A knowledge recommendation method, device, system and storage medium

Abstract

The present disclosure provides a knowledge recommendation method, device, system and storage medium, which relate to the fields of artificial intelligence technology, especially big data, natural language processing (NLP), and intelligent customer service technology. The specific implementation scheme is as follows: obtaining a problem to be solved submitted by a terminal device; obtaining first knowledge that solves the problem to be solved from a preset knowledge base; determining the correlation score between each other knowledge and the first knowledge based on the subject of the first knowledge and the subject of other knowledge in the preset knowledge base; determining second knowledge related to the first knowledge from other knowledge based on the correlation score; recommending the title of the second knowledge and the first knowledge to the terminal device, wherein the first knowledge is the answer to the problem to be solved and the title of the second knowledge is a related question to the problem to be solved.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, in particular to the field of big data, natural language processing (NLP), and intelligent customer service technology, and specifically to a knowledge recommendation method, device, system, and storage medium. Background Art

[0002] With the development and implementation of AI (Artificial Intelligence) technology, intelligent customer service has become the standard solution to replace manual customer service. Knowledge recommendation systems are a key component of intelligent customer service, used to recommend solutions to user problems. Summary of the Invention

[0003] The present disclosure provides a knowledge recommendation method, device, system and storage medium.

[0004] According to one aspect of the present disclosure, a knowledge recommendation method is provided, comprising:

[0005] Obtain pending issues submitted by terminal devices;

[0006] Acquire first knowledge for solving the problem to be solved from a preset knowledge base;

[0007] Determining, based on the subject of the first knowledge and the subjects of other knowledge in the preset knowledge base, an association score between each of the other knowledge and the first knowledge;

[0008] determining, from the other knowledge, second knowledge associated with the first knowledge according to the association score;

[0009] The title of the second knowledge and the first knowledge are recommended to the terminal device, wherein the first knowledge is the answer to the problem to be solved, and the title of the second knowledge is a related question of the problem to be solved.

[0010] According to another aspect of the present disclosure, a knowledge recommendation device is provided, comprising:

[0011] A first obtaining unit is used to obtain the unresolved issues submitted by the terminal device;

[0012] A second acquiring unit is configured to acquire first knowledge for solving the problem to be solved from a preset knowledge base;

[0013] a first determining unit, configured to determine, based on a subject of the first knowledge and subjects of other knowledge in the preset knowledge base, an association score between each of the other knowledge and the first knowledge;

[0014] a second determining unit, configured to determine, from the other knowledge, second knowledge associated with the first knowledge based on the association score;

[0015] A recommendation unit is used to recommend the title of the second knowledge and the first knowledge to the terminal device, wherein the first knowledge is the answer to the problem to be solved, and the title of the second knowledge is a related question of the problem to be solved.

[0016] According to a third aspect of the present disclosure, a knowledge recommendation system is provided, comprising:

[0017] at least one processor; and

[0018] a memory communicatively connected to the at least one processor; wherein,

[0019] The memory stores instructions that can be executed by the at least one processor. The instructions are executed by the at least one processor to enable the at least one processor to perform any of the above-mentioned knowledge recommendation methods.

[0020] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute any of the above-mentioned knowledge recommendation methods.

[0021] According to a fifth aspect of the present disclosure, a computer program product is provided, comprising a computer program, which implements any of the above-mentioned knowledge recommendation methods when executed by a processor.

[0022] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings are provided to facilitate a better understanding of the present invention and do not constitute a limitation of the present disclosure.

[0024] Figure 1 This is a first flow chart of the knowledge recommendation method provided by the embodiment of the present disclosure;

[0025] Figure 2 is a schematic diagram of a dialog box for associated knowledge provided by an embodiment of the present disclosure;

[0026] Figure 3 This is a second flow chart of the knowledge recommendation method provided by the embodiment of the present disclosure;

[0027] Figure 4 This is a schematic diagram of a dialog box for recommending hot questions provided by an embodiment of the present disclosure;

[0028] Figure 5This is a third flow chart of the knowledge recommendation method provided by the embodiment of the present disclosure;

[0029] Figure 6 This is a schematic diagram of a dialog box for recommending candidate titles provided by an embodiment of the present disclosure;

[0030] Figure 7 This is a fourth flow chart of the knowledge recommendation method provided by the embodiment of the present disclosure;

[0031] Figure 8 This is a fifth flow chart of the knowledge recommendation method provided by the embodiment of the present disclosure;

[0032] Figure 9 This is a sixth flow chart of the knowledge recommendation method provided by the embodiment of the present disclosure;

[0033] Figure 10 This is a seventh flow chart of the knowledge recommendation method provided by the embodiment of the present disclosure;

[0034] Figure 11 This is a schematic diagram of a dialog box for a directory question provided by an embodiment of the present disclosure;

[0035] Figure 12 This is an eighth flow chart of the knowledge recommendation method provided by the embodiment of the present disclosure;

[0036] Figure 13 This is a ninth flow chart of the knowledge recommendation method provided by the embodiment of the present disclosure;

[0037] Figure 14 This is a tenth flow chart of the knowledge recommendation method provided by the embodiment of the present disclosure;

[0038] Figure 15 This is a schematic diagram of the eleventh flow chart of the knowledge recommendation method provided by the embodiment of the present disclosure;

[0039] Figure 16 is a schematic diagram of a heterogeneous graph provided by an embodiment of the present disclosure;

[0040] Figure 17 This is a twelfth flow chart of the knowledge recommendation method provided by the embodiment of the present disclosure;

[0041] Figure 18 This is a structural diagram of a knowledge recommendation device provided by an embodiment of the present disclosure;

[0042] Figure 19 This is a first block diagram of a knowledge recommendation system for implementing the knowledge recommendation method of the embodiment of the present disclosure;

[0043] Figure 20This is a second block diagram of a knowledge recommendation system for implementing the knowledge recommendation method according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0044] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0045] The knowledge recommendation system is a key component of intelligent customer service, used to recommend solutions to user problems. Currently, knowledge recommendation systems rely on manual configuration to recommend solutions to users' problems. Specifically, the knowledge recommendation system receives a question from the user, determines the target knowledge corresponding to the question from pre-stored correspondences between questions and knowledge, and obtains related knowledge associated with the pre-configured target knowledge. The system then constructs a list of related questions based on the titles of the related knowledge, and recommends the target knowledge and related question list to the user.

[0046] The above knowledge recommendation method has the following problems:

[0047] First, the adaptation scenario is single and only supports knowledge recommendation in knowledge-related problem scenarios.

[0048] Second, it is necessary to manually associate related knowledge for each piece of knowledge. The greater the amount of knowledge, the more manpower is consumed.

[0049] Third, the related knowledge recommended by manual association is one-sided, especially for large knowledge bases with a knowledge level of more than 100,000. It is difficult for humans to master all the knowledge in the knowledge base, and it is difficult to globally associate related knowledge with each piece of knowledge, and thus it is difficult to globally recommend relevant knowledge to users.

[0050] To solve the above problems, the present disclosure provides a knowledge recommendation method, such as Figure 1 As shown, this method can be applied to a knowledge recommendation system. The knowledge recommendation system can be located on a terminal device where the user inputs questions. The terminal device is a device that interacts with the user and can be an electronic device such as a mobile phone, a PC (Personal Computer), or a tablet computer. The knowledge recommendation system can also be an independent physical machine, such as a knowledge recommendation system. For ease of understanding, the following description uses the knowledge recommendation system as the execution subject, which is not limiting. The knowledge recommendation method includes the following steps:

[0051] Step S11: Obtain the unresolved issues submitted by the terminal device.

[0052] In the embodiment of the present disclosure, a user can input a problem to be solved, i.e., a problem to be solved, into a terminal device. The terminal device submits the problem to be solved input by the user to the knowledge recommendation system, and the knowledge recommendation system then obtains the problem to be solved submitted by the terminal device.

[0053] Step S12: Acquire first knowledge for solving the problem to be solved from a preset knowledge base.

[0054] In the embodiment of the present disclosure, the preset knowledge base contains multiple pieces of knowledge. The preset knowledge base can be implemented using a MySQL (structured query language, relational database management system) database, an ES (ElasticSearch, distributed full-text search engine) database, or other forms of databases, without limitation.

[0055] After obtaining the problem to be solved, the knowledge recommendation system searches the preset knowledge base to obtain the answer to the problem, which is the first knowledge.

[0056] Step S13: Based on the subject of the first knowledge and the subjects of other knowledge in the preset knowledge base, determine the association score of each other knowledge with the first knowledge.

