Search and recommendation method and system based on tourism knowledge graph

Abstract

The application provides a search and recommendation method based on a tourism knowledge graph. The application comprises: in a label preparation stage, extracting nouns and corresponding characteristics from a tourism site introduction and comments according to rules and processing the nouns and the corresponding characteristics into labels; in a model training stage, transmitting the labeled data into a PEGASUS model pre-trained by LCSTS for training, and then transmitting unlabeled data and debugging; in a search construction stage, associating the graph with a source site and storing the graph in a database to construct a visual module and a query transmission module of a system; and in a recommendation construction stage, constructing a recommendation calculation module of the system. The application has the following advantages: 1. The labels retain the original meaning and the superior-inferior relationship, which facilitates future classification of the graph; 2. Synonyms are processed according to the principle of majority covering minority, so that the labels are generated as uniformly as possible, long-term work is reduced, and the growth of traditional inverted index vocabulary is avoided; 3. Site reputation feature retrieval is more in line with user needs than text retrieval; and 4. Compared with traditional recommendation algorithms, the application is more interpretable.

Description

TECHNICAL FIELD

[0001] The present application relates to a search and recommendation method based on a tourism knowledge graph, belonging to the technical field of network information retrieval. BACKGROUND

[0002] From the technical meaning, information retrieval technology is that users search for specific network data, information and other resources through network retrieval terminals according to certain purposes, and use the obtained resources. Although knowledge graph and multimedia retrieval technology have developed to a certain extent, the most mainstream network retrieval technology is still the vector space model, probability model and Boolean model based on text retrieval.

[0003] Traditional text retrieval technology faces the problems of solving ambiguity, synonymy and synonymy in natural language processing, which makes it difficult to meet the diversified needs of users and solve the problem of excessive growth of word index sequence. SUMMARY

[0004] The technical problem to be solved by the present application is to provide a search and recommendation method based on a tourism knowledge graph, thereby avoiding and solving the problems of low-efficiency index, ambiguity and rigid text retrieval that cannot meet the personalized search needs of users caused by synonymy and synonymy in the tourism field.

[0005] The technical solution of the present application is: a search and recommendation method based on a tourism knowledge graph, comprising the following steps,

[0006] Step 1: extracting nouns from the introduction and comments of the tourism site according to the rules, processing the corresponding characteristics and labeling as tags;

[0007] Step 2: inputting the tags in Step 1 into the PEGASUS model trained by LCSTS, and then inputting the unmarked data and debugging;

[0008] Step 3: associating the graph with the source site and storing it in the database to build the visualization module and query transmission module of the system;

[0009] Step 4: calculating the difference degree according to the difference between the proportion of the occurrence times of each graph, building the recommendation calculation module of the system, and the recommendation module calculating the difference degree between the current site and the searched site and recommending the corresponding site.

[0010] In one possible design, Step 1 is specifically:

[0011] Step 1.1: extracting all nouns from the introduction and comments of a part of the tourism site;

[0012] Step 1.2: extracting all characteristics corresponding to the nouns;

[0013] Step 1.3: Combine the noun with all its corresponding properties to generate a basic label;

[0014] Step 1.4: Standardize the basic label according to the rules of brevity, moderate granularity, and synonymous coverage. Simplify complex properties while preserving the original meaning and avoiding excessive generalization. Process labels with similar or close meanings according to the principle of majority covering minority. Finally, obtain standardized labels.

[0015] In one possible design, Step 2 includes the following steps:

[0016] Step 2.1: First, transfer the labeled data into the PEGASUS generative Chinese text model trained by LCSTS for transfer training. Let the model's output direction learn towards the standardized label direction to adapt to the rules;

[0017] Step 2.2: Then, transfer the unlabeled data into the PEGASUS model after transfer training, and standardize the label data that does not conform to the rules according to the processing rules in Step 1;

[0018] Step 2.3: Add the standardized labels in Step 2.2 to the training set and repeat Steps 2.1-2.3.

