An information recommendation method, device, apparatus, and storage medium

Abstract

The application provides an information recommendation method and related device. The method comprises: receiving an information query request sent by a terminal, the information query request carrying query information; obtaining a candidate information set associated with the query information according to the information query request, wherein the candidate information set is derived from at least two of a search recommendation log, a search click log and a third-party click log; obtaining a first hot word set and a second hot word set from the candidate information set, wherein the first hot word set is determined according to the heat values of each candidate information in a first time period, and the second hot word set is determined according to the heat values of each candidate information in a second time period; and generating an information recommendation index according to the first hot word set and the second hot word set and feeding back the information recommendation index. The application improves the coverage of hot word data, realizes high stability and high real-time performance of the hot word, and improves the conversion rate of the information recommendation index.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to an information recommendation method, apparatus, device, and storage medium. Background Technology

[0002] With the development of computer technology, people are increasingly relying on the Internet to search for information. How to quickly and accurately find the information they need from the vast amount of information on the Internet has become an increasingly important issue for users.

[0003] Traditional recommendation functions rely on search engines to retrieve information related to the user's query terms as a candidate set. The popularity of these candidate terms is then used as a key feature for ranking all candidates, generating a recommended list which is returned to the user. However, existing technologies have several drawbacks: the candidate set relies on a single data source, failing to meet users' search needs; and the trending keyword data in recommendations is generated based on long-term click data, resulting in poor real-time performance and difficulty in capturing short-term increases in popularity, thus failing to meet users' demands for real-time search capabilities. Summary of the Invention

[0004] This application provides an information recommendation method and related apparatus, which expands the data sources of candidate information sets, improves the coverage of hot word data, and achieves the characteristics of high stability and strong real-time performance of hot words, thus better meeting user needs.

[0005] One aspect of this application provides an information recommendation method, comprising:

[0006] The receiving terminal sends an information query request, wherein the information query request carries query information;

[0007] Based on the information query request, obtain a candidate information set associated with the query information. The candidate information set is derived from at least two of the following: retrieval recommendation logs, retrieval click logs, and third-party click logs. The candidate information set includes K candidate information items, each of which corresponds to a popularity score, where K is an integer greater than 1.

[0008] A first set of hot words and a second set of hot words are obtained from the candidate information set. The first set of hot words is determined based on the popularity score of each candidate information in the first time period, and the second set of hot words is determined based on the popularity score of each candidate information in the second time period. The first time period includes the second time period, and the duration of the first time period is longer than the duration of the second time period.

[0009] An information recommendation index is generated based on the first and second sets of hot words;

[0010] Feedback information to the terminal recommending indexes.

[0011] Another aspect of this application provides an information recommendation device, comprising:

[0012] The information query request receiving module is used to receive information query requests sent by the terminal, wherein the information query request carries query information;

[0013] The candidate information set acquisition module is used to acquire the candidate information set associated with the query information according to the information query request. The candidate information set is derived from at least two of the following: retrieval recommendation logs, retrieval click logs, and third-party click logs. The candidate information set includes K candidate information, each of which corresponds to a popularity score, where K is an integer greater than 1.

[0014] The hot word set acquisition module is used to acquire a first hot word set and a second hot word set from the candidate information set. The first hot word set is determined based on the popularity score of each candidate information in a first time period, and the second hot word set is determined based on the popularity score of each candidate information in a second time period. The first time period includes the second time period, and the duration of the first time period is longer than the duration of the second time period.

[0015] The information recommendation index generation module is used to generate an information recommendation index based on the first set of hot words and the second set of hot words.

[0016] The information recommendation index sending module is used to send the information recommendation index back to the terminal.

[0017] In another implementation of this application embodiment, the hot word set acquisition module includes: a first hot word set acquisition module and a second hot word set acquisition module;

[0018] The first hot word set acquisition module is used to acquire candidate information with a popularity score greater than a first threshold from the candidate information set within a first time period to obtain a first candidate hot word set; and to update the first candidate hot word set according to a first sub-time period to obtain a first hot word set.

[0019] The second hot word set acquisition module is used to acquire candidate information with a popularity score greater than a second threshold from the candidate information set within a second time period to obtain a second candidate hot word set; and to update the second candidate hot word set according to the second sub-time period to obtain a second hot word set.

[0020] The duration of the first time period is longer than the duration of the first sub-time period, and the duration of the second time period is longer than the duration of the second sub-time period.

[0021] In another implementation of this application embodiment, the information recommendation device further includes: a popularity score calculation module;

[0022] The popularity score calculation module is used to calculate the popularity score of each candidate information by weighting the effective clicks in the retrieval recommendation log, the effective clicks in the retrieval click log, and the effective clicks in the third-party click log.

[0023] In another implementation of this application embodiment, the information recommendation device further includes: an information recall module and a first module for updating popularity scores;

[0024] The information recall module is used to receive information click requests sent by the terminal based on the information recommendation index;

[0025] The first module for updating popularity scores is used to update the popularity scores of candidate information in the candidate information set based on information click requests.

[0026] In another implementation of this application, the information click request carries a target object identifier and target query information, where the target query information is any candidate information in the information recommendation index;

[0027] The first module for updating the popularity score also includes: a duplicate filtering module and a second module for updating the popularity score;

[0028] The deduplication module is used to deduplicat log records in the associated logs based on the target object identifier and target query information to obtain the effective click count of the target query information in the associated logs. The associated logs include: retrieval recommendation logs, retrieval click logs, and third-party click logs.

[0029] The second module for updating the popularity score is used to update the popularity score of the target query information based on the number of valid clicks in the associated logs.

[0030] In another implementation of this application embodiment, the information query request carries the location information of the terminal;

[0031] The candidate information set acquisition module is also used to acquire a candidate information set that is associated with both location information and query information, based on the information query request.

[0032] In another implementation of this application, the information recommendation index generation module is further configured to sort all candidate information in the first hot word set and the second hot word set in descending order based on the popularity scores of all candidate information in the first hot word set and the second hot word set, and generate an information recommendation index.

[0033] Another aspect of this application provides a computer device, comprising:

[0034] Memory, transceiver, processor, and bus system;

[0035] The memory is used to store programs;

[0036] The processor is used to execute programs in memory, including methods for performing the aspects mentioned above;

[0037] Bus systems are used to connect memory and processor to enable communication between them.

[0038] Another aspect of this application provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the methods described above.

[0039] Another aspect of this application provides a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the above aspects.