[0057] In the embodiment of the present disclosure, each piece of knowledge in the preset knowledge base may include one or more topics.

[0058] The knowledge recommendation system extracts the theme of each piece of knowledge from the preset knowledge base, specifically the theme of the first piece of knowledge, and then extracts the themes of other pieces of knowledge from the preset knowledge base. For each piece of other knowledge, the knowledge recommendation system determines a relevance score between the other piece of knowledge and the first piece of knowledge based on the theme of the first piece of knowledge and the themes of the other piece of knowledge. A higher relevance score indicates a higher degree of relevance between the other piece of knowledge and the first piece of knowledge, while a lower relevance score indicates a lower degree of relevance between the other piece of knowledge and the first piece of knowledge.

[0059] Step S14: Determine the second knowledge associated with the first knowledge from other knowledge based on the association score.

[0060] After obtaining the association score between each other knowledge and the first knowledge, the knowledge recommendation system can determine other knowledge associated with the first knowledge as the second knowledge according to the size of each association score.

[0061] In an optional embodiment, the knowledge recommendation system may determine a sixth preset number of other knowledge items with the highest association scores from other knowledge items, and then use the sixth preset number of other knowledge items as the second knowledge items associated with the first knowledge item.

[0062] The sixth preset number may be set according to actual needs. For example, the sixth preset number may be 5, 8 or 10, and there is no limitation on this.

[0063] In another optional embodiment, the knowledge recommendation system can determine other knowledge with a correlation score higher than a preset score threshold from other knowledge, and then use the other knowledge as the second knowledge associated with the first knowledge. The preset score threshold can be set according to actual needs and is not limited to this.

[0064] In the technical solution provided by the embodiments of the present disclosure, the relevance score of other knowledge items to the first knowledge item is proportional to the degree of relevance. This means that the knowledge recommendation system can obtain a limited number of pieces of knowledge items with the highest degree of relevance to the first knowledge item, namely, the second knowledge items. If the knowledge recommendation system recommends this second knowledge item to a terminal device, the user can use the terminal device to better solve the problem in question based on this second knowledge item with the highest degree of relevance, thereby improving the user's problem-solving efficiency. Furthermore, recommending only a limited number of pieces of second knowledge items to the user can conserve resources within the knowledge recommendation method and avoid inefficient resource consumption.

[0065] The knowledge recommendation system may also use other methods to determine the second knowledge, which is not limited.

[0066] Step S15: recommend the title of the second knowledge and the first knowledge to the terminal device, wherein the first knowledge is the answer to the problem to be solved, and the title of the second knowledge is a related question of the problem to be solved.

[0067] In the embodiment of the present disclosure, each piece of knowledge in the preset knowledge base has a corresponding title, and the title can be expressed as a question corresponding to the knowledge.

[0068] After obtaining the first and second pieces of knowledge, the knowledge recommendation system uses the first piece of knowledge as the answer to the problem to be solved and the title of the second piece of knowledge as a related question to the problem to be solved. The knowledge recommendation system then recommends the answer to the problem to be solved and the related questions to the terminal device. After receiving the answer to the problem to be solved and the related questions, the terminal device displays the answer to the problem to be solved and the related questions.

[0069] like Figure 2 In the dialog box of associated knowledge shown in FIG, user X enters question 1 in the input interface of the dialog box and submits question 1 to the knowledge recommendation system, as shown in FIG. Figure 2 Question 1 under user X on the right side of the display interface is displayed. Subsequently, after the terminal device receives the answer to question 1 and related questions of question 1, it displays the answer to question 1 and related questions of question 1 recommended by the knowledge recommendation system, such as Figure 2 The information displayed under the smart assistant on the left side of the display interface.

[0070] In the embodiment of the present disclosure, the knowledge recommendation system can sort multiple second knowledge in order of relevance scores from large to small. In this way, the knowledge recommendation system can form a recommendation list of related questions according to the titles of the sorted second knowledge, and recommend the recommendation list of related questions to the terminal device, so that the user can obtain the knowledge of interest from the recommendation list and improve the efficiency of the user in solving problems.

[0071] In the technical solution provided by the embodiments of the present disclosure, after acquiring the first knowledge, the knowledge recommendation system can determine the relevance score of each other knowledge with the first knowledge based on the subject of each other knowledge in the preset knowledge base and the subject of the first knowledge. Compared with the traditional method of determining relevance scores based on keywords, this knowledge recommendation method based on determining relevance scores based on topics is more accurate, especially when there are words with similar meanings but different word forms in the other knowledge and the first knowledge. The method based on keywords to determine relevance scores cannot determine the relevance relationship between these words, while the knowledge recommendation method based on topics to determine relevance scores can avoid this situation.

[0072] Furthermore, after determining the relevance score, the knowledge recommendation system can determine the second piece of knowledge associated with the first piece of knowledge based on the relevance score and recommend the title of the second piece of knowledge along with the first piece of knowledge to the terminal device, that is, to the user using the terminal device. This allows the user to solve the problem based on the first piece of knowledge. If the first piece of knowledge cannot solve the problem, the user can also view the title of the recommended second piece of knowledge, that is, view related questions of the problem. By viewing these related questions, the user can solve the problem, thereby improving the user's problem-solving efficiency.

[0073] In addition, the knowledge recommendation system can determine the second knowledge associated with the first knowledge based on the association score, and then determine the related issues. There is no need to manually configure the related knowledge, which reduces manpower consumption. Moreover, since there is no need to manually configure the related knowledge, even if the amount of knowledge stored in the knowledge base is large, the knowledge recommendation system can accurately compare the first knowledge with all the knowledge in the preset knowledge base, determine the second knowledge associated with the first knowledge, and then accurately recommend related issues, thereby improving the comprehensiveness of the recommended related issues.

[0074] In one embodiment of the present disclosure, a knowledge recommendation method is also provided. Figure 3 As shown, the method may include steps S31-S37, and steps S34-S37 are the same as the above steps S12-S15, which are not described in detail here. Steps S31-S33 are a possible implementation of step S11.

[0075] Step S31 , according to the preset weight of each question type, obtain questions of the question type from the question pool of the question type as hot questions.

[0076] In the embodiment of the present disclosure, the question type may include at least one of the following: a type of a first high-frequency question, a type of a second high-frequency question, and a type of a preset question.

[0077] The first high-frequency question is: the first preset number of knowledge titles with the highest number of accesses within a preset duration. The preset duration can be set based on actual needs, such as 12 hours, 1 day, 3 days, or 5 days, and there is no limit on this. The first preset number can also be set based on actual needs and there is no limit on this. The number of accesses to a piece of knowledge is the total number of times the knowledge recommendation system recommends that piece of knowledge to each terminal device. That is, each time the knowledge recommendation system recommends a piece of knowledge to a terminal device, the number of accesses to that piece of knowledge increases by 1.

[0078] The second most frequently asked question is: among the knowledge associated with the knowledge recommended to the terminal device, the titles of the second preset number of knowledge items with the highest number of accesses within a preset time period. The second preset number can be set based on actual needs. The first preset number and the second preset number can be the same or different, and there is no limitation on this. Regarding the determination of the associated knowledge items, please refer to the description of steps S13-S14 above regarding determining the second knowledge item associated with the first knowledge item.

[0079] Pre-set questions may include but are not limited to questions configured by the administrator in the background based on released new products, new features and promotional activities.

[0080] In the embodiment of the present disclosure, question types may also include other types, and questions of different question types may be the same or different, which is not limited to this.

[0081] The knowledge recommendation system can pre-set a corresponding weight for each question type, i.e., a preset weight. The preset weight for each question type can be set based on actual needs. For example, the preset weight for question type 1 is 3, the preset weight for question type 2 is 3, and the preset weight for question type 3 is 4. There is no limit to this.

[0082] In the embodiment of the present disclosure, the knowledge recommendation system can randomly obtain questions of the question type from the question pool of the question type according to the preset weight of each question type as hot questions.

[0083] For example, the knowledge recommendation system sets up a question pool of question types 1-3, where the preset weight of question type 1 is 3, the preset weight of question type 2 is 3, and the preset weight of question type 3 is 4. A total of 10 popular questions need to be obtained. The knowledge recommendation system then obtains 3 questions of question type 1 from the question pool of question type 1 as popular questions; obtains 3 questions of question type 2 from the question pool of question type 2 as popular questions; and obtains 4 questions of question type 3 from the question pool of question type 3 as popular questions.

