CN111914157A - Travel scheme generation method and system based on user preference - Google Patents

Abstract

The invention relates to a travel scheme generation method based on user preferences, which comprises the following steps that a cloud service platform receives at least one travel demand input by a first user through at least one of a plurality of computing devices of the first user, and executes the following steps: in response to at least one travel requirement, pushing at least one first text message related to the at least one travel requirement that can be captured and recorded by at least one of the number of computing devices; acquiring a first user preference information set in response to a specific operation of the first user on the first text information, and generating a plurality of first travel schemes at least partially matched with the first user preference information set aiming at the first text information; and the optimized first user preference information set is stored in the cloud service platform in a mode that a second user can match and retrieve at least according to user preferences.

Description

Travel scheme generation method and system based on user preference

Technical Field

The invention relates to the technical field of travel recommendation, in particular to a travel route recommendation method and system, and particularly relates to a travel scheme generation method and system based on user preferences.

Background

In recent years, with the explosive development of the internet, various kinds of information have been explosively increased. The birth of the recommendation technology can help people to acquire the resources in which the people are interested. The recommendation technology is developed more mature in the aspect of electronic commerce, and products in China, such as companies in Alibara, Tencent, Baidu, Jingdong and the like, adopt the recommendation technology to recommend various similar contents to users to different degrees. Currently, the adopted recommendation technologies mainly include collaborative filtering recommendation, content-based recommendation, knowledge-based recommendation, combined recommendation, and the like.

However, the travel route is influenced by many factors, such as real-time traffic flow, weather, user preference, and the like, and the condition is complicated and variable. The recommendation of a tour route is now in an immature stage. In the process of planning a tour route, tourists face complicated service information, the tourists are difficult to obtain the most favorable scheme for the tourists from numerous recommendation information, the consumption requirements of different consumers are different, and the traditional service information recommendation mode at the present stage mainly adopts a mode of information packaging and timing pushing (such as through a special tour recommendation website), namely what is recommended, the personal preference of the user and whether the consumption requirements exist at the present moment are not considered, and the information of scenic spots is not considered, so that the recommendation effect is not in line with the preference of the user.

For example, chinese patent publication No. CN105468679B discloses a travel information processing and scheme providing method, which includes: obtaining relevant information of tourist attractions; carrying out structuring processing on the relative information of the scenic spots to obtain effective information of the scenic spots, and storing the structured effective information of the scenic spots in a database; carrying out distributed label classification on the effective information of the scenic spots, dividing all the scenic spots into a plurality of scenic areas, and storing the divided information of the scenic spots in a database; acquiring user requirements, and generating a recommended road route according to the user requirements and scenic spot effective information and scenic spot division information in a database; and feeding back the recommended route to the user. The invention can acquire the effective information of the scenic spot from the complicated information of the internet, and provides a feasible scheme aiming at the personalized requirements of tourists to customize personalized services for users. However, although the travel scheme provided by the patent considers that a personalized travel route recommendation scheme aiming at the user requirement is generated according to the user requirement, the formulation of the travel route is limited by various factors, such as weather information, traffic information and the like, and the labeled mechanical acquisition of the user requirement cannot combine the user perception, the user psychological feeling and the user economic capability to obtain the preferential travel route scheme according with the user preference. Moreover, the method more combines the information of the scenic spots with the requirements of the user, rather than analyzing the behavior information of the user to obtain the preference of the user, and the situation that all the scenic spot information of the database cannot meet the requirements of the user may occur.