[0019] In one possible design, Step 3 includes the following steps:

[0020] Step 3.1: Associate the graph labels extracted after Step 2 training with their source places, so that the graph labels become the identification of their source places;

[0021] Step 3.2: Store the associated structured data in the MySQL database;

[0022] Step 3.3: Build a search system to realize visual search;

[0023] In one possible design, Step 3.3 includes the following steps:

[0024] Step 3.3.1: Configure the visualization module to provide users with three nodes of food, scenic spots, and accommodation, and a visual knowledge graph;

[0025] Step 3.3.2: Configure the query transmission module to receive user clicks and generate SQL statements to the database;

[0026] Step 3.3.3: Configure the database to return query results to the query transmission module;

[0027] Step 3.3.4: The configuration query transmission module transmits the result to the visualization module;

[0028] Step 3.3.5: The configuration visualization module displays the knowledge graph or the place of the query to the user.

[0029] In one possible design, the Step 4 includes the following steps:

[0030] Step 4.1: The user selects whether to click to emphasize the emphasis of one or more graph labels, and then clicks the intelligent recommendation button to send the information to the query transmission module;

[0031] Step 4.2: The query transmission module generates a SQL statement and transmits it to the database;

[0032] Step 4.3: The database returns all places containing the same label to the query transmission module;

[0033] Step 4.4: The query transmission module transmits the query result to the recommendation module;

[0034] Step 4.5: The recommendation module calculates the difference between the current place and the searched place according to the recommendation algorithm and recommends the corresponding place to the visualization module according to the threshold;

[0035] Step 4.6: The visualization module visualizes the recommended information.

[0036] In one possible design, the Step 4.5 includes the following steps:

[0037] Step 4.5.1: Calculate the data of the current facility and , is the proportion of the xth word cloud selected in the current place in all word clouds in the current place, is the proportion of the xth word cloud not selected in the current place in all word clouds in the current place;

[0038] Step 4.5.2: Calculate the data of the searched other facility and , is the proportion of the xth word cloud selected in the other place in all word clouds in the other place, is the proportion of the xth word cloud not selected in the other place in all word clouds in the other place;

[0039] Step 4.5.3: According to the data of Step 4.3.1 and Step 4.3.2, the corresponding , and 、 Subtracting and then processing x is the proportion of a certain word cloud in the present location minus the proportion of the selected location, wherein 、 Also multiply a constant greater than 1 before the result In order to amplify the difference of the unselected word cloud to make the place with small difference of the selected word cloud more likely to be recommended;

[0040] Step4.5.4: sum the calculation results in 4.5.3 to obtain the difference degree "D". ."

[0041] The application also provides a search and recommendation device based on a tourism knowledge graph, comprising a memory, a control processor, and a computer program stored on the memory and executable on the control processor, wherein the control processor executes the program to realize the search and recommendation method based on the tourism knowledge graph as described above.

[0042] The application also provides a search and recommendation system based on a tourism knowledge graph, comprising the pulse width modulation device as described above.

[0043] The application also provides a computer readable storage medium, which stores computer executable instructions for making a computer execute the search and recommendation method based on the tourism knowledge graph as described above.

[0044] The application has the following advantages:

[0045] 1. The label retains the original meaning and the superior-inferior relationship, which is convenient for future graph classification;

[0046] 2. The same word is processed according to the majority covering the minority, so that the label generation is as unified as possible, the long-term work is reduced, and the traditional inverted index vocabulary growth is avoided;

[0047] 3. The place reputation feature retrieval is more in line with user needs than the text retrieval;

[0048] 4. Compared with the traditional recommendation algorithm, it has better interpretability. DETAILED DESCRIPTION

[0049] Figure 1 is the flowchart of the application;

[0050] Figure 2 is the diagram for obtaining nouns and constructing label of the application;

[0051] Figure 3 is the graph label extraction diagram of the application.

[0052] Figure 4 is a part of the atlas label diagram;

[0053] Figure 5 is a visual search diagram of the present application Figure 1 ;

[0054] Figure 6 is a visual search diagram of the present application Figure 2 ;

[0055] Figure 7 is a recommended function diagram of the present application. DETAILED DESCRIPTION

[0056] The present application will be further described below in conjunction with the drawings and specific embodiments.