[0084] Step S32: recommend the acquired multiple hot questions to the terminal device, so that the terminal device can determine a target hot question from the multiple hot questions.

[0085] The knowledge recommendation system can recommend multiple hot questions to the terminal device. After receiving the multiple hot questions, the terminal device can randomly select a hot question from the multiple hot questions as the target hot question.

[0086] After receiving multiple hot questions, the terminal device can also output these multiple hot questions. Figure 4 The dialog box for popular questions is shown. Figure 4 10 popular questions are shown in the figure. The number of popular questions in the figure is only for example and is not limited.

[0087] The user can select a hot question from the hot questions displayed on the terminal device; after the terminal device detects the user's operation of selecting a hot question, it determines that the hot question selected by the user is the target hot question.

[0088] Step S33: receiving the target hot issues determined by the terminal device as issues to be resolved.

[0089] After the knowledge recommendation system receives the target hot question determined by the terminal device, the knowledge recommendation system may use the target hot question as a problem to be solved submitted by the terminal device.

[0090] In the technical solutions provided by the embodiments of the present disclosure, the knowledge recommendation system can identify popular questions of each question type according to a preset weight. The system can then recommend these popular questions to a terminal device to identify the questions to be solved. This allows the knowledge recommendation system to recommend knowledge to users based on different question types, broadening their problem-solving strategies from different perspectives and improving their problem-solving efficiency.

[0091] In one embodiment of the present disclosure, a knowledge recommendation method is also provided. Figure 5As shown, the method may include steps S51-S59, and steps S56-S59 are the same as the above steps S12-S15, which are not described in detail here. Steps S51-S55 are a possible implementation of step S11.

[0092] Step S51: Acquire the candidate text submitted by the terminal device.

[0093] In the embodiment of the present disclosure, the candidate text may be one or more keywords or a complete problem to be solved, which is not limited. The user can submit the candidate text to the knowledge recommendation system through the terminal device, and the knowledge recommendation system obtains the candidate text submitted by the terminal device.

[0094] Step S52: Acquire multiple candidate titles that match the candidate text from the titles of knowledge included in the preset knowledge base.

[0095] The knowledge recommendation system can match and retrieve the titles of the knowledge included in the preset knowledge base with the candidate texts submitted by the terminal device, and obtain multiple candidate titles that match the candidate texts from the titles of the knowledge included in the preset knowledge base.

[0096] Step S53: Calculate the similarity score between each candidate title and the candidate text.

[0097] For each candidate title, the knowledge recommendation system may use a TF-IDF (Term Frequency-Inverse Document Frequency) algorithm or other algorithms for calculating text similarity to calculate a similarity score between the candidate title and the candidate text.

[0098] Step S54: recommending a third preset number of candidate titles with the highest similarity scores to the terminal device, so that the terminal device determines the target association question from the third preset number of candidate titles.

[0099] The knowledge recommendation system can select a third preset number of candidate titles with the highest similarity scores from the obtained multiple candidate titles, and recommend these candidate titles to the terminal device. The third preset number can be set according to actual needs, for example, the third preset number can be 5, 8, or 10, and there is no limitation on this.

[0100] After receiving multiple candidate titles, the terminal device can randomly determine a candidate title from a third preset number of candidate titles as the target association question.

[0101] After receiving multiple candidate titles, the terminal device can also output these multiple candidate titles. Figure 6 The dialog box for recommending candidate titles is shown. Figure 6In the input interface, after the candidate text is entered, 5 candidate titles are displayed above the input interface. The number of candidate titles in the figure is only an example and is not limited.

[0102] The user can select a candidate title from the candidate titles displayed by the terminal device; after the terminal device detects the user's operation of selecting the candidate title, it determines that the candidate title selected by the user is the target association title.

[0103] In the disclosed embodiment, the knowledge recommendation system sorts a third preset number of candidate titles in descending order of similarity scores and recommends the third preset number of candidate titles to the terminal device based on the sorting results. The higher the similarity score, the greater the probability that the user will be interested in the candidate title. The candidate titles are recommended to the terminal device in descending order of similarity scores, facilitating the user's access to knowledge of interest from the recommendation list and improving the user's problem-solving efficiency.

[0104] Step S55: receiving the target association problem determined by the terminal device as the problem to be solved.

[0105] After the knowledge recommendation system receives the target association question determined by the terminal device, the knowledge recommendation system may use the target association question as the problem to be solved submitted by the terminal device.

[0106] In the technical solution provided by the embodiments of the present disclosure, the knowledge recommendation system can obtain multiple candidate titles that match the candidate text submitted by a terminal device, and recommend a third preset number of candidate titles to the terminal device to determine the problem to be solved. In this way, the user can select a candidate title that matches the candidate text without submitting a complete question, and solve the problem by viewing the knowledge corresponding to the selected candidate title, thereby improving the user's problem-solving efficiency.

[0107] Furthermore, each time a user enters a keyword, the knowledge recommendation system matches a candidate title. As keywords are added or modified in the candidate text entered by the user, the candidate titles matched by the knowledge recommendation system will also change accordingly. This way, when the user is unclear about the problem they need to solve, the knowledge recommendation system can support the user entering a single keyword or changing the keyword search at any time to expand the user's problem-solving ideas and further improve the user's problem-solving efficiency. When the user has a clear question direction for the problem to be solved, the knowledge recommendation system can also support the user entering multiple keywords or entering a complete question so that the knowledge recommendation system can provide more accurate feedback to the user on the answer to the problem.

[0108] In one embodiment of the present disclosure, a knowledge recommendation method is also provided. Figure 7As shown, the method may include steps S71-S710, and steps S71 and steps S74-S710 are the same as steps S51 and steps S53-S59 above, and are not described in detail here. Steps S72-S73 are a possible implementation of step S52.

[0109] Step S72: segment the candidate text to obtain multiple candidate segmentations.

[0110] In the disclosed embodiment, the knowledge recommendation system may segment the candidate text to obtain a plurality of candidate segmentations.

[0111] If the preset knowledge base is stored in ES, the knowledge recommendation system can use Chinese word segmenters such as ik_smart (coarsest strength splitting) or ik_max_word (finest strength splitting) to segment the candidate text and obtain multiple candidate word segments. The knowledge recommendation system can also use other methods to segment the candidate text, which are not limited to this method.

[0112] Step S73: Obtain titles including candidate participles from titles of knowledge included in the preset knowledge base as candidate titles.

[0113] In the embodiment of the present disclosure, the knowledge recommendation system may query the titles of the knowledge included in the preset knowledge base, obtain the titles including the candidate segmentation words, and use the titles as candidate titles.

[0114] In the embodiment of the present disclosure, there may be one or more candidate participles. If there are multiple candidate participles, a title including one candidate participle can be used as a candidate title.

[0115] In the technical solution provided by the embodiments of the present disclosure, the knowledge recommendation system can segment candidate texts and obtain titles containing the candidate segmentations as candidate titles. Since a candidate text may be a complete problem to be solved, segmenting the candidate texts can increase the flexibility of the obtained candidate titles, thereby increasing the flexibility of knowledge recommendation.

[0116] In one embodiment of the present disclosure, a knowledge recommendation method is also provided. Figure 8 As shown, the method may include steps S81-S811, steps S81-S83 and steps S86-S811 are the same as steps S71-S73 and steps S75-S710 above, and are not repeated here. Steps S84-S85 are a possible implementation of step S74.

[0117] Step S84, calculate the ratio of the first coefficient of each candidate participle associated with each candidate title to the second coefficient of the candidate participle, the first coefficient of each candidate participle is the number of times the candidate participle appears in the candidate title, and the second coefficient of each candidate participle is the number of candidate titles that include the candidate participle.

[0118] In an embodiment of the present disclosure, after obtaining the candidate titles, for each candidate participle, the knowledge recommendation system can determine the number of occurrences of the candidate participle in the candidate title as the first coefficient of the candidate participle associated with the candidate title, determine the number of candidate titles that include the candidate participle as the second coefficient of the candidate participle associated with the candidate title; and calculate the ratio of the first coefficient of the candidate participle associated with the candidate title to the second coefficient of the candidate participle.

[0119] This ratio can also be regarded as: using the TF-IDF method to calculate the similarity score between the candidate title and the candidate text, and the resulting TF-IDF value.

[0120] Step S85: For each candidate title, the sum of the ratios of all candidate word segments associated with the candidate title is used as the similarity score between the candidate title and the candidate text.