For example, chinese patent publication No. CN108681586A discloses a tourist route personalized recommendation method based on crowd sensing, which includes: modeling an actual road network, and constructing POI location relation clusters according to the established road network model; constructing a user multivariate constraint interest model; calculating interest matching scores of the user, the scenic spots and the restaurants according to the user multivariate constraint interest model and the POI relevant information, and integrating the crowd sensing social score of the POI and the crowd sensing location score of the POI to obtain a target function; for single POI type route recommendation without going to scenic spots, a distance attenuation value is fused into a target function to obtain a comprehensive score, and then a varying neighbor greedy tourism route recommendation algorithm is adopted to dynamically insert the scenic spot with the highest comprehensive score to obtain an optimal route which accords with the preference of a user; for the recommendation of scenic spots which need to go, a single/multi-POI type two-section greedy tourism route recommendation algorithm is adopted, a basic route which only contains the scenic spots which need to go is obtained by utilizing a random segmentation tourism route recommendation algorithm, and the basic route is expanded by utilizing a clustering sequencing team tourism route recommendation algorithm, so that the final route which accords with the preference of the user is obtained. According to the method, on the basis of user interest matching, the POI social score perceived by crowd sensing and the POI location score perceived by crowd sensing are integrated, so that the score is more comprehensive. In addition, the provided scenic spot recommendation algorithm applicable to the unnecessary scenic spots, namely the variable neighbor tourist route recommendation algorithm and the two-stage greedy tourist route recommendation algorithm applicable to the single/multiple POI type scenic spots with the unnecessary scenic spots, are low in time complexity, better accord with user preferences and more reasonable. However, the method for recommending a tourist route disclosed in the patent adopts a crowd-sourcing thought combining crowd-sourcing perception and a data acquisition mode perceived by a mobile device, and since the data acquisition mode is based on unconscious cooperation of a user, can perceive multivariate heterogeneous data and has a wide and uniform coverage, the method can analyze preference information of the user to formulate the tourist route by combining a structured multivariate constraint interest model of the user with POI (point of interest) related information perceived by the crowd-sourcing perception, but the method needs to process massive data brought by the crowd-sourcing perception, so that not only is the generation speed of a recommended route scheme slow, but also a large amount of information with low relevance or useless information such as tourism and scenic spots can be obtained due to weak relevance of the massive data and the tourism, and finally the recommended tourist route does not accord with the preference of the user.

How to fully utilize a large amount of related data in the current internet to automatically provide accurate travel route planning service for users is a problem to be solved urgently in the current travel field. In the traditional travel route planning research, most recommendation methods are used for modeling and solving a travel route planning problem, and the solution method is difficult to meet the individual requirements of users in actual travel, so that great discrepancy is often generated with the travel route planning problem in the real world. One research effort has indicated that over 87% of customers rely on online user-generated content to make decisions for travel, and online searching has become one of the main ways users make travel route planning, behind which is the support for massive amounts of user-generated information, and while many current travel websites provide information about destinations and route choices, integrating and comparing different types of information from massive amounts of users requires a great deal of time and effort, and numerous choices make it difficult for consumers to find what they are looking for, so mining some of the user-required content from this information helps users make decisions that is the focus of current research.

For example, chinese patent publication No. CN107679961A discloses a method for personalized tour route planning based on knowledge map, which utilizes knowledge map to depict semantic information in tour spots and tour routes, and utilizes a route generation model and a user-route similarity calculation method to comprehensively consider three important factors of time, cost and preference of users during travel to construct a personalized tour route that best meets the user preference and constraint. The method provided by the patent utilizes a crawler technology to extract historical travel routes from travel notes and add the travel routes to a knowledge-graph route base, and constructs travel routes according to time, cost and preference. However, the method for recommending the travel route provided by the patent predicts the travel preference of the tourist based on the tourist with similar behaviors, so that the recommended travel route cannot well meet the personalized requirements of the tourist, the travel route of the tourist in other cities is not used for learning the travel preference of the user, when the tourist comes to a new city, the data of the user behavior is sparse or the data distribution is not uniform, the problem of cold start occurs, the data in the knowledge map route library is possibly zero, and the personalized degree of the recommendation structure and the game experience degree are low.

For example, chinese patent publication No. CN108829852A discloses a personalized tour route recommendation method, which includes: step 1, acquiring tourist information of a tourist and preprocessing the tourist information; step 2, obtaining a scenic spot type representation vector based on the category information; step 3, acquiring a tourism preference representation vector of each user, a tourism preference representation vector of tourists in each month and a representation vector of each scenic spot; step 4, obtaining a candidate play route set according to the step 1; step 5, obtaining a to-be-candidate tour route from the candidate play route set according to personal constraints; step 6, obtaining a preference expression vector of each travel route; and 7, performing similarity matching on the playing preference of the user and the candidate route to obtain a playing route which is most matched with the playing preference of the user and is used as a travel route recommended to the user finally. Although the recommendation method provided by the patent obtains the personalized tourism preference of the tourist according to each scenic spot in the historical playing track of the tourist and the scenic spot category information to which the scenic spot belongs, the cold start problem is relieved to some extent, but the problems of data sparseness and uneven data distribution are still not solved.