[0057] A search and recommendation method based on a tourism knowledge graph, as shown in Figures 1-6 , includes the following steps,

[0058] Step 1: Extract the nouns and corresponding properties from the tourism site introduction and comments according to the rules and process them into labels;

[0059] Step 2: Transfer the labeled data into the PEGASUS model pre-trained by LCSTS for training, and then transfer the unlabeled data and debug;

[0060] Step 3: Associate the graph with the source site and store it in the database to realize search;

[0061] Step 4: Calculate the difference degree according to the difference between the proportion of each graph occurrence times to realize recommendation.

[0062] The Step 1 includes the following steps:

[0063] Step 1.1: Extract all nouns from a part of the introduction and comments of the tourism sites;

[0064] In this example, the introduction and comments about scenic spots, hotels, and food stores are crawled by using a web crawler. A total of 50,000 data are labeled before and after the migration training and debugging, and the data scale involved in the whole graph is: 949 rural tourism themes, 3133 rural scenic spots, 2152 scenic spot labels, 5002 food restaurants, 5 food labels, 4286 hotels and inns, 381300 hotel reviews, and 5156 hotel labels

[0065] Step 1.2: Extract all the properties corresponding to the nouns;

[0066] Step 1.3: Combine the nouns with all the properties corresponding thereto to generate basic labels;

[0067] The following two sentences are used as examples in this example:

[0068] "The room is clean, but a bit small. Overall, it's okay, and the transportation is very convenient." The extracted preliminary tags are room clean, room small, and transportation convenient.

[0069] "The room is clean, but a bit small. Overall, it's okay, and the transportation is very convenient." The extracted preliminary tags are room clean, room small, and transportation convenient.

[0070] Step 1.4: According to the rules of simplicity, efficiency, moderate granularity, and synonym coverage, the basic tags are standardized to simplify complex characteristics while preserving the original meaning and avoiding excessive generalization. At the same time, tags with the same or similar meanings are processed according to the principle of majority covering minority, and finally standardized tags are obtained.

[0071] In this example, the tags from Step 1.3 are standardized according to the rules: room clean and tidy, room small, and transportation convenient; room clean and tidy, service good, and environment noisy. For moderate granularity, when we encounter fruit picking and grape picking, if we take a larger granularity, we choose to keep the former, which undoubtedly reduces the workload but loses the original meaning of the tags. We take the appropriate granularity, i.e., grape picking, although we will handle different types of fruits in the future, but it is obvious that we have preserved the original meaning and the hierarchical relationship between the two tags. This facilitates the hierarchical association and classification of the graph when the graph is large and simplifies the graph.

[0072] The Step 2 includes the following steps:

[0073] Step 2.1: First, transfer the annotated data into the PEGASUS generative Chinese text model trained by LCSTS for transfer training, so that the output direction of the model learns to adapt to the standardized tags.

[0074] Step 2.2: Then, transfer the unannotated data into the PEGASUS model after transfer training, and standardize the tags that do not meet the rules according to the processing rules in Step 1.

[0075] Step 2.3: Add the standardized tags from Step 2.2 to the training set and repeat Step 2.

[0076] In this example, there are some tags that do not meet the rules after the first 5000 data are labeled and trained. The proportion of the results that do not meet the requirements is calculated and the 5000 data are standardized and added to the training set for further training. After the fourth time, the overall results have no significant changes.

[0077] The Step3 comprises the following steps:

[0078] Step3.1: associate the extracted graph label with its source place, so that the graph label becomes the identification of its source place;

[0079] Step3.2: store the associated structured data into the MySQL database;

[0080] Step3.3: realize visual search through dijango and Echart;

[0081] The Step3.3 comprises the following steps:

[0082] Step3.3.1: the visualization module provides the user with three nodes of food, scenic spots and accommodation or visual knowledge graph;

[0083] In this example, first, the Echart chart is used to display the node or knowledge graph. And the JavaScript language is used to set the click event, and every time the click event is triggered, the function of the query transmission module will be run.