[0040] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:

[0041] This application provides an information recommendation method, mainly comprising: first, the server obtains a set of related candidate information based on the query information in the information query request sent by the terminal; the candidate information set originates from retrieval recommendation logs, retrieval click logs, and third-party click logs; then, the server divides the candidate information set into a first hot word set and a second hot word set based on the time span and the popularity score of the candidate information; finally, the server generates an information recommendation index based on the first hot word set and the second hot word set, and feeds the information recommendation index back to the terminal. The embodiments provided in this application improve the coverage of hot word data by expanding the data sources of the candidate information set; and by distinguishing hot word sets with different time spans, it simultaneously achieves the characteristics of high stability and strong real-time performance of hot words, better meeting user needs and improving the conversion rate of the information recommendation index. Attached Figure Description

[0042] Figure 1 A schematic diagram of the architecture of an information recommendation system provided in a certain embodiment of this application;

[0043] Figure 2 A flowchart illustrating an information recommendation method provided in one embodiment of this application;

[0044] Figure 3 A flowchart of an information recommendation method provided in another embodiment of this application;

[0045] Figure 4 A flowchart of an information recommendation method provided in another embodiment of this application;

[0046] Figure 5 A flowchart of an information recommendation method provided in another embodiment of this application;

[0047] Figure 6 A flowchart of an information recommendation method provided in another embodiment of this application;

[0048] Figure 7 A flowchart of an information recommendation method provided in yet another embodiment of this application;

[0049] Figure 8 A schematic diagram illustrating an information recommendation method in an application scenario provided by a certain embodiment of this application;

[0050] Figure 9 A schematic diagram illustrating multi-source data fusion processing in an application scenario provided by a certain embodiment of this application;

[0051] Figure 10 A test result diagram of an information recommendation method provided in a certain embodiment of this application;

[0052] Figure 11 A schematic diagram of an information recommendation device provided in one embodiment of this application;

[0053] Figure 12 A schematic diagram of an information recommendation device provided in another embodiment of this application;

[0054] Figure 13 A schematic diagram of an information recommendation device provided in another embodiment of this application;

[0055] Figure 14 A schematic diagram of an information recommendation device provided in another embodiment of this application;

[0056] Figure 15 A schematic diagram of an information recommendation device provided in yet another embodiment of this application;

[0057] Figure 16 This is a schematic diagram of a server structure provided in one embodiment of this application. Detailed Implementation

[0058] This application provides an information recommendation method and related apparatus, which expands the data sources of candidate information sets, improves the coverage of hot word data, and achieves the characteristics of high stability and strong real-time performance of hot words, thus better meeting user needs.

[0059] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “corresponding to,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0060] It should be understood that current information recommendation functions on the market involve users submitting information query requests through a terminal and entering query information (query) at the query entry point. The server first obtains the query information (query) and the terminal's location information (adcode) sent by the terminal. Then, it retrieves information related to both the query information and the location information from the suggestion (SUG) log, generating a candidate information set. Next, it sorts all candidate information in descending order based on the click volume of all candidate information in the candidate information set, generating an information recommendation index. Finally, the information recommendation index is fed back to the terminal. However, the problems with existing technology are: the data source of the candidate information set is singular, which cannot meet users' search needs; the candidate information in the information recommendation index is generated based on click volume over a long period of time, resulting in poor real-time performance and difficulty in capturing information with short-term popularity increases, thus failing to meet users' requirements for the real-time performance of the search system.

[0061] The problem this application solves is that by expanding the data sources of the candidate information set, the coverage of hot word data is improved; and by distinguishing hot word sets with different time spans, the stability and real-time performance of hot words are simultaneously achieved, better meeting user needs and improving the conversion rate of the information recommendation index. The information recommendation method provided in this application is mainly deployed on a server. The server first obtains the information query request sent by the client on the terminal, then generates an information recommendation index based on the information query request, and finally sends the information recommendation index to the terminal.

[0062] For easier understanding, please refer to Figure 1 , Figure 1 This is a diagram illustrating the application environment of the information recommendation method in the embodiments of this application, such as... Figure 1As shown, the information recommendation method in this embodiment is applied to an information recommendation system. The information recommendation system includes a server and a terminal; wherein the server can be an independent physical server, a server cluster composed of multiple physical servers, or a distributed system, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. The terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, etc., but is not limited to these. The terminal and server can be directly or indirectly connected via wired or wireless communication, which is not limited in this embodiment.

[0063] Users submit information query requests through their terminals and enter query information at the query entry point. The server first receives the query request from the terminal, which carries the query information. Next, based on the query request, the server retrieves a set of candidate information associated with the query information. This set of candidate information originates from at least two of the following: search recommendation logs, search click logs, and third-party click logs. The set includes K candidate information items, each corresponding to a popularity score, where K is an integer greater than 1. The server then retrieves a first set of hot words and a second set of hot words from the candidate information set. The first set of hot words is determined based on the popularity scores of each candidate information item within a first time period, and the second set of hot words is determined based on the popularity scores of each candidate information item within a second time period. The first time period includes the second time period, and the duration of the first time period is greater than the duration of the second time period. Finally, the server generates an information recommendation index based on the first and second hot word sets and sends the index back to the terminal.

[0064] The information recommendation method in this application will be described from the server's perspective below. Please refer to [link / reference]. Figure 2 The information recommendation method provided in this application embodiment includes steps S110 to S150. Specifically:

[0065] S110, Receive an information query request sent by the terminal. The information query request carries query information.

[0066] It should be noted that the terminal can run a client or mini-program with search functionality. This client or mini-program is configured with a query entry point for inputting query information. When running, the client or mini-program can connect to the corresponding server and interact with it. The client can be a map client or a web client, etc.; the mini-program can be a map mini-program or a web mini-program, etc. The client or mini-program can detect information query requests applied to the query entry point. Upon detection, it retrieves the query information carried in the query request and sends the query request and its carried query information to the server via the terminal. The information query request can be a triggered action generated when the user clicks the query entry point. The query information can be the query terms entered by the user in the query entry point.

[0067] Understandably, a user clicks on the query entry point on the terminal and enters their query information. The client or app on the terminal first responds to the query request and retrieves the query information. Then, the client or app sends the query request and the query information carried in the query request to the server via the terminal. The server receives the query request and the query information carried in the query request.

[0068] Optionally, after receiving an information query request from a terminal, the server preprocesses the query information. Preprocessing includes regularization of the query information, such as converting full-width characters to half-width characters, traditional Chinese characters to simplified Chinese characters, removing punctuation, case conversion, and encoding conversion.

[0069] S130. Based on the information query request, obtain a candidate information set associated with the queried information. The candidate information set is derived from at least two of the following: retrieval recommendation logs, retrieval click logs, and third-party click logs. The candidate information set includes K candidate information items, each corresponding to a popularity score, where K is an integer greater than 1.

[0070] It should be noted that the candidate information set refers to a collection of K candidate information items. Candidate information can be Points of Interest (POIs). A POI is an object that can be located on a map; a POI includes at least a name, category, and geographic location. The candidate information set associated with the query information refers to a collection of K candidate information items that are associated with the query information in name, category, or geographic location. Association can be understood as the query information having the same name as the POI, the same prefix, the same keywords, the same category, or the same geographic location. The Suggestion (SUG) log refers to the log generated by clicking on target query information in the recommended index generated based on the input query information. The Search (Search) Click Log refers to the log generated by performing a search operation directly after inputting the query information. The Third-Party Click Log refers to the log generated by clicking on links accessed through this client within a third-party client or mini-program. The popularity score of a candidate message can be used to reflect the number of clicks on that candidate message; that is, the more clicks, the higher the popularity score.

[0071] Understandably, based on the information query request, the server retrieves K1 pieces of information related to the information query request from the retrieval recommendation log as a candidate information set, K2 pieces of information related to the information query request from the retrieval click log as a candidate information set, and K3 pieces of information related to the information query request from the third-party click log as a candidate information set, where K1, K2, and K3 are all integers greater than 1. The server merges at least two of the candidate information sets from the retrieval recommendation log, the retrieval click log, and the third-party click log to generate a candidate information set associated with the query information that contains K pieces of candidate information, satisfying K = K1 + K2, K = K1 + K3, K = K2 + K3, or K = K1 + K2 + K3. Each piece of candidate information corresponds to a popularity score, which represents the popularity and / or click volume of the candidate information.