[0121] In the disclosed embodiment, for each candidate title, the knowledge recommendation system may sum the ratios (i.e., TF-IDF values) of all candidate word segments associated with the candidate title, and use the sum as the similarity score between the candidate title and the candidate text.

[0122] In the technical solution provided by the embodiments of this disclosure, for each candidate title, the knowledge recommendation system can obtain the sum of the ratios of all candidate word segments associated with the candidate title as the similarity score between the candidate title and the candidate text. In this way, the knowledge recommendation system can determine the association between each candidate title and the candidate word segment, and thus determine the similarity score between the candidate title and the candidate text. This can increase the probability that the candidate title is a question of interest to the user and improve the user's efficiency in solving problems.

[0123] In one embodiment of the present disclosure, a knowledge recommendation method is also provided. Figure 9 As shown, in this method, the knowledge recommendation system has multiple query algorithms for obtaining candidate titles, and the query algorithms may include but are not limited to match query and match_phrase query.

[0124] Optionally, when the knowledge recommendation system queries the titles of the knowledge included in the preset knowledge base, it can also add the professional terms or product names contained in the input text to the ES dictionary, and then query the titles of the knowledge included in the preset knowledge base to obtain the titles including the candidate segmentation words as candidate titles. In this way, the accuracy of obtaining candidate titles can be improved, thereby improving the accuracy of the knowledge recommendation method.

[0125] Figure 9 The knowledge recommendation method shown may include steps S91-S910. Steps S91, S92, and S95-S910 are the same as steps S51, S52, and S54-S59 described above, and are not described in detail here. Steps S93-S94 are a possible implementation of step S53.

[0126] Step S93: For each query algorithm, calculate the initial similarity score between each candidate title and candidate text obtained by using the query algorithm.

[0127] For each query algorithm, the knowledge recommendation system can calculate the initial similarity score corresponding to the query algorithm, which can be seen in the above Figure 8 The calculation of the similarity score is described in detail here and will not be repeated here.

[0128] Step S94: For each candidate title, the sum of the multiple initial similarity scores of the candidate title is calculated as the similarity score between the candidate title and the candidate text.

[0129] For each candidate title, the knowledge recommendation system may calculate the sum of multiple initial similarity scores of the candidate title as the similarity score between the candidate title and the candidate text.

[0130] In the technical solution provided by the embodiment of the present disclosure, the knowledge recommendation system can adopt multiple query algorithms to calculate similarity scores and obtain candidate titles that need to be recommended to users. This can make up for the shortcomings of different query algorithms and improve the flexibility and accuracy of recommending candidate titles.

[0131] In one embodiment of the present disclosure, a knowledge recommendation method is also provided. Figure 10 As shown, the method may include steps S101-S108, and steps S105-S108 are the same as the above steps S12-S15, and are not described in detail here. Steps S101-S104 are a possible implementation of step S11.

[0132] Step S101, recommend multiple hot directories to the terminal device so that the terminal device displays the hot directories in a navigation form and determines a target directory from the hot directories. The hot directory is a directory whose access times and values ​​of multiple knowledge items are higher than a preset access threshold.

[0133] The knowledge recommendation system pre-sets an access threshold, also known as a preset access threshold. The knowledge in the preset knowledge base is organized into multiple categories, each containing multiple pieces of knowledge. The knowledge recommendation system counts the number of visits and values ​​for each piece of knowledge in each category within the preset knowledge base. It then identifies categories with access counts and values ​​exceeding the preset access threshold as popular categories and recommends them to the terminal device.

[0134] After receiving the hot directory, the terminal device displays the hot directory in the form of navigation, such as Figure 4 The dialog box shown, Figure 4 In the figure, multiple popular directories are displayed at the bottom of the display interface, such as directories 1-5, etc., where the number of candidate titles in the figure is only an example and is not limited. Figure 4 In the example, an update button is set on the right side of the navigation directories 1-5. When the currently displayed popular directories do not contain the directory that the user is interested in, the user can click the update button to update the popular directories in the navigation.

[0135] The user can select a hot directory from the hot directories displayed on the terminal device; after the terminal device detects the user's operation of selecting a hot directory, it determines that the hot directory selected by the user is the target directory.

[0136] In order to improve the efficiency of knowledge recommendation, the knowledge recommendation system can sort multiple popular directories in descending order according to the number of visits and values ​​corresponding to the popular directories, and recommend the sorting results to the terminal device, so that users can obtain the directories they are interested in from the sorting results, thereby improving the efficiency of knowledge recommendation.

[0137] Step S102: receiving a target directory submitted by a terminal device.

[0138] After the terminal device determines the target directory, it submits the target directory to the knowledge recommendation system.

[0139] Step S103: recommending a fourth preset number of knowledge titles with the highest number of access times under the target directory to the terminal device, so that the terminal device can determine the target directory problem from the titles under the target directory.

[0140] In the embodiment of the present disclosure, the fourth preset number can be set according to actual needs. For example, the fourth preset number can be 5, 8 or 10, and there is no limitation on this.

[0141] The knowledge recommendation system can determine the fourth preset number of knowledge with the highest number of accesses from the multiple knowledge in the target directory, and recommend the titles of the fourth preset number of knowledge to the terminal device. After receiving the titles of the fourth preset number of knowledge, the terminal device can display the titles of the fourth preset number of knowledge. Figure 11In the dialog box of the directory question shown, after the terminal device submits directory 1, the titles of multiple knowledge under directory 1 are respectively displayed in area 1 and area 2 in the display interface. In the embodiment of the present disclosure, the terminal device can also use other methods to display the titles of the fourth preset number of knowledge, such as only in Figure 11 The titles of the fourth preset number of knowledge items are displayed in area 1 or area 2. Figure 11 In the navigation, an update button is set on the right side of the directory questions 7-10. When the currently displayed directory questions do not contain questions that the user is interested in, the user can click the update button to update the directory questions of the navigation.

[0142] The user can select a hot directory from the hot directories displayed on the terminal device; after the terminal device detects the user's operation of selecting a hot directory, it determines that the hot directory selected by the user is the target directory.

[0143] In order to improve the efficiency of users in solving problems, the knowledge recommendation system can sort multiple popular directories in descending order according to the number of visits and values ​​corresponding to the popular directories, and recommend the sorting results to the terminal device, so that users can obtain the questions they are interested in from the sorting results, thereby improving the efficiency of knowledge recommendation.

[0144] Optionally, if the terminal device includes multiple areas for displaying questions, the terminal device can configure the priority of each area. The higher the priority area, the greater the number of visits and values ​​of the popular directory corresponding to the displayed question. The questions displayed in each area are sorted from large to small according to the number of visits to the knowledge. In this way, the efficiency of users solving problems can be further improved. For example, Figure 11 As shown, both area 1 and area 2 are used to display questions, wherein the priority of area 1 is higher than that of area 2, that is, the number of visits to the catalog questions 1-6 displayed in area 1 is higher than the number of visits to the catalog questions 7-9 displayed in area 2.

[0145] Step S104: receiving the target directory problem determined by the terminal device as a problem to be solved.

[0146] The knowledge recommendation system can receive the target directory problem determined by the terminal device and use the target directory problem as a problem to be solved.

[0147] In the technical solution provided by the embodiment of the present disclosure, the knowledge recommendation system can recommend questions to users based on popular directories and multiple pieces of knowledge with the highest number of visits under the popular directories, thereby improving the efficiency of determining problems to be solved and thereby improving the efficiency of users in solving problems.

[0148] In one embodiment of the present disclosure, a knowledge recommendation method is also provided. Figure 12As shown, the method may include steps S121-S128, where steps S121 and S126-S128 are the same as the above-mentioned steps S11 and S13-S15, and are not described in detail here. Steps S122-S125 are a possible implementation of step S12.

[0149] Step S122: segment the problem to be solved to obtain multiple target segmentations.

[0150] In the embodiment of the present disclosure, the knowledge recommendation system can segment the problem to be solved to obtain multiple target segmentations. The knowledge recommendation system can use Chinese segmentation tools such as ik_smart or ik_max_word for segmentation. The knowledge recommendation system can also use other methods for segmentation, which is not limited to this.

[0151] In an optional embodiment, the knowledge recommendation system may segment the problem to be solved to obtain multiple initial segmentations. The knowledge recommendation system may then preprocess the multiple initial segmentations to obtain multiple target segmentations. The preprocessing may include at least one of the following operations: removing stop words, removing filter words, correcting spelling errors, and adding synonyms for the initial segmentations.