[0084] Step3.3.2: the query transmission module receives the information of the user's click and transmits the generated SQL statement to the database;

[0085] The transmission module function will obtain the text content of the clicked element, and judge whether the user clicks the node or the label graph according to the text content. The front and back ends are combined to send an ajax request through dijango, and then an appropriate SQL statement is generated according to the judgment result and the query is performed.

[0086] Step3.3.3: the database returns the query result to the query transmission module;

[0087] Step3.3.4: the query transmission module transmits the result to the visualization module;

[0088] The query transmission module triggers the event according to the previous judgment after obtaining the data through the ajax request. According to the judgment, it will choose to create a new chart or replace the data in the chart.

[0089] Step3.3.5: the visualization module displays the knowledge graph or the place of the query to the user.

[0090] The Step4 comprises the following steps:

[0091] Step4.1: the user selects whether to click to emphasize the emphasis degree of one or more graph labels, and then clicks the intelligent recommendation button to send the information to the query transmission module;

[0092] In this example, we choose to contain "leisure picking, wine making, farming culture" characteristics of Jinsu organic grape picking garden as the basis and select the leisure picking label to search, trigger the query transmission module event.

[0093] Step4.2: Query transmission module generates SQL statement and transmits to database;

[0094] Step4.3: The database returns all the places containing similar labels to the query transmission module;

[0095] Step4.4: Query transmission module transmits the query result to the recommendation module;

[0096] We send an ajax request to search all the scenic spots containing the three labels in the MySQL database and their corresponding labels and their occurrence times. The following results are obtained: ['Jifa agricultural dream kingdom', 'Chaihe town know young city', 'Jiufengshan Yangxin Valley', '......', 'Oriental giant turtle garden', 'Akira ecological garden', 'Sichuan peach garden', 'Yuchun village', 'Bangli farm', 'Nantangzi village'] 165

[0097] Step4.5: The recommendation module calculates the difference between the current place and the searched place according to the recommendation algorithm And according to the threshold, recommend the corresponding place to the visualization module;

[0098] Step4.6: The visualization module visualizes the recommended information.

[0099] Step4.5 is as follows:

[0100] Step4.5.1: Calculate the data of the current facility And , is the proportion of the xth word cloud selected in the current place in all word clouds in the current place, is the proportion of the xth word cloud not selected in the current place in all word clouds in the current place;

[0101] In this example, first calculate And The following results are obtained: =[('Jifa agricultural dream kingdom', 'leisure picking', 0.333333)], =[('Jifa Agricultural Dream Kingdom', 'Leisure Picking', 0.25), ('Quchaihe Town Educated Youth City', 'Leisure Picking', 0.25), ('Jiufeng Mountain Yangxin Valley', 'Leisure Picking', 0.125), ('Yunzhai Village', 'Leisure Picking', 0.25), ('Xiji Big Cherry Plantation', 'Leisure Picking', 0.25), ..., ('Lukai Ecological Manor', 'Leisure Picking', 0.2), ('Ancient Post Road Town', 'Leisure Picking', 0.25), ('Seventy-Eighth Regiment', 'Leisure Picking', 0.25), ('Yucun Village', 'Leisure Picking', 0.2), ('Nantangzi Village', 'Leisure Picking', 0.166667)] 165

[0102] Step 4.5.2: Calculate data for other facilities being searched. and , This refers to the percentage of the x-th word cloud selected from other locations out of all word clouds in other locations. This refers to the percentage of the xth word cloud that was not selected in other locations out of all word clouds in other locations;

[0103] Similarly, we calculate The following results were obtained: =[('Jin Su Organic Grape Harvesting', 'Winemaking', 0.142857), ('Jin Su Organic Grape Harvesting', 'Agricultural Culture', 0.25)]2, =[('Qingtianzhai Farm', 'Agricultural Culture', 0.142857), ('Jinguo Tiandi (Beijing) Ecological Technology Co., Ltd.', 'Agricultural Culture', 0.333333), ('Green Manor', 'Agricultural Culture', 0.25)]3

[0104] Step 4.5.3: Based on the data from Step 4.3.1 and Step 4.3.2, convert the corresponding... , and corresponding , Subtract and then pass Processing. Because the original expression In this context, x represents the percentage of a word cloud representing the current location minus the percentage representing the selected location. If x is negative, it means the selected location has a higher percentage in the word cloud for that term than the current location, indicating that the selected location is superior in that aspect. Therefore, the word cloud value is 0, signifying no difference. , You also need to multiply the result by a constant greater than 1. (1.5 by default), to amplify the difference of the unselected word cloud to make the place where the selected word cloud is less likely to be recommended.