[0072] For example, assuming the client used to make the input information query request is a map client, if a user enters "Peking University" in the query entry and clicks "Peking University North Gate" in the information recommendation index, then "Peking University North Gate" will be recorded in the search recommendation log; if a user enters "Peking University" in the query entry and directly searches, and then clicks "Peking University" in the search results, then "Peking University" will be recorded in the search click log; if a user clicks "Peking University West Gate (Subway Station)" through a link in a third-party application's map client, then "Peking University North Gate Subway Station" will be recorded in the third-party click log. If a user enters "Peking University" in the query entry, location points of interest may include: Peking University, Peking University West Gate (Subway Station), etc.; if a user enters "universities" in the query entry, location points of interest may include: Beijing University Student Entrepreneurship Park, Peking University, and Beijing Jiaotong University, etc.; if a user enters "No. 5, Yiheyuan Road, Haidian District, Beijing" in the query entry, location points of interest may include Peking University, etc.

[0073] S150. Obtain a first set of hot words and a second set of hot words from the candidate information set. The first set of hot words is determined based on the popularity score of each candidate information within a first time period, and the second set of hot words is determined based on the popularity score of each candidate information within a second time period. The first time period includes the second time period, and the duration of the first time period is longer than the duration of the second time period.

[0074] It should be noted that the first and second hot keyword sets can be collectively referred to as hot keyword sets. Hot keyword sets represent candidate information with high click-through rates and / or high popularity over a certain period. The first hot keyword set refers to the collection of candidate information with high click-through rates and / or high popularity over a long period; it is also called long-term hot keyword set. For example, the first hot keyword set includes candidate information with high click-through rates and / or high popularity within the past year, where "past year" refers to the past year starting from the date the query request was received, excluding the date of receipt. The second hot keyword set refers to the collection of candidate information with high click-through rates and / or high popularity within a short period; it is also called short-term hot keyword set. For example, the second hot keyword set includes candidate information with high click-through rates and / or high popularity within the past two weeks, where "past two weeks" refers to the past two weeks starting from the date the query request was received, excluding the date of receipt.

[0075] It is understandable that, since the first and second sets of hot words were selected at the same time, there is an overlap between the first and second sets of hot words, or the first set of hot words may contain the second set of hot words.

[0076] For example, if the date the query request is received is January 15, 2022, then the first hot word set includes all candidate information in the candidate information set whose popularity score meets the threshold from January 15, 2021 to January 14, 2022, and the second hot word set includes all candidate information in the candidate information set whose popularity score meets the threshold from January 1, 2022 to January 14, 2022.

[0077] S170. Generate an information recommendation index based on the first set of hot words and the second set of hot words.

[0078] It should be noted that the information recommendation index refers to a recommendation list composed of locations of interest with high click-through rates and / or high popularity.

[0079] It is understandable that the information recommendation index can either identify whether each candidate information belongs to the first or second set of hot words, or it can choose not to identify whether each candidate information belongs to the first or second set of hot words.

[0080] Optionally, all candidate information can be sorted according to the popularity score of each candidate information in the first hot word set and the second hot word set, thereby generating an information recommendation index.

[0081] S190: Feedback information to the terminal recommending an index.

[0082] Understandably, the server will send the generated information recommendation index to the terminal.

[0083] This application provides an information recommendation method, mainly including: First, the server obtains a set of related candidate information based on the query information in the information query request sent by the terminal. The candidate information set originates from retrieval recommendation logs, retrieval click logs, and third-party click logs. Next, the server divides the candidate information set into a first hot word set and a second hot word set based on the time span and the popularity score of the candidate information. Finally, the server generates an information recommendation index based on the first and second hot word sets and feeds the information recommendation index back to the terminal. This application solves the problem of a single data source for the candidate set. By expanding the data sources of the candidate information set, this application improves the coverage of hot word data. In this application, the data sources of the candidate set include: retrieval recommendation logs, retrieval click logs, and third-party click logs. This application also solves the problems of long coverage periods, high time consumption, and large data volume of hot word data. By distinguishing hot word sets with different time spans, this application simultaneously achieves the characteristics of high stability and strong real-time performance of hot words. The first hot word set contains long-term hot words, resulting in higher stability, while the second hot word set contains short-term hot words, resulting in stronger real-time performance. This application can better meet user needs and improve the conversion rate of the information recommendation index.

[0084] In this application Figure 2 In the first optional embodiment of the information recommendation method provided in the corresponding embodiment, please refer to... Figure 3 Step S150 further includes: steps S151 to S157; specifically:

[0085] S151. Within the first time period, obtain candidate information with a popularity score greater than the first threshold from the candidate information set to obtain the first candidate hot word set.

[0086] It is understandable that the candidate information in the first candidate hot word set satisfies the condition that the popularity score is greater than the first threshold within the first time period.

[0087] S153. Update the first candidate hot word set according to the first sub-time period to obtain the first hot word set.

[0088] The duration of the first time period is longer than the duration of the first sub-time period.

[0089] It should be noted that the first sub-time period refers to the update cycle for updating the first candidate hot word set.

[0090] Understandably, the candidate information in the first candidate hot word set is periodically updated, and the candidate information that still satisfies the condition of having a popularity score greater than a first threshold within the first time period after the periodic update is formed into the first hot word set. The first hot word set is a subset of the first candidate hot word set.

[0091] For example, the first set of candidate hot words with a popularity score greater than the first threshold is updated every two weeks to obtain the first set of hot words; where the first time period is the past year and the first sub-time period is the past two weeks.

[0092] S155. During the second time period, candidate information with a popularity score greater than the second threshold is obtained from the candidate information set to obtain the second candidate hot word set.

[0093] It is understandable that the candidate information in the second candidate hot word set has the characteristic that the popularity score is greater than the second threshold in the second time period.

[0094] S157. Update the second candidate hot word set according to the second sub-time period to obtain the second hot word set.

[0095] The duration of the second time period is longer than the duration of the second sub-time period.

[0096] It should be noted that the second sub-time period refers to the update cycle for updating the second candidate hot word set. The first and second hot word sets will have some overlap. For overlapping candidate information, there are both long-term and short-term click counts. For non-overlapping candidate information, the click attributes that do not exist are assigned 0. These parameters will all participate as features in the ranking of candidate information in the information recommendation index during the information recommendation index generation process.

[0097] Understandably, the candidate information in the second candidate hot word set is periodically updated, and the candidate information that still satisfies the condition of having a popularity score greater than the second threshold within the second time period after the periodic update is formed into the second hot word set. The second hot word set is a subset of the second candidate hot word set.

[0098] For example, the second set of candidate hot words with a popularity score greater than the second threshold is updated daily to obtain the second set of hot words; where the second time period is the past two weeks and the second sub-time period is the past day.