[0152] In the technical solution provided by the embodiment of the present disclosure, the knowledge recommendation system performs word segmentation and data preprocessing on the problem to be solved, which can make the knowledge recommendation system more accurate in determining the association score between each other knowledge and the first knowledge, thereby improving the accuracy of knowledge recommendation.

[0153] Step S123: convert the multiple target word segments into target feature vectors.

[0154] In the embodiment of the present disclosure, the method for converting multiple target word segments into target feature vectors by the knowledge recommendation system includes but is not limited to tf-idf, word2vec, and seq2seq.

[0155] Step S124 , calculating the similarity between the target feature vector and the feature vector of each knowledge title in the preset knowledge base.

[0156] The knowledge recommendation system can use algorithms such as Euclidean distance or cosine distance to calculate the similarity between the target feature vector and the feature vector of each knowledge title in the preset knowledge base. Here, the smaller the distance, the more similar it is.

[0157] Step S125: The knowledge corresponding to the title with the highest similarity is taken as the first knowledge.

[0158] The knowledge recommendation system can sort the titles in descending order of similarity, and take the knowledge corresponding to the title with the highest similarity as the first knowledge.

[0159] In the technical solution provided by the disclosed embodiments, the knowledge recommendation system can segment the problem to be solved, convert multiple target segmentations into target feature vectors, and select the knowledge corresponding to the title with the highest similarity as the first knowledge. In this way, users can determine the similarity based on the feature vectors and provide the best answer, improving the accuracy and efficiency of knowledge recommendations.

[0160] In one embodiment of the present disclosure, a knowledge recommendation method is also provided. Figure 13 As shown, the method may include steps S131-S139. Steps S131-S138 are the same as the above steps S121-S128 and are not repeated here.

[0161] Step S139: recommending other titles except the title of the first knowledge among the fifth preset number of titles with the highest similarity to the terminal device, so that the terminal device displays other titles in a navigation form.

[0162] In the embodiment of the present disclosure, the fifth preset number can be set according to actual needs. For example, the fifth preset number can be 5, 8 or 10, and there is no limitation on this.

[0163] The knowledge recommendation system can determine the fifth preset number of knowledge with the highest similarity from the preset knowledge base, remove the title of the first knowledge in the fifth preset number of titles, obtain other titles, and recommend the other titles to the terminal device. The terminal device displays the other titles in the form of navigation. Figure 2 As shown, similar questions 1-4 are displayed at the bottom of the display interface. In the figure, the number of other titles displayed is only for example and not limiting. Figure 2 In the navigation, an update button is set on the right side of similar questions 1-4. When the currently displayed similar questions do not contain the questions that the user is interested in, the user can click the update button to update the similar questions in the navigation.

[0164] In order to improve the efficiency of users in solving problems, the knowledge recommendation system can sort multiple titles in descending order according to the number of visits to the knowledge corresponding to the titles, and recommend the sorting results to the terminal device, so that users can obtain the questions they are interested in from the sorting results, thereby improving the efficiency of knowledge recommendation.

[0165] In the embodiment of the present disclosure, for each problem to be solved, the knowledge recommendation system needs to calculate the similarity between each problem to be solved and the feature vector and the feature vectors of all knowledge titles in the preset knowledge base, and the time complexity is O(n), where n is the number of knowledge in the preset knowledge base. In step S149, the fifth preset number of titles with the highest similarity can be obtained with the help of a big root heap, and its time complexity is O(n*logK), where n is the number of knowledge in the preset knowledge base, and K is the fifth preset number. It can be seen that the time complexity of the entire process is O(n*logK), which is very large. In order to quickly obtain the fifth preset number of titles, the knowledge recommendation system can use SimHash to reduce the dimensionality of the above-mentioned feature vector, and use the index to obtain the fifth preset number of titles with the highest similarity, so as to reduce the time complexity of the calculation.

[0166] In the technical solution provided by the disclosed embodiments, the knowledge recommendation system recommends the remaining titles, excluding the first knowledge title, from a fifth predetermined number of titles with the highest similarity to the terminal device, which then displays them. In the navigation knowledge recommendation scenario, users can use the navigation knowledge recommendation window to solve pending problems, improving their problem-solving efficiency.

[0167] In one embodiment of the present disclosure, a knowledge recommendation method is also provided. Figure 14 As shown, the method may include steps S141-S147, steps S141-S142, steps S146-S147 are the same as the above steps S11-S12, steps S14-step S15, and are not repeated here. Steps S143-S145 are a possible implementation of step S13.

[0168] Step S143 , mapping the subject of the first knowledge to a preset multi-dimensional feature space to obtain a first feature vector of the first knowledge.

[0169] In the embodiment of the present disclosure, after determining the first knowledge, the knowledge recommendation system may determine the association score between the first knowledge and other knowledge based on the similarity of the topic feature vector, which may specifically include steps S143-S145.

[0170] The knowledge recommendation system pre-defines a multidimensional feature space, such as an n-dimensional feature space, in which each dimension corresponds to a topic. A piece of knowledge has one or more topics. The knowledge recommendation system maps all topics of a piece of knowledge to the pre-determined multidimensional feature space, obtaining a first feature vector for the first piece of knowledge. The mapping rules can be customized based on actual needs.

[0171] In the embodiment of the present disclosure, the topic in the knowledge recommendation system may be a classification directory set based on business experience. The knowledge recommendation system may also determine the topic based on indicators such as perplexity or MPI-score.

[0172] Step S144 maps the subject of each other knowledge in the preset knowledge base to a preset multidimensional feature space to obtain a second feature vector of each other knowledge. Similar to step S143, please refer to the relevant description of step S143 above for details.

[0173] In the embodiment of the present disclosure, the execution order of steps S143 and S144 is not limited.

[0174] Step S145 , calculating the cosine similarity between the first eigenvector and each second eigenvector, and obtaining the association score between the first knowledge and each other knowledge.

[0175] In the embodiment of the present disclosure, the knowledge recommendation system may use the following formula to determine the association score between the first knowledge and each other knowledge:

[0176]

[0177] Among them, score(A,B) represents the correlation score between knowledge A and knowledge B, n is the dimension of the preset dimensional feature space, A i Represents the eigenvalue of dimension i in the eigenvector of knowledge A, B i Represents the eigenvalue of dimension i in the eigenvector of knowledge B.

[0178] In the technical solution provided by the embodiments of the present disclosure, the knowledge recommendation system determines the cosine similarity based on the subject of the knowledge, and then determines the association score to obtain the second knowledge associated with the first knowledge. In this process, there is no need to manually configure the related knowledge, which reduces manpower consumption. Moreover, since there is no need to manually configure the related knowledge, even if the amount of knowledge stored in the knowledge base is large, the knowledge recommendation system can accurately compare the first knowledge with all the knowledge in the preset knowledge base, determine the second knowledge associated with the first knowledge, and then accurately recommend related issues, thereby improving the comprehensiveness of the recommended related issues.

[0179] In one embodiment of the present disclosure, a knowledge recommendation method is also provided. Figure 15 As shown, the method may include steps S151-S155, where steps S151, S152, S154 and S155 are the same as the above-mentioned steps S11, S12, S14 and S15, and are not described in detail here. Step S153 is a possible implementation of step S13.

[0180] Step S153: determining the number of common topics between the first knowledge and each other knowledge in the preset knowledge base, and using this as the association score between the first knowledge and the other knowledge.

[0181] In the embodiment of the present disclosure, after determining the first knowledge, the knowledge recommendation system may perform link prediction based on local structural similarity to determine the association score between the first knowledge and other knowledge, which may specifically include step S153.

[0182] In step S153, the knowledge recommendation system counts the topics included in the first knowledge and the topics included in each other knowledge in the preset knowledge base; for each other knowledge, the knowledge recommendation system compares the topics included in the first knowledge and the topics included in the other knowledge, obtains the common topics of the first knowledge and the other knowledge, and then obtains the number of common topics of the first knowledge and the other knowledge. The number of common topics can be used as the correlation score between the first knowledge and the other knowledge.

[0183] In an optional embodiment, the knowledge recommendation system can use knowledge or topics as nodes to construct a heterogeneous graph. In this heterogeneous graph, there are no edges between knowledge nodes, but there are edges between knowledge nodes and topic nodes. For example, if knowledge Q includes topic T, there is an edge between the knowledge Q node and the topic T node; the two nodes connected by the edge are neighbor nodes. Figure 16 The heterogeneous graph shown includes knowledge 1-4 nodes and topic 1-4 nodes, wherein there is no connection between the knowledge 1-4 nodes, and there is a connection between the knowledge 1-4 nodes and the topic 1-4 nodes.

