405 may be used to help the user send and receive emails, browse web pages, and access streaming media.

[0248] Although not shown, the server may also include a display unit, etc., which will not be described in detail herein. Specifically in this embodiment, the processor 401 in the server will load the executable files corresponding to the processes of one or more application programs into the memory 402 according to the following instructions, and the processor 401 will run the application programs stored in the memory 402, thereby implementing the steps in the methods of the embodiments of the present application.

[0249] The specific implementation of the above operations can be found in the previous embodiments, which will not be described in detail here.

[0250] From the above, it can be seen that target recall recommended search results corresponding to each preset search scenario can be planned, thereby ensuring that a targeted and comprehensive search experience can be provided in different situations, and search results with corresponding relevance levels can be provided for different search scenarios, which can improve the user's search experience.

[0251] A person of ordinary skill in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be completed by instructions, or by controlling related hardware through instructions. The instructions may be stored in a computer-readable storage medium and loaded and executed by a processor.

[0252] To this end, an embodiment of the present application provides a computer-readable storage medium, in which a plurality of instructions are stored, and the instructions can be loaded by a processor to execute the steps in any one of the information search methods provided in the embodiments of the present application. For example, the instructions can execute the following steps:

[0253] Obtaining information to be searched and search results to be recommended associated with the information to be searched;

[0254] Based on multiple preset recall strategies, multi-way recall processing is performed on the recommended search results to obtain the recalled recommended search results corresponding to each preset recall strategy;

[0255] Calculate the relevance between the recalled recommended search results and the information to be searched;

[0256] Determining the relevance level corresponding to the recalled recommended search results according to the relevance and the relevance range specified by the preset relevance level;

[0257] Based on the relevance level corresponding to the recall recommendation search results and the relevance level information corresponding to each preset search scenario, a target recall recommendation search result corresponding to each preset search scenario is determined from the recall recommendation search results.

[0258] The storage medium may include: a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0259] According to one aspect of the present application, a computer program product or computer program is provided, the computer program product or computer program including a computer program / instruction, the computer program / instruction being stored in a computer-readable storage medium. A processor of an electronic device reads the computer program / instruction from the computer-readable storage medium, and the processor executes the computer program / instruction, so that the electronic device executes the method provided in various optional implementations of the information search provided in the above embodiments.

[0260] Since the instructions stored in the storage medium can execute the steps in any information search method provided in the embodiments of the present application, the beneficial effects that can be achieved by any information search method provided in the embodiments of the present application can be achieved. Please refer to the previous embodiments for details and will not be repeated here.

[0261] The above is a detailed introduction to an information search method, device, electronic device, storage medium and program product provided in the embodiments of the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for technical personnel in this field, according to the ideas of the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. An information search method, characterized in that: include: Acquire information to be searched and search results to be recommended associated with the information to be searched; Based on multiple preset recall strategies, multi-way recall processing is performed on the search results to be recommended to obtain recall recommendation search results corresponding to each preset recall strategy; Calculate the relevance between the recalled recommended search results and the information to be searched ； Determining the relevance level corresponding to the recalled recommendation search result according to the relevance and the relevance range specified by the preset relevance level; Based on the relevance level corresponding to the recall recommendation search results and the relevance level information corresponding to each preset search scenario, a target recall recommendation search result corresponding to each preset search scenario is determined from the recall recommendation search results.

2. The information search method according to claim 1, characterized in that: The obtaining of the information to be searched includes: Receiving a service search instruction, and acquiring information to be searched indicated by the service search instruction, wherein the service search instruction corresponds to a target search scenario; After determining the target recall recommendation search results corresponding to each preset search scenario from the recall recommendation search results based on the relevance level corresponding to the recall recommendation search results and the relevance level information corresponding to each preset search scenario, the method further includes: Return the target recall recommendation search results corresponding to the target search scenario.

3. The information search method according to claim 1, characterized in that: The method of performing multi-channel recall processing on the search results to be recommended based on multiple preset recall strategies to obtain the recalled recommended search results corresponding to each preset recall strategy includes: Based on the keywords in the information to be searched, determining candidate recommended search results that match the keywords from the search results to be recommended; Obtaining attribute features of candidate recommendation search results; Based on the attribute characteristics of the candidate recommended search results, the candidate recommended search results are sorted to obtain a sorted candidate recommended search result list; Based on a plurality of preset recall strategies, a multi-way recall process is performed on the candidate recommendation search result list to obtain the recall recommendation search results corresponding to each preset recall strategy.

4. The information search method according to claim 3, characterized in that: After obtaining the attribute features of the candidate recommendation search results, the method further includes: Building an index list based on the attribute features of the candidate recommended search results and the candidate recommended search results; The step of sorting the candidate recommendation search results based on the attribute features of the candidate recommendation search results to obtain a sorted candidate recommendation search result list includes: Based on the attribute characteristics of the candidate recommended search results in the index list, the candidate recommended search results in the index list are sorted to obtain a sorted index list; The method of performing multi-way recall processing on the candidate recommendation search result list based on multiple preset recall strategies to obtain the recall recommendation search results corresponding to each preset recall strategy includes: Based on multiple preset recall strategies, a multi-way recall process is performed on the candidate recommended search results in the sorted index list to obtain the recalled recommended search results corresponding to each preset recall strategy.

5. The information search method according to claim 1, characterized in that: The calculating the relevance between the recalled recommendation search results and the information to be searched includes: Merging the recall recommendation search results corresponding to the preset recall strategies to obtain a recall recommendation search result set; Deduplication processing is performed on the recall recommendation search result set to obtain a deduplication recall recommendation search result set; The relevance between the recalled recommended search results in the removed duplicate recalled recommended search result set and the information to be searched is calculated.

6. The information search method according to claim 5, characterized in that: The calculating the relevance between the recalled recommendation search results in the removed duplicate recalled recommendation search results set and the information to be searched includes: Based on the information to be searched, determining each quality evaluation index of the recall recommendation search results in the deduplicated recall recommendation search result set; A weighted fusion process is performed on each quality evaluation index of the recalled recommendation search result to obtain the relevance between the recalled recommendation search result and the information to be searched.

7. An information search device, characterized in that: include: An information acquisition unit, used to acquire information to be searched and search results to be recommended associated with the information to be searched; A recall unit, configured to perform multi-way recall processing on the search results to be recommended based on multiple preset recall strategies, and obtain recall recommendation search results corresponding to each preset recall strategy; A calculation unit, used to calculate the relevance between the recalled recommended search results and the information to be searched ； A level determination unit, configured to determine the relevance level corresponding to the recalled recommendation search result according to the relevance and a relevance range specified by a preset relevance level; The scenario recommendation unit is used to determine the target recall recommendation search results corresponding to each preset search scenario from the recall recommendation search results based on the relevance level corresponding to the recall recommendation search results and the relevance level information corresponding to each preset search scenario.

8. An electronic device, characterized in that: The invention comprises a processor and a memory, wherein the memory stores a plurality of instructions; the processor loads instructions from the memory to execute the steps in the information search method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor to execute the steps in the information search method according to any one of claims 1 to 6.

10. A computer program product comprising a plurality of instructions, characterized in that: When the instructions are executed by a processor, the steps of the information search method according to any one of claims 1 to 6 are implemented.

Images