[0099] The information recommendation method provided in this application generates a first set of hot words consisting of candidate information with high popularity scores over a long period and a second set of hot words consisting of candidate information with high popularity scores over a short period. The first set of hot words contains candidate information with a longer timeframe, making the data more reliable. However, to ensure the stability of the first set, updates are slower and the update cycle is longer. The second set of hot words contains candidate information with a shorter timeframe and smaller data volume, allowing for faster reflection of candidate information with rising popularity in the short term. It has a shorter update cycle and better real-time characteristics. The information recommendation index generated using the first and second hot word sets combines stability and real-time performance, improving search performance.

[0100] In this application Figure 2 In a second optional embodiment of the information recommendation method provided in the corresponding embodiment, please refer to... Figure 4 The process includes step S100 before step S110; specifically:

[0101] S100. The popularity score of each candidate information is obtained by weighting the effective clicks in the retrieval recommendation log, the effective clicks in the retrieval click log, and the effective clicks in the third-party click log.

[0102] It should be noted that the popularity score for each candidate piece of information is calculated by weighting the number of valid clicks in the retrieval recommendation log, the number of valid clicks in the retrieval click log, and the number of valid clicks in the third-party click log. Specifically, the popularity score can be calculated using the following formula:

[0103] clkscore =w sug *sug_clk+w search *search_clk+w app *app_clk;

[0104] Among them, clk score w represents the popularity score corresponding to the candidate information. sug This represents the weight coefficient of the retrieval recommendation log, sug_clk represents the effective click count of candidate information in the retrieval recommendation log, and w search This represents the weighting coefficient of the search click log, search_clk represents the effective click count of candidate information in the search click log, and w app This represents the weighting coefficient of the third-party click logs, and app_clk represents the effective click count of the candidate information in the third-party click logs.

[0105] Optionally, since candidate information is mainly used in retrieval and recommendation, the calculation of the popularity score corresponding to the candidate information is mainly based on the effective click volume of the candidate information in the retrieval and recommendation log. The effective click volumes from other sources are added together with a certain weight, i.e., w sug Satisfy w sug =1, w search Satisfy 0 <w search <1, w app Satisfy 0 <w app <1. When the number of valid clicks on candidate information in third-party click logs is too high, the weighting coefficient of third-party click logs needs to be appropriately reduced.

[0106] Optionally, to ensure the stability of the candidate information in the first hot word set, w can be set. sug =1, w search =0.5, w app =0.5, and retain the heat score to satisfy clk. score Candidates with ≥4 or more words are used as the first candidate hot word set. To ensure the real-time nature of the second hot word set, w can be set. sug =1, w search =1, w app =0.5, and retain the heat score to satisfy clk. score Candidate information with a value of ≥2.5 is used as the second candidate hot word set.

[0107] The information recommendation method provided in this application improves the accuracy of the heat score calculation and the reliability of the ranking of candidate information in the information recommendation index by weighting the click volume contained in the logs from three sources when calculating the heat score of candidate information.

[0108] In this application Figure 2In a third optional embodiment of the information recommendation method provided in the corresponding embodiment, please refer to... Figure 5 The process further includes steps S210 to S230 after step S190; specifically:

[0109] S210, The receiving terminal sends an information click request based on the information recommendation index.

[0110] It should be noted that an information click request refers to a user's response action to candidate information in the information recommendation index via a terminal.

[0111] Understandably, users first enter their query information through the query portal, and then click on any candidate information in the information recommendation index displayed below; after the terminal detects the information click request acting on the information recommendation index, it sends the information click request to the server.

[0112] For example, a user first enters "Peking University" into the search portal of the map client on the terminal, and then clicks "North Gate of Peking University" in the displayed information recommendation index; the client first receives the user's response action, and then sends the information click request corresponding to the response action to the server; the server receives the information click request sent by the terminal.

[0113] S230. Update the popularity score of candidate information in the candidate information set according to the information click request.

[0114] It should be noted that each click request for a candidate message will count as a click on that candidate message, so the popularity score of each candidate message needs to be updated after it is clicked.

[0115] Understandably, the process involves first recalling the clicked candidate information to the candidate information set, then calculating the popularity score of the candidate information, and finally classifying the candidate information into the first or second hot word set based on the updated popularity score.

[0116] Optionally, during the recall process, the number of candidate information recalled to the first and second hot word sets needs to be controlled. For candidate information recalled to the first hot word set, the original click volume in the retrieval recommendation log is used as a reference standard. Based on the original click volume in the retrieval recommendation log, all candidate information recalled to the first hot word set is sorted in descending order to obtain the first hot word candidate recall set. A recall quantity for the first hot word set is set, and candidate information in the first hot word candidate recall set that meets the recall quantity of the first hot word set is categorized into the first hot word set. For candidate information recalled to the second hot word set, both the original click volume in the retrieval recommendation log and the original click volume in the retrieval click log are used as reference standards. Based on the original click volume in the retrieval recommendation log and the original click volume in the retrieval click log, all candidate information recalled to the second hot word set is sorted in descending order to obtain the second hot word candidate recall set. A recall quantity for the second hot word set is set, and candidate information in the second hot word candidate recall set that meets the recall quantity of the second hot word set is categorized into the second hot word set. If the number of candidate information in the second hot word candidate recall set is insufficient to meet the recall quantity of the second hot word set, then the candidate information in the third-party click log is sorted in descending order according to the original click volume in the third-party click log to obtain the candidate recall set of the third-party click log, and candidate information classified into the second hot word set is selected from the candidate recall set of the third-party click log.

[0117] The information recommendation method provided in this application recalls candidate information clicked by users and updates the popularity score of the clicked candidate information during the recall. The updated popularity score is then used as the basis for classifying the candidate information into a first hot word set or a second hot word set. This achieves the recall of candidate information, making the information recommendation index composed of the first hot word set and the second hot word set more reliable and improving the conversion rate of retrieval and recommendation.

[0118] In practical applications, users may repeatedly click on a particular piece of information in order to increase its appearance in the information recommendation index or to improve its ranking. This increases the number of clicks and thus the popularity score of that information, ultimately leading to its inclusion or higher ranking in the index. This behavior diminishes the authenticity and reliability of the rankings within the information recommendation index.

[0119] To address the aforementioned issues, this application... Figure 5 In one optional embodiment of the information recommendation method provided in the corresponding implementation, the information click request carries a target object identifier and target query information, where the target query information is any candidate information in the information recommendation index. Please refer to... Figure 6Step S230 further includes: steps S2301 to S2303; specifically:

[0120] S2301. Based on the target object identifier and the target query information, perform deduplication processing on the log records in the associated logs to obtain the effective click volume of the target query information in the associated logs. The associated logs include: retrieval recommendation logs, retrieval click logs, and third-party click logs.

[0121] It should be noted that the target object identifier refers to unique identification data used to represent the object sending the information click request; for example, the target object identifier can be the unique identification data of the terminal or the user's account ID logged in on the client. The associated log refers to the data source of the target query information. Deduplication means that when the same target object identifier is clicked repeatedly on a certain target query information, only one click is retained as a valid click.

[0122] Understandably, in order to prevent high click counts caused by the same user repeatedly clicking on the same target query information, it is necessary to filter the click data in the associated log records based on the target object identifier that initiated the information click request, so that multiple clicks by the same user on the same target query information are counted as only one click.

[0123] For example, if user A enters "barbecue restaurant" into the search bar 5 times within a certain period of time, and then clicks "ABC Barbecue" in the information recommendation index each time, the effective new click count for "ABC Barbecue" is 1; if user B enters "barbecue restaurant" into the search bar multiple times, and then clicks "ABC Barbecue" only once in the information recommendation index, the effective new click count for "ABC Barbecue" is 1.