[0184] The knowledge recommendation system is based on a heterogeneous graph to determine the neighbor nodes (i.e., neighbor topic nodes) of the first knowledge node and the neighbor nodes of each other knowledge node. For each other knowledge node, the knowledge recommendation system can determine the same neighbor nodes that the first knowledge node and the other knowledge node have. The number of the same neighbor nodes is the number of the same topics that the first knowledge and the other knowledge have, that is, the association score between the first knowledge and the other knowledge.

[0185] The above-mentioned knowledge recommendation system can use the following formula to determine the number of common neighbor nodes of two knowledge nodes.

[0186] score(A,B)=|N(A)∩N(B)|.

[0187] Among them, score(A,B) represents the number of common neighbor nodes of knowledge node A and knowledge node B, that is, the association score between knowledge node A and knowledge node B, N(A) represents the set of neighbor nodes of knowledge node A, and N(B) represents the set of neighbor nodes of knowledge node B.

[0188] In the technical solution provided by the embodiments of this disclosure, based on the principle of heterogeneous graphs, the number of common topics shared by the first knowledge and other knowledge is determined as the association score between the first knowledge and the other knowledge. The greater the number of common topics, the greater the probability that an edge exists between the first knowledge and the other knowledge, and the higher the association score between the first knowledge and the other knowledge. Compared with methods that determine association scores based on complex cosine similarity, this method of determining association scores based on the principle of heterogeneous graphs improves the efficiency of association scoring, thereby improving the efficiency of knowledge recommendation.

[0189] In one embodiment of the present disclosure, a knowledge recommendation method is also provided. Figure 17 As shown, the method may include steps S171-S176, where steps S171, S172, S175 and S176 are the same as steps S11, S12, S14 and S15 above, and are not described in detail here. Steps S173-S174 are a possible implementation of step S13.

[0190] Step S173: Determine the multiple common themes between the first knowledge and each other knowledge in the preset knowledge base. The method for determining the common themes can be found in the relevant description of step S163 above and will not be repeated here.

[0191] Step S174, calculate the sum of the influence coefficients of multiple identical topics as the association score between the first knowledge and each other knowledge; wherein the influence coefficient of the identical topic is: the reciprocal of the number of knowledge with the identical topic, or the reciprocal of the logarithm of the number of knowledge with the identical topic.

[0192] For each other knowledge, for each common topic between the first knowledge and the other knowledge, the knowledge recommendation system can calculate the inverse of the number of knowledge with the common topic as the influence coefficient of the common topic corresponding to the other knowledge; the knowledge recommendation system calculates the sum of the influence coefficients of the common topic corresponding to the other knowledge as the association score between the first knowledge and the other knowledge.

[0193] The above-mentioned knowledge recommendation system can use the following formula to calculate the sum of the influence coefficients of other knowledge items with the same topic.

[0194]

[0195] Among them, score(A,B) represents the influence coefficient and value of the same topic shared by knowledge node A and knowledge node B, that is, the association score between knowledge node A and knowledge node B, N(A) represents the set of neighbor nodes of knowledge node A, N(B) represents the set of neighbor nodes of knowledge node B, z represents the same topic nodes shared by knowledge node A and knowledge node B, k(z) represents the number of knowledge nodes with topic z nodes, and in a heterogeneous graph, k(z) can also be called the degree of topic z nodes.

[0196] For each other knowledge, for each common topic between the first knowledge and the other knowledge, the knowledge recommendation system can calculate the inverse of the logarithm of the number of knowledge with the common topic as the influence coefficient of the common topic corresponding to the other knowledge; the knowledge recommendation system calculates the sum of the influence coefficients of the common topic corresponding to the other knowledge as the association score between the first knowledge and the other knowledge.

[0197] The above-mentioned knowledge recommendation system can use the following formula to calculate the sum of the influence coefficients of other knowledge items with the same topic.

[0198]

[0199] Among them, score(A,B) represents the influence coefficient and value of the same topic shared by knowledge node A and knowledge node B, that is, the association score between knowledge node A and knowledge node B, N(A) represents the set of neighbor nodes of knowledge node A, N(B) represents the set of neighbor nodes of knowledge node B, z represents the same topic nodes shared by knowledge node A and knowledge node B, k(z) represents the number of knowledge with topic z nodes, and in a heterogeneous graph, k(z) can also be called the degree of topic z nodes.

[0200] In the technical solution provided by the embodiments of this disclosure, based on the principle of heterogeneous graphs, the number of common topics shared by the first knowledge and other knowledge is determined as the association score between the first knowledge and the other knowledge. The greater the number of common topics, the greater the probability that an edge exists between the first knowledge and the other knowledge, and the higher the association score between the first knowledge and the other knowledge. Compared with methods that determine association scores based on complex cosine similarity, this method of determining association scores based on the principle of heterogeneous graphs improves the efficiency of association scoring, thereby improving the efficiency of knowledge recommendation.

[0201] In addition, in the disclosed embodiments, neighbor nodes are used as a medium for transferring resources, and the influence of degree on node association is taken into account. Neighbor nodes with smaller degrees have a greater impact on the association between knowledge nodes, while neighbor nodes with larger degrees have a smaller impact on the association between knowledge nodes. This effectively improves the accuracy and rationality of association scoring, and thus the accuracy and rationality of knowledge recommendations.

[0202] The technical solution provided by the embodiment of the present disclosure is applicable to various scenarios, such as popular knowledge recommendation scenarios (such as Figure 3 As shown), input the recommended scenario (such as Figure 5 、 Figure 7-Figure 9 As shown in ), navigation knowledge recommendation scenario (such as Figure 12-13 As shown in ), related knowledge recommendation scenarios (such as Figure 14 、 Figure 15 、 Figure 17 This greatly enriches the application scenarios of knowledge recommendation.

[0203] Corresponding to the above-mentioned knowledge recommendation method, the embodiment of the present disclosure also provides a knowledge recommendation device, such as Figure 18 As shown, this may include:

[0204] The first acquiring unit 181 is configured to acquire an unresolved issue submitted by a terminal device;

[0205] The second acquisition unit 182 is used to acquire first knowledge for solving the problem to be solved from a preset knowledge base;

[0206] A first determining unit 183 is configured to determine, based on the subject of the first knowledge and the subjects of other knowledge in the preset knowledge base, an association score between each other knowledge and the first knowledge;

[0207] A second determining unit 184 is configured to determine, from other knowledge, second knowledge associated with the first knowledge based on the association score;

[0208] The recommendation unit 185 is configured to recommend the title of the second knowledge and the first knowledge to the terminal device, wherein the first knowledge is the answer to the problem to be solved and the title of the second knowledge is a related question of the problem to be solved.

[0209] Optionally, the first acquiring unit 181 may be specifically configured to:

[0210] According to the preset weight of each question type, questions of this question type are obtained from the question pool of this question type as hot questions;

[0211] Recommending the acquired multiple hot questions to the terminal device, so that the terminal device determines a target hot question from the multiple hot questions;

[0212] Receive the target hot issues determined by the terminal device as issues to be solved.

[0213] Optionally, the question type includes at least one of the following: a type of the first high-frequency question, a type of the second high-frequency question, and a type of a preset question;

[0214] The first high-frequency question is: the titles of the first preset number of pieces of knowledge with the highest number of visits within a preset time period;

[0215] The second most frequently asked question is: among the knowledge associated with the knowledge recommended to the terminal device, the titles of the second preset number of pieces of knowledge with the highest number of visits within a preset time period.

[0216] Optionally, the first acquiring unit 181 may include:

[0217] A first acquiring subunit is used to acquire the candidate text submitted by the terminal device;

[0218] A second acquisition subunit is configured to acquire a plurality of candidate titles matching the candidate text from titles of knowledge included in a preset knowledge base;

[0219] A calculation subunit, used to calculate the similarity score between each candidate title and the candidate text;

[0220] A recommendation subunit, configured to recommend a third preset number of candidate titles with the highest similarity scores to the terminal device, so that the terminal device determines a target association question from the third preset number of candidate titles;

[0221] The receiving subunit is used to receive the target association problem determined by the terminal device as the problem to be solved.

[0222] Optionally, the second acquiring subunit may include:

[0223] Segment the candidate text to obtain multiple candidate segmentations;

[0224] From the titles of knowledge included in a preset knowledge base, titles including candidate word segments are obtained as candidate titles.