[0105] Step4.5.4: Sum the results in 4.4.3 to get the difference "D" cx is the proportion of the xth selected word cloud in other places in all word clouds in other places; ux is the proportion of the xth unselected word cloud in other places in all word clouds in other places; Then according to the threshold value (0.5 by default), select places with a difference less than 0.5 for recommendation.

[0106] Finally, calculate = [('8th winery', 0.0), ('"Fengyi Flower" agricultural demonstration park', 0.0), ('"Kash food" na culture industry park', 0.0), ('"Fishing village thick" ecological leisure farm', 0.0), ('10,000 mu of ecological sightseeing garden', 0.0), ('10,000 mu of grape picking garden', 0.0), ('Three teams', 0.0),...., ('Galaxy sunshine ecological agricultural park', 0.16666599999999998), ('Jifah agricultural dream kingdom', 0.16666599999999998), ('Liu Leng countryside town green picking base', 0.26666599999999996), ('Beautiful leg agricultural park', 0.26666599999999996), ('He Run Fude Manor', 0.33333199999999996), ('Forget worry farm', 0.33333199999999996), ('Camphor garden', 0.33333199999999996), ('Tian Dian garden folk tourism resort', 0.33333199999999996), ('Shimenwan four seasons orchard', 0.33333199999999996), ('Green snow bud white tea manor', 0.33333199999999996), ('Qingtian village farm', 0.38095199999999996), ('Wild tree forest mushroom garden', 0.44444399999999995), ('Nine peaks mountain heart valley', 0.624999), ('Khanka Lake Ecological Tourism Resort', 0.624999)] 169

[0107] Here, according to the default threshold value, we can select all the scenic spots except Khanka Lake Ecological Tourism Resort for recommendation.

[0108] ​The application further provides a search and recommendation device based on a tourism knowledge graph, comprising a memory, a control processor, and a computer program stored on the memory and executable on the control processor, wherein the control processor executes the program to implement the search and recommendation method based on the tourism knowledge graph.

[0109] The application further provides a search and recommendation system based on a tourism knowledge graph, comprising the search and recommendation device based on the tourism knowledge graph.

[0110] The modulation method according to the embodiments of the present disclosure can be written as a computer program and stored on a computer readable storage medium. When the computer program is executed by a processor, the search and recommendation method based on the tourism knowledge graph can be implemented.

[0111] The application can be further illustrated by the following experimental results.

[0112] Experimental environment:

[0113] General operating environment: CPU is Intel(R) Core(TM) i5-10200H CPU @ 2.40GHz, memory is 16GB, operating system is windows 10, environment is Python3.9, dijango4.1, and MySQL8.0.

[0114] Model training environment: GPU is Tesla V100, environment is Paddle

[0115] Experimental data: pre-training uses LCSTS data set, small sample migration training uses the introduction and comment data about scenic spots, accommodation, and food obtained by web crawler, and extracts the summary of Chinese text.

[0116] Experimental result analysis: the application adopts small sample migration training, and after 5000 standard graph labels are processed according to the rules, the pre-trained model is subjected to small sample migration training. Then, the results are sequentially corrected by 5000 standardization and re-added to the training set for training, so that the model gradually learns in the direction of adapting to the rules.

[0117] Table one is the rule adaptation rate in the migration training result, and it can be seen that the adaptation rate of the last four times has basically stabilized.

[0118] Table one

[0119]

[0120] Table two is the number of individual labels of the first seven accommodations, and it can be seen that the model generation result is in a relatively unified direction.