[0124] S2303. Update the popularity score of the target query information based on the number of valid clicks in the associated logs.

[0125] Understandably, after deduplication, the effective click count of the target query information is obtained, and the popularity score corresponding to the target query information is recalculated based on the associated logs.

[0126] The information recommendation method provided in this application performs deduplication processing on repeated clicks of the same target object identifier and the same target query information by the target object identifier and the target query information. This ensures that multiple clicks of the same target query information by the same user in the information recommendation index generated by the same query information are counted as only one valid click. This effectively prevents users from cheating, increases the authenticity of the popularity score of candidate information, and improves the authenticity and reliability of the ranking of candidate information in the information recommendation index.

[0127] Please see Figure 7 Another embodiment of this application provides an information recommendation method including steps S120 to S200. Specifically:

[0128] S120: Receive an information query request sent by the terminal. The information query request carries query information and the terminal's location information.

[0129] S140. Based on the information query request, obtain a candidate information set that is associated with both location information and the query information. The candidate information set is derived from at least two of the following: retrieval recommendation logs, retrieval click logs, and third-party click logs. The candidate information set includes K candidate information items, each corresponding to a popularity score, where K is an integer greater than 1.

[0130] S160. Obtain a first set of hot words and a second set of hot words from the candidate information set. The first set of hot words is determined based on the popularity score of each candidate information within a first time period, and the second set of hot words is determined based on the popularity score of each candidate information within a second time period. The first time period includes the second time period, and the duration of the first time period is longer than the duration of the second time period.

[0131] S180. Based on the terminal's location information, generate an information recommendation index according to the first hot word set and the second hot word set.

[0132] S200: Feedback information to the terminal recommending an index.

[0133] It should be noted that location information is the location information obtained by the server after the terminal grants the server permission to access the terminal's location information. The location information can be the terminal's current prefecture-level administrative region, county-level administrative region, township-level administrative region, or the terminal's current street, or the terminal's current geographic coordinates, etc.

[0134] Understandably, after detecting the query information entered by the user at the query entry point, the terminal uploads the query information to the server, along with the terminal's geographical location information. When generating the information recommendation index, the server uses the popularity scores corresponding to candidate information in the first and second hot word sets, along with the terminal's location information, as the basis for ranking the candidate information in the information recommendation index.

[0135] The information recommendation method provided in this application improves the conversion rate of retrieval and recommendation by obtaining the location information of the terminal and using the location information of the terminal as one of the criteria for ranking candidate information in the generation of information recommendation index.

[0136] In this application Figure 2 In a fourth optional embodiment of the information recommendation method provided in the corresponding embodiment, step S170 further includes:

[0137] Based on the popularity scores of all candidate information in the first and second hot word sets, the candidate information in the first and second hot word sets is sorted in descending order to generate an information recommendation index.

[0138] It is understandable that when generating the information recommendation index, all candidate information from the first and second hot word sets will be displayed in the information recommendation index in descending order of popularity score.

[0139] The information recommendation method provided in this application arranges all candidate information in the first hot word set and the second hot word set in descending order of popularity score in the information recommendation index, and prioritizes displaying candidate information with high popularity score at the top of the information recommendation index to improve interactivity.

[0140] To facilitate understanding, the following will combine... Figure 8 This paper introduces an information recommendation scenario in a map search application. The goal of this application is to optimize the search process and improve the conversion rate of the information recommendation index by combining long-term and short-term hot keyword data with multi-source click logs. The input data for the information recommendation method in this application scenario includes: the information query request sent by the terminal and the terminal's location information; the output data is an information recommendation index sorted according to the popularity scores of all hot keywords in both long-term and short-term hot keywords. The information query request carries query information, which refers to the query text entered by the user when using map search, and the terminal's location information refers to the geographical location of the user when initiating the search request. Based on candidate information from the long-term and short-term hot keyword sets, a hot keyword recall and ranking process is performed. (See also...) Figure 8 , Figure 8 This is a schematic diagram of the information recommendation method in the application scenario of this application, as shown in the figure. Specifically:

[0141] Offline portion:

[0142] The server obtains a candidate information set, which includes: candidate information from search and recommendation logs over the past year, candidate information from search and click logs over the past year, candidate information from third-party click logs over the past year, candidate information from search and recommendation logs over the past two weeks, candidate information from search and click logs over the past two weeks, and candidate information from third-party click logs over the past two weeks.

[0143] The server integrates candidate information from nearly a year's worth of search recommendation logs, candidate information from nearly a year's worth of search click logs, and candidate information from nearly a year's worth of third-party click logs to generate a long-term hot keyword set; the long-term hot keyword set is updated every two weeks.

[0144] The server performs multi-source data fusion processing on candidate information from the search recommendation logs of the past two weeks, candidate information from the search click logs of the past two weeks, and candidate information from the third-party click logs of the past two weeks to generate a short-term hot keyword set; the short-term hot keyword set is updated daily.

[0145] Online section:

[0146] The server sorts all candidate information in both the long-term and short-term hot word sets based on the long-term click volume in the retrieval recommendation log, the long-term click volume in the retrieval click log, and the long-term click volume in the third-party click log for each candidate information in the long-term hot word set, as well as the real-time click volume in the retrieval recommendation log, the real-time click volume in the retrieval click log, and the real-time click volume in the third-party click log for each candidate information in the short-term hot word set, and obtains an information recommendation index list.

[0147] The server performs hot word recall on all candidate information in the long-term hot word set and all candidate information in the short-term hot word set, generating a recall set.

[0148] It should be noted that multi-source data fusion processing refers to merging candidate information from retrieval recommendation logs, retrieval click logs, and third-party click logs. Multi-source data fusion processing is achieved through a multi-source data fusion model; both long-term and short-term hot keyword sets are generated based on this model.

[0149] Please see Figure 9 , Figure 9 This is a schematic diagram of the multi-source data fusion processing in this application, as shown in the figure. Specifically:

[0150] The server first obtains candidate information from the retrieval recommendation log, retrieval click log, and third-party click log, thus obtaining candidate information for the retrieval recommendation log, candidate information for the retrieval click log, and candidate information for the third-party click log.

[0151] The server then performs user anti-fraud processing and regularization processing on the candidate information of the retrieval recommendation log, the candidate information of the retrieval click log, and the candidate information of the third-party click log in sequence to obtain the valid candidate information of the retrieval recommendation log, the valid candidate information of the retrieval click log, and the valid candidate information of the third-party click log.

[0152] The server then merges the valid candidate information from the retrieval recommendation log, the valid candidate information from the retrieval click log, and the valid candidate information from the third-party click log to obtain a hot word set.

[0153] The server then expands the valid candidate information in the hot word set to obtain expanded information.

[0154] The server finally generates an information recommendation index based on the extended information.

[0155] It should be noted that, in order to prevent the situation of high click-through rates caused by the same user repeatedly clicking on the same candidate information, this application proposes user anti-fraud processing. Specifically, the click data is filtered for deduplication based on the target object identifier of the request, so that multiple clicks by the same user on the same candidate information for the same query are counted as only one click.