In a conventional recommendation system, collaborative filtering or SVD matrix decomposition is generally used for recommendation, but the problems of cold start and the like need to be overcome. In the aspect of processing text content, such as text classification, a common method is to classify the text content by a probabilistic model, for example, naive bayes, but the probabilistic model may encounter situations of data sparseness, data maldistribution, and the like, and does not relate to relevant information of travel places. In the location-based algorithm recommendation, the tourist attraction distance of the residence of the tourist is usually adopted as important reference information. Moreover, some hidden variable models use a matrix decomposition to find the hidden features of a place. Information that is easily ignored, such as a user's hidden feature and a place's hidden feature, includes rich text information.

For example, chinese patent publication No. CN106934056A discloses a personalized travel note recommendation method based on a probability map model, which includes the following steps: step 1, initializing a travel note main body, segmenting words of a travel note article, obtaining main body distribution of each travel note and topic distribution of each word by adopting a standard article topic model and Gibbs sampling, copying relevant parameters of gamma distribution of the travel note and the words by using the calculated topic distribution, and endowing initial values for the relevant parameters of hidden characteristics of a place by using random numbers; step 2, for each note in each note, calculating a logarithm value of a word frequency relation through the distribution of a word theme and an article main body, and updating each note and a shape parameter in gamma distribution parameters of words in the note; step 3, aiming at each travel note of each user comment, calculating a logarithmic value of the user participating in the travel note comment according to user preference distribution, travel note main body distribution and location hidden characteristics, and updating shape parameters in gamma distribution parameters of the user, the travel note and the location; step 4, updating the scale parameters of all gamma distributions; and 5, predicting from the verification data set through the user preference and the hidden place characteristic trained by the training set. However, although the recommendation method provided by the patent adopts a poisson decomposition method, and finds out the potential hidden features by utilizing gamma distribution, rich text information is obtained, and the problem of tourist route recommendation in a data sparse scene is solved to a certain extent, the method provided by the patent does not consider the influence of geographic features on location recommendation for users, so that in the data sparse scene, such as remote location recommendation, the performance is poor, and the problem of cold start of new users and new articles cannot be effectively solved.

For example, chinese patent publication No. CN105718576A discloses a personalized position recommendation system related to geographic features, which mainly solves the problem of poor performance of the conventional collaborative filtering recommendation algorithm in a data sparse scene. The system comprises a data acquisition module DC, a user database module UD and a geographic database module DD, wherein the data acquisition module DC is used for acquiring the sign-in record of a target website user and providing data for the user database module UD and the geographic database module DD; the user database module UD is used for storing all information captured by the data acquisition module DC in a user database by taking the user ID as an index and providing user information for the user preference mining module UM; the geographic database module DD is used for storing all information captured by the data acquisition module DC in a geographic database by taking a geographic city as an index and providing geographic information for the geographic feature mining module DM; the user preference mining module UM is used for mining the preference of the user by using the data in the user database module UD to obtain a candidate position recommendation List1 sorted according to the preference of the user; the geographic characteristic mining module DM is used for mining geographic characteristics by using data in the geographic database module DD to obtain a candidate position recommendation List2 sorted according to the influence of the geographic characteristics; and the recommendation module RD is used for generating a final recommendation Result according to the candidate position recommendation List1 sorted according to the user preference and the candidate position recommendation List2 sorted according to the geographic characteristic influence. However, the method provided by the patent does not consider automatic learning of the user preferences, and only adopts a data mining technology to obtain the candidate location recommendation List1 of the user preferences to perform sorting recommendation in a linear manner, so that the method provided by the patent cannot acquire the user preferences in an automatic learning manner and cannot dynamically express the user preferences, and thus the degree of personalization of the user for obtaining the recommended route is not high, and the recommendation accuracy is low.