[0121] Table two

[0122]

[0123] Examples of computer-readable storage media include read-only memory (ROM), programmable read-only memory (PROM), electrically programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), random-access memory (RAM), dynamic random-access memory (DRAM), static random- access memory (SRAM), flash memory, non-volatile memory, CD-ROM, CD-R, CD+R, CD-RW, CD+RW, DVD-ROM, DVD-R, DVD+R, DVD-RW, DVD+RW, DVD-RAM, BD-ROM, BD-R, BD-R LTH, BD-RE, Blu-ray or optical disk storage, hard disk drives (HDDs), solid-state drives (SSDs), card-type memories (such as multimedia cards, secure digital (SD) cards, or extreme digital (XD) cards), magnetic tapes, floppy disks, magneto-optical data storage devices, optical data storage devices, hard disks, solid-state disks, and any other devices configured to store computer programs and any associated data, data files, and data structures in a non-transitory manner and provide the computer programs and any associated data, data files, and data structures to a processor or computer so that the processor or computer can execute the computer programs. In one example, the computer programs and any associated data, data files, and data structures are distributed over a networked computer system so that the computer programs and any associated data, data files, and data structures are stored, accessed, and executed in a distributed manner by one or more processors or computers.

[0124] Although the methods are illustrated and described above as a series of actions, it will be appreciated that the methods are not limited by the order of actions, as some actions can occur in different orders or concurrently with other actions from those illustrated and described herein or in other actions not specifically mentioned herein, depending on the implementation. Those skilled in the art will further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the embodiments disclosed herein can be implemented as electronic hardware, computer software, or combinations of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality, without reference to a particular logical arrangement of hardware or software. Such functionality can be implemented with up to an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), discrete components, or any combination(s) thereof. Such functionality can also be implemented with a general purpose processor, a digital signal processor (DSP), a microprocessor, or other processing component that can be programmed to perform the functions described herein. The various illustrative logical blocks, modules, and circuits described in connection with the embodiments disclosed herein can be implemented or performed with a general purpose processor, a DSP, an ASIC, an FPGA or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor can be a microprocessor, but in the alternative, the processor can be any conventional processor, controller, microcontroller, or state machine. A processor can also be implemented as a combination of computing devices, e.g., a combination of a general purpose processor and a DSP, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration. The steps of a method or algorithm described in connection with the embodiments disclosed herein can be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module can reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor such that the processor can read information from, and write information to, the storage medium. In the alternative, the storage medium can be integral to the processor. The processor and the storage medium can reside in an ASIC. The ASIC can reside in a user terminal. In the alternative, the processor and the storage medium can reside as discrete components in a user terminal. In one or more exemplary embodiments, the functions described can be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions can be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Computer-readable media include both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another.Computer-readable media includes both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. A storage media can be any available media that can be accessed by a computer. By way of example, and not limitation, such computer-readable media can comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. Also, any connection is properly termed a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of medium. Disk and disc, as used herein, include compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk and Blu-ray® disc where disks usually reproduce data magnetically, while discs reproduce data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0125] The above descriptions are only the preferred embodiment of the application, not intended to limit the application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall be included in the protection scope of the application.