[0156] Regularization refers to processing query information by converting full-width characters to half-width characters, traditional Chinese characters to simplified Chinese characters, removing punctuation, converting case, and converting encoding.

[0157] Merging refers to combining click data from search recommendation logs, search click logs, and third-party click logs, padding missing click counts with zeros. The merged data format is: query|adcode|clk score |sug_clk|search_clk|app_clk, where query represents query information, adcode represents location information, sug_clk represents the click volume from the search recommendation click log, search_clk represents the click volume from the search click log, app_clk represents the click volume from the third-party click log, and clk represents the click volume from the search recommendation click log. score This represents the popularity score calculated based on a weighted average of three click-through rates. By setting a popularity threshold, candidate keywords with scores higher than the threshold are retained to generate a set of trending keywords. The popularity score is calculated as follows:

[0158] clk score =sug_clk+w search *search_clk+w app *app_clk;

[0159] Among them, w search w represents the weighting coefficient of the retrieval click logs. app This represents the weighting coefficient of third-party click logs. Since hot keyword information is primarily used in search and recommendation services, clk... score The calculation primarily relies on clicks from the recommended click logs, with clicks from other sources weighted accordingly. This means that a minimum of 0. <w search <1,0 <w app <1. If the number of clicks in the third-party click logs is too high, the number of clicks in the third-party click logs can be appropriately reduced, that is, the weight coefficient of the third-party click logs can be lowered.

[0160] For long-term hot word sets, to ensure the stability of candidate information, w can be set. search =0.5,wapp =0.5; For short-term hot word sets, to improve the real-time nature of candidate information, w can be set. search =1.0,w app =0.5. When merging the long-term hot word set and the short-term hot word set, clk can be retained. score Candidate information with a value of ≥4 is used as candidate information in the long-term hot word set, and clk is retained. score Candidate information with a value of ≥2.5 is used as candidate information in the short-term hot word set.

[0161] Data expansion refers to the phenomenon in retrieval and recommendation services where shorter data points corresponding to the user's target query indicate better service performance. Therefore, for information recommendation indexes, the prefix of the user's target query is highly relevant. If the target query is recommended based on the user's input of a prefix, the user can quickly find the desired query. Thus, prefix expansion based on query information is necessary. Specifically:

[0162] If the query information is in full Pinyin, then a certain number of prefixes are used to expand the Pinyin into new data. If the query information is not in full Pinyin, then a certain number of prefixes are first used to expand the original query information, and then the query information is converted into full Pinyin. Based on the full Pinyin, a certain number of Pinyin prefixes are expanded. For the expanded candidate information, the click volume becomes the click volume of the corresponding expanded information. The original clicks in the retrieval recommendation log, the original clicks in the retrieval click log, and the original clicks in the third-party log are all expanded into four types of click counts: original clicks, expanded clicks, Pinyin clicks, and Pinyin expanded clicks. Therefore, the dimensions of the candidate information become 12 dimensions. After the data is expanded, the click volumes of these 12 dimensions for the same query|adcode|md are merged.

[0163] Understandably, the method provided in this application scenario generates a long-term hot keyword set containing clicks over the past year and a short-term hot keyword set containing clicks over the past two weeks. The long-term hot keyword set covers candidate information with a longer click cycle and higher click volume, and its update cycle is only two weeks, resulting in slower updates and greater stability. Conversely, the short-term hot keyword set covers candidate information with a shorter click cycle and lower click volume, and its update cycle is daily, allowing for faster updates and reflecting candidate information with rapidly increasing popularity in the short term, thus ensuring high real-time performance. By merging the long-term and short-term hot keyword sets during the information recommendation index ranking process, the information recommendation index achieves both stability and real-time performance, improving search performance.

[0164] During the recall process, it is necessary to control the number of candidate information from both the long-term and short-term hot keyword sets added to the recall set. For the long-term hot keyword set, the original click volume in the search recommendation log is used as a reference standard. Candidate information from the long-term hot keyword set is sorted in descending order based on the original click volume in the search recommendation log, and the candidate information with the highest ranking is prioritized for addition to the recall set according to the set recall quantity. For the short-term hot keyword set, candidate information with high click volume in the search click log needs to be reflected in the search recommendations in a timely manner. Therefore, the candidate information in the short-term hot keyword set needs to be sorted in descending order based on both the original click volume in the search recommendation log and the original click volume on the search click day. The candidate information with the highest ranking is prioritized for addition to the recall set according to the set recall quantity. If the number of candidate information added to the recall set is insufficient, candidate information with only original clicks from third-party click logs is added to the recall set.

[0165] Since the time span of candidate data in the long-term hot word set includes the time span of candidate data in the short-term hot word set, there will be overlap between the candidate information in the short-term hot word set and the candidate information in the long-term hot word set. For the overlapping candidate information, there are both long-term clicks and short-term clicks. For the non-overlapping candidate information, the click attributes that do not exist are assigned 0. Both long-term clicks and short-term clicks are used as features in the training of the ranking model.

[0166] In the ranking process, it is necessary to consider not only the long-term click volume in the retrieval recommendation click log, the long-term click volume in the retrieval click log, and the long-term click volume in the third-party click log, but also the real-time click volume in the retrieval recommendation click log, the real-time click volume in the retrieval click log, and the real-time click volume in the third-party click log. All the click volumes are input into the ranking model to obtain the information recommendation index.

[0167] To verify the functional characteristics of the technical solution, the method proposed in this application was tested. Using 1000 randomly generated online query results, the existing technologies and the method proposed in this application were compared to evaluate the success rate of the method provided in this application. The test showed that the success rate of the method provided in this application was 59%. The success rate was calculated as follows:

[0168]

[0169] Comparing the conversion rates of the experimental container containing the method proposed in this application and the control container containing the prior art between November 5th and November 11th, 2021, the comparison results are as follows: Figure 10 As shown, A represents the conversion rate of candidate data in the retrieval recommendation log in the method provided in this application, and B represents the conversion rate of candidate data in the retrieval recommendation log in the existing scheme.

[0170] The information recommendation device in this application is described in detail below. Please refer to [link / reference]. Figure 11 , Figure 11 This is a schematic diagram of one embodiment of the information recommendation device 10 in this application. The information recommendation device 10 includes:

[0171] The information query request receiving module 110 is used to receive information query requests sent by the terminal. The information query request carries query information.

[0172] The candidate information set acquisition module 130 is used to acquire a candidate information set associated with the queried information based on the information query request. The candidate information set is derived from at least two of the following: retrieval recommendation logs, retrieval click logs, and third-party click logs. The candidate information set includes K candidate information items, each corresponding to a popularity score, where K is an integer greater than 1.

[0173] The hot word set acquisition module 150 is used to acquire a first hot word set and a second hot word set from the candidate information set. The first hot word set is determined based on the popularity score of each candidate information within a first time period, and the second hot word set is determined based on the popularity score of each candidate information within a second time period. The first time period includes the second time period, and the duration of the first time period is longer than the duration of the second time period.

[0174] The information recommendation index generation module 170 is used to generate an information recommendation index based on the first set of hot words and the second set of hot words.

[0175] The information recommendation index sending module 190 is used to send the information recommendation index back to the terminal.