For example, chinese patent publication No. CN108681739A discloses a travel destination recommendation method based on user emotion and time dynamics, which includes the following steps: firstly, an opinion mining technology is used for obtaining a quantitative value of emotional tendency of a user, the emotional degree is integrated into a matrix decomposition model, and a time dynamic mechanism is adopted to represent user preference and the popularity of a travel destination along with the change of time. And combining the two elements of user emotion and time influence and fusing with an SVD + + method to construct a synthetic recommendation model. By analyzing the explicit or implicit feedback, the preference of the user is automatically learned, and the characteristics of the tourism and leisure resources are matched with the requirements of the user. However, the recommendation method provided by the patent acquires an emotional tendency matrix of a user, namely user preference, by adopting opinion mining based on user comments, and constructs the emotional tendency matrix related to the user preference to alleviate the problem of cold start, but the training mode only related to the emotional tendency of a single user cannot effectively acquire potential factors of user behaviors, and the problem of cold start cannot be solved. Moreover, the resulting data sparseness and uneven data distribution not only result in lack of data samples for automatic learning of user preferences, but also result in poor accuracy of recommended routes, which do not meet the user preferences. In addition, the method provided by the patent only adopts the emotional tendency of the user and the popularity of the tourist destination as recommendation bases, and ignores the objective requirements of the user.

In summary, there is a need to improve the prior art, and for the problem of cold start and data sparseness in the travel route recommendation technology, there is a need to provide a travel scheme generation method and system that is oriented to the objective needs of users for travel and can avoid the way of training and learning only according to the historical data of the users themselves and the preferences of the users to alleviate cold start and data sparseness.

Moreover, on the one hand, since the skilled person in the art who is understood by the applicant is necessarily different from the examination department; on the other hand, since the inventor made the present invention while studying a large number of documents and patents, the disclosure should not be limited to the details and contents listed in the specification, but the present invention should not have the features of the prior art, but the present invention should have the features of the prior art, and the applicant reserves the right to increase the related art in the background art at any time according to the related specification of the examination guideline.

Disclosure of Invention

Aiming at the problems of cold start and sparse data in the travel route recommendation technology, a travel text recommendation-feedback mechanism which is guided by the objective requirements of users for travel is provided to obtain preference information of the users, and other users are dynamically associated based on the preference information of the users, so that preference data are provided for the cold start users, and the objective requirements of the users can be met. The method and the system can avoid relieving cold start and data sparsity only by training a learning mode according to the historical data of the user and the preference of the user, and are not limited to popularity and geographical position of scenic spots for recommendation.

A travel scheme generation method based on user preferences comprises the steps that a cloud service platform receives at least one travel demand input by a first user through at least one of a plurality of computing devices of the first user. The method further comprises the following steps executed by the cloud service platform: in response to at least one of the travel needs, pushing at least one first textual information relating to at least one of the travel needs that is capable of being retrieved and recorded by at least one of the number of computing devices;

acquiring a first user preference information set of the first user in response to a specific operation of the first user on the first text information, and generating a plurality of first travel schemes at least partially matching the first user preference information set for at least one first text information;

optimizing a first user preference information set based on the selection result of the first user about the plurality of first travel schemes, and storing the optimized first user preference information set in the cloud service platform in a mode that a second user can match and retrieve at least according to user preferences.

According to a preferred embodiment, the cloud service platform obtains at least one second text message related to the travel demand of the second user through a plurality of computing devices of the second user. The cloud service platform acquires a second user preference information set of the second user in response to a specific operation of the second user on the second text information. The cloud service platform at least partially obtains user preference information matched with each other based on the similarity of the second user preference information set and the first user preference information set so as to update the second user preference information set and/or the first user preference information set.

According to a preferred embodiment, the cloud service platform generates a plurality of second travel schemes for the second text information, wherein the second travel schemes at least partially match the updated second user preference information set. The cloud service platform optimizes a second user preference information set based on the selection result of the second user about the plurality of second travel schemes. And the optimized second user preference information set is stored in the cloud service platform according to a mode that the first user and the third user can match and retrieve at least according to the user preference.

According to a preferred embodiment, the cloud service platform further performs the following steps:

obtaining second text information and a second user preference information set of a second user matching the travel needs of the first user in response to the travel needs;

obtaining reference user preference information that does not match each other and includes the first user preference information based at least in part on similarities of a first user preference information set and the second user preference information set of the first user;

generating a number of reference travel solutions for the first and second textual information that at least partially match the reference user preference information.