[0126] Finally, it should be noted that the above examples are used to illustrate the technical solutions of the present application, but not to limit the application; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacements for part of the technical features; and these modifications or replacements do not drive the essence of the corresponding technical solutions out of the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A search and recommendation method based on a tourism knowledge graph, characterized in that, Comprising the following steps, Step1: according to the rules, extract the nouns from the introduction and comments of the tourist sites, and process the corresponding characteristics to be labeled as tags; Step2: the tags in Step1 are transmitted into the LCSTS pre-trained PEGASUS model for training, and then transmitted into unannotated data and debugged; Step3: the graph is associated with the source place and stored in the database to construct the visualization module and query transmission module of the system; the Step3.3 comprises the following steps: Step3.3.1: configure the visualization module to provide users with three nodes of food, scenic spots and accommodation and a visual knowledge graph; Step3.3.2: configure the query transmission module to receive the information clicked by the user and transmit the generated SQL statement to the database; Step3.3.3: configure the database to return the query result to the query transmission module; Step3.3.4: configure the query transmission module to transmit the result to the visualization module; Step3.3.5: configure the visualization module to display the knowledge graph or the searched place to the user; Step4: according to the difference between the proportion of the occurrence of each graph, the difference degree is calculated to construct the recommendation calculation module of the system, and the difference degree between the current place and the searched place is calculated by the recommendation module, and the corresponding place is recommended, and the Step4 comprises the following steps: Step4.1: the user selects whether to click to emphasize the emphasis degree of one or more graph labels, and then clicks the intelligent recommendation button to send the information to the query transmission module; Step4.2: the query transmission module generates an SQL statement and transmits it to the database; Step4.3: the database returns all the places containing the same label to the query transmission module; Step4.4: the query transmission module transmits the query result to the recommendation module; Step 4.5: The recommendation module calculates the difference degree between the current place and the searched place according to the recommendation algorithm and recommends the corresponding place to the visualization module according to the threshold; the Step 4.5 comprises the following steps: Step4.5.1: Calculate the data of the current facility and , is the proportion of the xth word cloud selected in the current site in all word clouds in the current site, is the proportion of the xth word cloud not selected in the current site in all word clouds in the current site. Step4.5.2: Calculate the data of other facilities searched and , is the proportion of the xth word cloud selected in other places in all word clouds in other places, is the proportion of the xth word cloud not selected in other places in all word clouds in other places; Step4.5.3: Subtract the corresponding , and corresponding , after processing via , x is the proportion of the current place of a certain word cloud minus the proportion of the selected place, where , Also multiply a constant greater than 1 before the result , to amplify the difference of the unselected word cloud to make the selected word cloud difference small place more likely to be recommended; Step4.5.4: sum the calculation results in 4.5.3 to obtain the difference degree ； Step4.6: the visualization module visualizes the recommendation information. 2.The method of claim 1, wherein, The Step1 is specifically: Step1.1: first, extract all the nouns from the introduction and comments of a part of the tourist sites; Step1.2: extract all the characteristics corresponding to the nouns; Step1.3: combine the nouns with all the characteristics corresponding thereto to generate basic labels; Step1.4: according to the rules of being concise and skilled, having moderate granularity and synonym coverage, the basic labels are standardized to simplify complex characteristics while retaining the original meaning, and the labels with the same or similar meanings are processed according to the principle of majority covering minority, and finally the standardized labels are obtained. 3.The method of claim 1 or 2, wherein, The Step2 comprises the following steps: Step2.1: first, the labeled data is transmitted into the LCSTS pre-trained PEGASUS generative Chinese text model for training to migrate training, so that the output direction of the model learns to the standardized label direction to adapt to the rules; Step2.2: then, the unannotated data is transmitted into the PEGASUS model after migration training, and the label data not meeting the rules is standardized according to the processing rules in Step1; Step 2.3: Add the normalized label in Step 2.2 to the training set, repeat Step 2.1-2.3 until the label meets the adaptation rules and the adaptation rate is stable. 4.The method of claim 1 or 2, wherein, The Step 3 comprises the following steps: Step 3.1: Associate the graph label extracted after Step 2 training with its source place, so that the graph label becomes the identification of its source place; Step 3.2: Store the associated structured data into the MySQL database; Step 3.3: Build a search system to realize visual search. 5.A search and recommendation device based on a tourism knowledge graph, characterized in that, A computer readable storage medium stores a computer executable instruction, and the computer executable instruction is used to make a computer execute the search and recommendation method based on the tourism knowledge graph as claimed in any one of claims 1-4. 6.A search and recommendation system based on a tourism knowledge graph, characterized in that, The search and recommendation device based on the tourism knowledge graph as claimed in claim 5.

7. A computer readable storage medium characterized in that, The computer readable storage medium stores a computer executable instruction, and the computer executable instruction is used to make a computer execute the search and recommendation method based on the tourism knowledge graph as claimed in any one of claims 1-4.

Images