[0225] Optional computing subunit, specifically used for:

[0226] Calculate the ratio of the first coefficient of each candidate participle associated with each candidate title to the second coefficient of the candidate participle, where the first coefficient of each candidate participle is the number of occurrences of the candidate participle in the candidate title, and the second coefficient of each candidate participle is the number of candidate titles that include the candidate participle;

[0227] For each candidate title, the sum of the ratios of all candidate word segments associated with the candidate title is used as the similarity score between the candidate title and the candidate text.

[0228] Optionally, there are multiple query algorithms for obtaining candidate titles, and the calculation subunit can be used to:

[0229] For each query algorithm, calculate the initial similarity score between each candidate title and candidate text obtained by using the query algorithm;

[0230] For each candidate title, the sum of multiple initial similarity scores of the candidate title is calculated as the similarity score between the candidate title and the candidate text.

[0231] Optionally, the first acquiring unit 181 may be specifically configured to:

[0232] Recommending multiple hot directories to the terminal device so that the terminal device displays the hot directories in a navigational form and determines a target directory from the hot directories, wherein the hot directory is a directory including multiple pieces of knowledge with access times and values ​​higher than a preset access threshold;

[0233] Receive the target directory submitted by the terminal device;

[0234] recommending to the terminal device a fourth preset number of knowledge titles with the highest number of access times under the target directory, so that the terminal device determines the target directory problem from the titles under the target directory;

[0235] The target directory problem determined by the terminal device is received as a problem to be solved.

[0236] Optionally, the second obtaining unit 182 may include:

[0237] The word segmentation sub-unit is used to segment the problem to be solved and obtain multiple target word segments;

[0238] A conversion subunit, used to convert multiple target word segmentations into target feature vectors;

[0239] A calculation subunit, used to calculate the similarity between the target feature vector and the feature vector of each knowledge title in the preset knowledge base;

[0240] A subunit is determined, which is used to take the knowledge corresponding to the title with the highest similarity as the first knowledge.

[0241] Optional, word segmentation subunit, specifically can be used for:

[0242] Perform word segmentation on the problem to be solved to obtain multiple initial word segmentations;

[0243] Preprocess multiple initial segmented words to obtain multiple target segmented words;

[0244] Preprocessing includes at least one of the following operations: removing stop words, removing filter words, correcting spelling errors, and adding synonyms of initial segmentation words.

[0245] Optionally, the knowledge recommendation device may further include:

[0246] The other titles except the title of the first knowledge among the fifth preset number of titles with the highest similarity are recommended to the terminal device, so that the terminal device displays the other titles in a navigation form.

[0247] Optionally, the first determining unit 183 may be specifically configured to:

[0248] Mapping the subject of the first knowledge to a preset multidimensional feature space to obtain a first feature vector of the first knowledge;

[0249] Mapping the subject of each other knowledge in the preset knowledge base to a preset multi-dimensional feature space to obtain a second feature vector of each other knowledge;

[0250] The cosine similarity between the first eigenvector and each second eigenvector is calculated to obtain the association score between the first knowledge and each other knowledge.

[0251] Optionally, the first determining unit 183 may be specifically configured to:

[0252] The number of common topics between the first knowledge and each other knowledge in the preset knowledge base is determined as the association score between the first knowledge and the other knowledge.

[0253] Optionally, the first determining unit 183 may be specifically configured to:

[0254] determining a plurality of common themes shared by the first knowledge and each other knowledge in a preset knowledge base;

[0255] Calculate the sum of the influence coefficients of multiple identical topics as the correlation score between the first knowledge and the other knowledge;

[0256] The influence coefficient of the same topic is: the reciprocal of the number of knowledge with the same topic, or the reciprocal of the logarithm of the number of knowledge with the same topic.

[0257] Optionally, the second determining unit 184 may be specifically configured to:

[0258] From the other knowledge, a sixth preset number of other knowledge with the highest association scores are selected as the second knowledge associated with the first knowledge.

[0259] In the technical solution provided by the embodiments of the present disclosure, after acquiring the first knowledge, the knowledge recommendation system can determine the relevance score of each other knowledge with the first knowledge based on the subject of each other knowledge in the preset knowledge base and the subject of the first knowledge. Compared with the traditional method of determining relevance scores based on keywords, this knowledge recommendation method based on determining relevance scores based on topics is more accurate, especially when there are words with similar meanings but different word forms in the other knowledge and the first knowledge. The method based on keywords to determine relevance scores cannot determine the relevance relationship between these words, while the knowledge recommendation method based on topics to determine relevance scores can avoid this situation.

[0260] Furthermore, after determining the relevance score, the knowledge recommendation system can determine the second piece of knowledge associated with the first piece of knowledge based on the relevance score and recommend the title of the second piece of knowledge along with the first piece of knowledge to the terminal device, that is, to the user using the terminal device. This allows the user to solve the problem based on the first piece of knowledge. If the first piece of knowledge cannot solve the problem, the user can also view the title of the recommended second piece of knowledge, that is, view related questions of the problem. By viewing these related questions, the user can solve the problem, thereby improving the user's problem-solving efficiency.

[0261] In addition, the knowledge recommendation system can determine the second knowledge associated with the first knowledge based on the association score, and then determine the related issues. There is no need to manually configure the related knowledge, which reduces manpower consumption. Moreover, since there is no need to manually configure the related knowledge, even if the amount of knowledge stored in the knowledge base is large, the knowledge recommendation system can accurately compare the first knowledge with all the knowledge in the preset knowledge base, determine the second knowledge associated with the first knowledge, and then accurately recommend related issues, thereby improving the comprehensiveness of the recommended related issues.

[0262] In the technical solutions disclosed herein, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0263] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0264] Figure 19 A block diagram of a knowledge recommendation system 1900 for implementing the knowledge recommendation method of an embodiment of the present disclosure is shown. An electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. An electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0265] like Figure 19As shown, device 1900 includes a computing unit 1901, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1902 or a computer program loaded from a storage unit 1908 into a random access memory (RAM) 1903. Various programs and data required for the operation of device 1900 can also be stored in RAM 1903. Computing unit 1901, ROM 1902, and RAM 1903 are connected to each other via a bus 1904. An input / output (I / O) interface 1905 is also connected to bus 1904.

[0266] Various components in device 1900 are connected to I / O interface 1905, including: an input unit 1906, such as a keyboard, mouse, etc.; an output unit 1907, such as various types of displays, speakers, etc.; a storage unit 1908, such as a magnetic disk, optical disk, etc.; and a communication unit 1909, such as a network card, modem, wireless communication transceiver, etc. The communication unit 1909 allows device 1900 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0267] The computing unit 1901 can be a variety of general-purpose and / or specialized processing components with processing and computing capabilities. Some examples of the computing unit 1901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 1901 performs the various methods and processes described above, such as the knowledge recommendation method. For example, in some embodiments, the knowledge recommendation method can be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as the storage unit 1908. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 1900 via the ROM 1902 and / or the communication unit 1909. When the computer program is loaded into the RAM 1903 and executed by the computing unit 1901, one or more steps of the knowledge recommendation method described above can be performed. Alternatively, in other embodiments, the computing unit 1901 can be configured to perform the knowledge recommendation method by any other appropriate means (e.g., by means of firmware).

[0268] Figure 20 A block diagram of a knowledge recommendation system for implementing the knowledge recommendation method of an embodiment of the present disclosure includes:

[0269] at least one processor 2001; and

[0270] A memory 2002 in communication with at least one processor 2001; wherein,

[0271] The memory 2002 stores instructions that can be executed by the at least one processor 2001 . The instructions are executed by the at least one processor 2001 so that the at least one processor 2001 can perform any knowledge recommendation method.

[0272] The embodiments of the present disclosure also provide a non-transitory computer-readable storage medium storing computer instructions, where the computer instructions are used to enable a computer to execute any of the above-mentioned knowledge recommendation methods.

[0273] An embodiment of the present disclosure further provides a computer program product, including a computer program, which implements any of the above-mentioned knowledge recommendation methods when executed by a processor.

[0274] Various embodiments of the systems and techniques described above can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0275] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0276] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0277] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0278] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0279] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact through a communication network. The client-server relationship arises through computer programs running on the respective computers and having a client-server relationship with each other. The server may be a cloud server, a server in a distributed system, or a server integrated with a blockchain.

[0280] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not limited herein.