[0176] Compared to existing technologies where the candidate set data source is singular, this application's embodiment improves the coverage of hot word data by expanding the data sources of the candidate information set. In this embodiment, the candidate set data sources include: search recommendation logs, search click logs, and third-party click logs. Compared to existing technologies where hot word data is based on click volume over the past year, resulting in long coverage periods, high time consumption, and large data volumes, this application's embodiment distinguishes between hot word sets with different time spans, achieving both high stability and strong real-time performance. The first hot word set contains long-term hot words, offering higher stability, while the second hot word set contains short-term hot words, providing stronger real-time performance. This application's embodiment can better meet user needs and improve the conversion rate of information recommendation indexes.

[0177] In this application Figure 11 In the first optional embodiment of the information recommendation system provided in the corresponding implementation, please refer to... Figure 12 The hot word set acquisition module 150 includes: a first hot word set acquisition module 1501 and a second hot word set acquisition module 1502.

[0178] The first hot word set acquisition module 1501 is used to acquire candidate information with a popularity score greater than a first threshold from the candidate information set within a first time period to obtain a first candidate hot word set; and to update the first candidate hot word set according to a first sub-time period to obtain a first hot word set.

[0179] The duration of the first time period is longer than the duration of the first sub-time period.

[0180] The second hot word set acquisition module 1502 is used to acquire candidate information with a popularity score greater than a second threshold from the candidate information set within a second time period to obtain a second candidate hot word set; and to update the second candidate hot word set according to the second sub-time period to obtain a second hot word set.

[0181] The duration of the second time period is longer than the duration of the second sub-time period.

[0182] The information recommendation device provided in this application generates a first hot word set consisting of candidate information with high popularity scores over a long period and a second hot word set consisting of candidate information with high popularity scores over a short period. The first hot word set contains candidate information with a longer timeframe, making the data more reliable. However, to ensure the stability of the first hot word set, updates are slower and the update cycle is longer. The second hot word set contains candidate information with a shorter timeframe and smaller data volume, allowing for faster reflection of candidate information with rising popularity in the short term. It has a shorter update cycle and better real-time characteristics. The information recommendation index generated using the first and second hot word sets combines stability and real-time performance, improving search performance.

[0183] In this application Figure 11 In a second optional embodiment of the information recommendation device provided in the corresponding embodiment, please refer to... Figure 13 In another embodiment of this application, the information recommendation device further includes a popularity score calculation module 100.

[0184] The popularity score calculation module 100 is used to perform weighted calculation based on the effective clicks in the retrieval recommendation log, the effective clicks in the retrieval click log, and the effective clicks in the third-party click log for each candidate information to obtain the popularity score for each candidate information.

[0185] The information recommendation device provided in this application improves the accuracy of the popularity score calculation and enhances the reliability of the ranking of candidate information in the information recommendation index by weighting the click volume contained in the logs from three sources when calculating the popularity score of candidate information. This fully considers the click volume of candidate information in different sources and the weight of each source.

[0186] In this application Figure 11 In a third optional embodiment of the information recommendation device provided in the corresponding embodiment, please refer to... Figure 14 The information recommendation device 10 also includes: an information recall module 210 and a first module 230 for updating popularity scores.

[0187] The information recall module 210 is used to receive information click requests sent by the terminal based on the information recommendation index.

[0188] The first module 230 for updating the popularity score is used to update the popularity score of candidate information in the candidate information set based on the information click request.

[0189] The information recommendation device provided in this application recalls candidate information clicked by users and updates the popularity score of the clicked candidate information during the recall. The updated popularity score is then used as the basis for classifying the candidate information into a first hot word set or a second hot word set. This achieves the recall of candidate information, making the information recommendation index composed of the first hot word set and the second hot word set more reliable and improving the conversion rate of retrieval recommendation.

[0190] In practice, users may repeatedly click on a particular piece of information in order to increase its appearance or ranking in the recommendation index. This increases the number of clicks and thus the popularity score, ultimately leading to its inclusion or higher ranking. This behavior diminishes the authenticity and reliability of the rankings within the recommendation index.

[0191] To address the aforementioned issues, this application... Figure 14 In one optional embodiment of the information recommendation device provided in the corresponding implementation, the information click request carries a target object identifier and target query information, where the target query information is any candidate information in the information recommendation index. Please refer to... Figure 15 The first module 230 for updating the popularity score also includes: a weight filtering module 2301 and a second module 2302 for updating the popularity score.

[0192] The deduplication module 2301 is used to perform deduplication processing on log records in the associated logs based on the target object identifier and target query information to obtain the effective click volume of the target query information in the associated logs. The associated logs include: retrieval recommendation logs, retrieval click logs, and third-party click logs.

[0193] The second module 2302 for updating the popularity score is used to update the popularity score of the target query information based on the number of valid clicks on the target query information in the associated logs.

[0194] The information recommendation device provided in this application performs deduplication processing on repeated clicks of the same target object identifier and the same target query information by the target object identifier and the target query information. This ensures that multiple clicks of the same target query information by the same user in the information recommendation index generated by the same query information are counted as only one valid click. This effectively prevents users from cheating, increases the authenticity of the popularity score of candidate information, and improves the authenticity and reliability of the ranking of candidate information in the information recommendation index.

[0195] In this application Figure 10 In a fourth optional embodiment of the information recommendation device provided in the corresponding embodiment, the information query request carries the location information of the terminal. The candidate information set acquisition module 130 is further configured to acquire a candidate information set that is associated with the location information and the query information according to the information query request.

[0196] The information recommendation device provided in this application improves the conversion rate of retrieval and recommendation by obtaining the location information of the terminal and using the location information of the terminal as one of the criteria for sorting candidate information in the generation of information recommendation index.

[0197] In this application Figure 10 In a fifth optional embodiment of the information recommendation device provided in the corresponding embodiment, the information recommendation index generation module 170 is further configured to sort all candidate information in the first hot word set and the second hot word set in descending order according to the popularity scores of all candidate information in the first hot word set and the second hot word set, and generate an information recommendation index.

[0198] The information recommendation device provided in this application arranges all candidate information in the first hot word set and the second hot word set in descending order of popularity score in the information recommendation index, and prioritizes displaying candidate information with high popularity score at the top of the information recommendation index to improve interactivity.

[0199] Figure 16This is a schematic diagram of a server structure provided in an embodiment of this application. The server 300 can vary significantly due to different configurations or performance. It may include one or more central processing units (CPUs) 322 (e.g., one or more processors) and memory 332, and one or more storage media 330 (e.g., one or more mass storage devices) for storing application programs 342 or data 344. The memory 332 and storage media 330 can be temporary or persistent storage. The program stored in the storage media 330 may include one or more modules (not shown in the diagram), each module may include a series of instruction operations on the server. Furthermore, the CPU 322 may be configured to communicate with the storage media 330 and execute the series of instruction operations stored in the storage media 330 on the server 300.

[0200] Server 300 may also include one or more power supplies 326, one or more wired or wireless network interfaces 350, one or more input / output interfaces 358, and / or one or more operating systems 341, such as Windows Server. TM Mac OS X TM Unix TM Linux TM FreeBSD TM etc.

[0201] The steps performed by the server in the above embodiments can be based on this Figure 16 The server structure shown.