[0281] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A knowledge recommendation method, comprising: Obtain pending issues submitted by terminal devices; Acquire first knowledge for solving the problem to be solved from a preset knowledge base; determining a plurality of common themes between the first knowledge and each other knowledge in the preset knowledge base; Calculating the sum of the influence coefficients of the multiple same topics as the association score between the first knowledge and each other knowledge; wherein the influence coefficient of the same topic is: the reciprocal of the number of knowledge with the same topic, or the reciprocal of the logarithm of the number of knowledge with the same topic; determining, from the other knowledge, second knowledge associated with the first knowledge according to the association score; The title of the second knowledge and the first knowledge are recommended to the terminal device, wherein the first knowledge is the answer to the problem to be solved, and the title of the second knowledge is a related question of the problem to be solved.

2. The method according to claim 1, wherein The step of obtaining the unresolved problem submitted by the terminal device includes: According to the preset weight of each question type, questions of this question type are obtained from the question pool of this question type as hot questions; recommending the acquired multiple hot questions to a terminal device, so that the terminal device determines a target hot question from the multiple hot questions; The target hot issue determined by the terminal device is received as a problem to be solved.

3. The method according to claim 2, wherein: The question type includes at least one of the following: a type of a first high-frequency question, a type of a second high-frequency question, and a type of a preset question; The first high-frequency question is: the titles of the first preset number of pieces of knowledge with the highest number of access times within a preset time period; The second high-frequency question is: titles of a second preset number of pieces of knowledge with the highest number of access times within the preset time period among the knowledge associated with the knowledge recommended to the terminal device.

4. The method according to claim 1, wherein The step of obtaining the unresolved problem submitted by the terminal device includes: Obtain the candidate text submitted by the terminal device; Acquire multiple candidate titles matching the candidate text from titles of knowledge included in a preset knowledge base; Calculating a similarity score between each candidate title and the candidate text; recommending a third preset number of candidate titles with the highest similarity scores to a terminal device, so that the terminal device determines a target association question from the third preset number of candidate titles; The target association problem determined by the terminal device is received as a problem to be solved.

5. The method according to claim 4, wherein The step of obtaining a plurality of candidate titles matching the candidate text from titles of knowledge included in a preset knowledge base includes: Performing word segmentation on the candidate text to obtain multiple candidate word segmentations; From the titles of knowledge included in a preset knowledge base, titles including candidate word segments are obtained as candidate titles.

6. The method according to claim 5, wherein: The step of calculating the similarity score between each candidate title and the candidate text comprises: Calculate the ratio of the first coefficient of each candidate participle associated with each candidate title to the second coefficient of the candidate participle, where the first coefficient of each candidate participle is the number of occurrences of the candidate participle in the candidate title, and the second coefficient of each candidate participle is the number of candidate titles that include the candidate participle; For each candidate title, the sum of the ratios of all candidate word segments associated with the candidate title is used as the similarity score between the candidate title and the candidate text.

7. The method according to claim 4 or 5, wherein: There are multiple query algorithms for obtaining candidate titles, and the step of calculating the similarity score between each candidate title and the candidate text includes: For each query algorithm, calculating an initial similarity score between each candidate title obtained using the query algorithm and the candidate text; For each candidate title, a sum of multiple initial similarity scores of the candidate title is calculated as the similarity score between the candidate title and the candidate text.

8. The method according to claim 1, wherein The step of obtaining the unresolved problem submitted by the terminal device includes: Recommending a plurality of hot directories to a terminal device so that the terminal device displays the hot directories in a navigational form and determines a target directory from the hot directories, wherein the hot directories are directories including a plurality of knowledge items whose access times and values ​​are higher than a preset access threshold; receiving the target directory submitted by the terminal device; recommending to the terminal device a fourth preset number of knowledge titles with the highest number of access times under the target directory, so that the terminal device determines a target directory problem from the titles under the target directory; The target directory problem determined by the terminal device is received as a problem to be solved.

9. The method according to claim 1, wherein The step of acquiring first knowledge for solving the problem to be solved from a preset knowledge base includes: Segment the problem to be solved to obtain multiple target segmentations; Converting the multiple target word segments into target feature vectors; Calculating the similarity between the target feature vector and the feature vector of each knowledge title in the preset knowledge base; The knowledge corresponding to the title with the highest similarity is taken as the first knowledge.

10. The method according to claim 9, wherein: The step of segmenting the problem to be solved to obtain multiple target segmentations includes: Perform word segmentation on the problem to be solved to obtain multiple initial word segmentations; Preprocessing the multiple initial segmented words to obtain multiple target segmented words; The preprocessing includes at least one of the following operations: removing stop words, removing filter words, correcting spelling errors, and adding synonyms of initial segmented words.

11. The method according to claim 9, further comprising: The other titles except the title of the first knowledge among the fifth preset number of titles with the highest similarity are recommended to the terminal device, so that the terminal device displays the other titles in a navigation form.

12. The method according to claim 1, wherein The step of determining, from the other knowledge, second knowledge associated with the first knowledge based on the association score, includes: A sixth preset number of pieces of other knowledge with the highest association scores are selected from the other knowledge as the second knowledge associated with the first knowledge.

13. A knowledge recommendation device, comprising: A first obtaining unit is used to obtain the unresolved issues submitted by the terminal device; A second acquiring unit is configured to acquire first knowledge for solving the problem to be solved from a preset knowledge base; A first determining unit is configured to determine multiple common themes between the first knowledge and each other knowledge in the preset knowledge base; calculate the sum of the influence coefficients of the multiple common themes as the association score between the first knowledge and each other knowledge; wherein the influence coefficient of the common themes is: the reciprocal of the number of knowledge pieces having the common themes, or the reciprocal of the logarithm of the number of knowledge pieces having the common themes; a second determining unit, configured to determine, from the other knowledge, second knowledge associated with the first knowledge based on the association score; A recommendation unit is used to recommend the title of the second knowledge and the first knowledge to the terminal device, wherein the first knowledge is the answer to the problem to be solved, and the title of the second knowledge is a related question of the problem to be solved.

14. The device according to claim 13, wherein The first acquiring unit is specifically configured to: According to the preset weight of each question type, questions of this question type are obtained from the question pool of this question type as hot questions; recommending the acquired multiple hot questions to a terminal device, so that the terminal device determines a target hot question from the multiple hot questions; The target hot issue determined by the terminal device is received as a problem to be solved.

15. The device according to claim 14, wherein The question type includes at least one of the following: a type of a first high-frequency question, a type of a second high-frequency question, and a type of a preset question; The first high-frequency question is: the titles of the first preset number of pieces of knowledge with the highest number of access times within a preset time period; The second high-frequency question is: titles of a second preset number of pieces of knowledge with the highest number of access times within the preset time period among the knowledge associated with the knowledge recommended to the terminal device.

16. The device according to claim 13, wherein The first acquiring unit includes: A first acquiring subunit is used to acquire the candidate text submitted by the terminal device; A second acquisition subunit is configured to acquire a plurality of candidate titles matching the candidate text from titles of knowledge included in a preset knowledge base; A calculation subunit, configured to calculate a similarity score between each candidate title and the candidate text; A recommendation subunit, configured to recommend a third preset number of candidate titles with the highest similarity scores to a terminal device, so that the terminal device determines a target association question from the third preset number of candidate titles; The receiving subunit is used to receive the target association problem determined by the terminal device as a problem to be solved.

17. The device according to claim 16, wherein The second acquiring subunit includes: Performing word segmentation on the candidate text to obtain multiple candidate word segmentations; From the titles of knowledge included in a preset knowledge base, titles including candidate word segments are obtained as candidate titles.

18. The device according to claim 17, wherein The computing subunit is specifically configured to: Calculate the ratio of the first coefficient of each candidate participle associated with each candidate title to the second coefficient of the candidate participle, where the first coefficient of each candidate participle is the number of occurrences of the candidate participle in the candidate title, and the second coefficient of each candidate participle is the number of candidate titles that include the candidate participle; For each candidate title, the sum of the ratios of all candidate word segments associated with the candidate title is used as the similarity score between the candidate title and the candidate text.

19. The device according to claim 16 or 17, wherein There are many query algorithms for obtaining candidate titles, and the calculation subunit is specifically used to: For each query algorithm, calculating an initial similarity score between each candidate title obtained using the query algorithm and the candidate text; For each candidate title, a sum of multiple initial similarity scores of the candidate title is calculated as the similarity score between the candidate title and the candidate text.

20. A knowledge recommendation system comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 12.

21. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-12.

22. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 12.

Images