[0202] It is understood that in the specific implementation of this application, user information, user click data and user query data and other related data are involved. When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0203] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0204] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0205] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0206] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0207] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0208] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. An information recommendation method, characterized in that, include: The receiving terminal sends an information query request, wherein the information query request carries query information; Based on the information query request, a candidate information set associated with the query information is obtained. The candidate information set is derived from at least two of the following: retrieval recommendation logs, retrieval click logs, and third-party click logs. The retrieval recommendation logs are logs generated by clicking on target query information in the information recommendation index generated based on the input query information. The retrieval click logs are logs generated by performing a search operation directly after inputting the query information. The third-party logs are logs generated by clicking on links in third-party clients or mini-programs. The candidate information set includes K candidate information items, each corresponding to a popularity score, where K is an integer greater than 1. A first hot word set and a second hot word set are obtained from the candidate information set. The first hot word set is a long-term hot word set, and the second hot word set is a short-term hot word set. The first hot word set is determined based on the popularity score of each candidate information within a first time period. The popularity score is determined by weighting the effective clicks of each candidate information in the search recommendation log, the effective clicks in the search log, the effective clicks in the third-party click log, and the weight coefficients corresponding to different types of logs. The second hot word set is determined based on the popularity score of each candidate information within a second time period. The first time period includes the second time period, and the duration of the first time period is longer than the duration of the second time period. An information recommendation index is generated based on the first hot word set and the second hot word set; The information recommendation index is fed back to the terminal; The step of obtaining the first hot word set and the second hot word set from the candidate information set includes: Within a first time period, candidate information with a popularity score greater than a first threshold is obtained from the candidate information set to obtain a first candidate hot word set; The first candidate hot word set is updated according to the first sub-time period to obtain the first hot word set; During the second time period, candidate information with a popularity score greater than the second threshold is obtained from the candidate information set to obtain the second candidate hot word set; The second candidate hot word set is updated based on the second sub-time period to obtain the second hot word set; Wherein, the duration of the first time period is greater than the duration of the first sub-time period, and the duration of the second time period is greater than the duration of the second sub-time period.

2. The information recommendation method as described in claim 1, characterized in that, After feeding back the information recommendation index to the terminal, the method further includes: Receive information click requests sent by the terminal based on the information recommendation index; The popularity score of the candidate information in the candidate information set is updated based on the click request.

3. The information recommendation method as described in claim 2, characterized in that, The information click request carries the target object identifier and target query information, wherein the target query information is any candidate information in the information recommendation index; The step of updating the popularity score of candidate information in the candidate information set based on the information click request includes: Based on the target object identifier and the target query information, the log records in the associated logs are deduplicated to obtain the effective click volume of the target query information in the associated logs. The associated logs include: retrieval recommendation logs, retrieval click logs, and third-party click logs. The popularity score of the target query information is updated based on the number of valid clicks in the associated log.

4. The information recommendation method as described in claim 1, characterized in that, The information query request carries the location information of the terminal; The step of obtaining a candidate information set associated with the query information according to the information query request includes: Based on the information query request, obtain a set of candidate information that is associated with the location information and the query information.

5. The information recommendation method as described in claim 1, characterized in that, The step of generating an information recommendation index based on the first hot word set and the second hot word set includes: Based on the popularity scores of all candidate information in the first and second hot word sets, the candidate information in the first and second hot word sets is sorted in descending order to generate the information recommendation index.

6. An information recommendation device, characterized in that, include: An information query request receiving module is used to receive information query requests sent by a terminal, wherein the information query request carries query information; The candidate information set acquisition module is used to acquire a candidate information set associated with the query information according to the information query request. The candidate information set is derived from at least two of the following: retrieval recommendation logs, retrieval click logs, and third-party click logs. The retrieval recommendation logs are logs generated by clicking on target query information in the information recommendation index generated based on the input query information. The retrieval click logs are logs generated by performing a search operation directly after inputting the query information. The third-party logs are logs generated by clicking on links in third-party clients or mini-programs. The candidate information set includes K candidate information items, each corresponding to a popularity score, where K is an integer greater than 1. The hot word set acquisition module is used to acquire a first hot word set and a second hot word set from the candidate information set. The first hot word set is a long-term hot word set, and the second hot word set is a short-term hot word set. The first hot word set is determined based on the popularity score of each candidate information within a first time period. The popularity score is determined by weighting the effective clicks of each candidate information in the search recommendation log, the effective clicks in the search log, the effective clicks in the third-party click log, and the weight coefficients corresponding to different types of logs. The second hot word set is determined based on the popularity score of each candidate information within a second time period. The first time period includes the second time period, and the duration of the first time period is longer than the duration of the second time period. The information recommendation index generation module is used to generate an information recommendation index based on the first hot word set and the second hot word set. The information recommendation index sending module is used to send the information recommendation index back to the terminal; The hot word set acquisition module includes: a first hot word set acquisition module and a second hot word set acquisition module; The first hot word set acquisition module is used to acquire candidate information with a popularity score greater than a first threshold from the candidate information set within a first time period, thereby obtaining a first candidate hot word set; The first candidate hot word set is updated according to the first sub-time period to obtain the first hot word set; The second hot word set acquisition module is used to acquire candidate information with a popularity score greater than a second threshold from the candidate information set within a second time period, thereby obtaining a second candidate hot word set; The second candidate hot word set is updated based on the second sub-time period to obtain the second hot word set; Wherein, the duration of the first time period is greater than the duration of the first sub-time period, and the duration of the second time period is greater than the duration of the second sub-time period.

7. The apparatus according to claim 6, characterized in that, The information recommendation device further includes: an information recall module and a first module for updating popularity scores; The information recall module is used to receive information click requests sent by the terminal based on the information recommendation index; The first module for updating the popularity score is used to update the popularity score of candidate information in the candidate information set according to the information click request.

8. The apparatus according to claim 7, characterized in that, The information click request carries the target object identifier and target query information, wherein the target query information is any candidate information in the information recommendation index. The first module for updating the popularity score also includes: a deduplication module and a second module for updating the popularity score. The deduplication module is used to perform deduplication processing on log records in the associated logs based on the target object identifier and the target query information to obtain the effective click volume of the target query information in the associated logs. The associated logs include: retrieval recommendation logs, retrieval click logs, and third-party click logs. The second module for updating the popularity score is used to update the popularity score of the target query information based on the number of valid clicks on the target query information in the associated log.

9. The apparatus according to claim 6, characterized in that, The information query request carries the location information of the terminal; the candidate information set acquisition module is further configured to acquire a candidate information set that is associated with the location information and the query information according to the information query request.

10. The apparatus according to claim 6, characterized in that, The information recommendation index generation module is further configured to sort all candidate information in the first hot word set and the second hot word set in descending order based on the popularity scores of all candidate information in the first hot word set and the second hot word set, and generate the information recommendation index.

11. A computer device, characterized in that, include: Memory, transceiver, processor, and bus system; The memory is used to store programs; The processor is configured to execute a program in the memory, including executing the information recommendation method as described in any one of claims 1 to 5; The bus system is used to connect the memory and the processor to enable communication between the memory and the processor.

12. A computer-readable storage medium comprising instructions that, when executed on a computer, cause the computer to perform the information recommendation method as described in any one of claims 1 to 5.

13. A computer program product, comprising a computer program, characterized in that, The computer program is executed by a processor using the information recommendation method as described in any one of claims 1 to 5.

Images