According to a preferred embodiment, the plurality of reference travel solutions are recommended to the first user on a priority basis greater than a plurality of first travel solutions of the first user. The cloud service platform optimizes the preference information set of the first user based on the selection result of the first user about the plurality of reference travel schemes and the plurality of first travel schemes.

According to a preferred embodiment, the method further comprises: the cloud service platform constructs a mapping chart at least comprising the user, the user preference information set and the travel scheme in a correlated mode through at least one of the plurality of computing devices of the first user and the plurality of computing devices of the second user. Presenting on the map a corresponding travel itinerary and a corresponding set of user preference information associatively labeled. The corresponding preference weight index is associatively labeled at the corresponding travel itinerary. The preference weight index represents a preference weight of the user's selection of the corresponding travel itinerary. The cloud service platform provides modification permission of the preference weight index for the user so that the user can modify the preference weight index within a preset modification range.

A personalized travel text information pushing method based on semantic understanding comprises the following steps executed by a cloud service platform:

generating at least one topic keyword comprising at least one sequence semantic based on information of the user's travel needs;

acquiring at least one initial text message based on at least one topic keyword associated user corpus content;

and dividing at least one piece of initial text information based on a limited domain formed by each topic keyword and the related intra-domain feature words thereof to extract corresponding travel text information meeting travel requirements in a mode of removing irrelevant high-frequency words and recommend the travel text information to users.

According to a preferred embodiment, the method further comprises: and the cloud service platform generates the sentence of the travel text information in a mode of extracting the initial text information along the continuous direction of the semantics of the in-domain feature words of the limited domain edge by taking the in-domain feature words of the limited domain edge as a starting point.

A travel itinerary generation system based on user preferences, the system including at least a cloud service platform and a number of computing devices of a first user in communication with each other. The cloud service platform is configured to:

receiving, by at least one of a number of computing devices of a first user, at least one travel demand input by the first user;

in response to at least one of the travel needs, pushing at least one first textual information relating to at least one of the travel needs that is capable of being retrieved and recorded by at least one of the first user's computing devices;

acquiring a first user preference information set of the first user in response to a specific operation of the first user with respect to the first text information;

when the first user preference information set data is sparse, at least partially obtaining user preference information matched with each other based on the similarity of the preference information set of at least one second user and the first user preference information set so as to update the first user preference information set;

a number of first travel itineraries are generated for at least one of the first textual information that at least partially match the updated first set of user preference information.

According to a preferred embodiment, the cloud service platform is further configured to: obtaining text information and a set of user preference information of at least one other user matching the travel needs of the first user in response to the travel needs;

obtaining, based at least in part on similarity of a first set of user preference information of the first user to the other sets of user preference information, a reference set of user preference information that do not match each other and that includes the first user preference information;

and the cloud service platform generates a plurality of reference travel schemes at least partially matched with the preference information of the reference users aiming at the first text information and the information set of other users.

According to a preferred embodiment, the cloud service platform is further configured to:

obtaining text information and a user preference information set of at least one second user matching the travel needs of the first user in response to the travel needs;

obtaining a reference user preference information set that does not match each other and that includes the first user preference information based at least in part on similarities of the first user preference information set and a second user preference information set;

generating a number of reference travel solutions for the first and second textual information that at least partially match the set of reference user preference information.

The invention provides a travel scheme generation method and system based on user preference, which at least have the following advantages:

(1) aiming at the cold start problem of the new user, the invention adopts a tourism text information recommendation-feedback mechanism which takes the tourism demand of the client as the guide to obtain the information data of the new user about the user preference, enriches the user preference data set of the new user by associating the user preference information sets of other users, and solves the cold pneumatic problem that the new user behavior data is insufficient;

(2) according to the method and the device, the user preference information is further obtained through the secondary feedback of the user about the first travel schemes, so that the user preference information set of the user is optimized, the accuracy of the user preference is improved, and the fitness of the travel schemes and the user preference is increased;

(3) according to the method and the device, under the condition that the travel demands of the first user and the second user are irrelevant, the first user preference information set and the second user preference information set are compared and analyzed, and the data with the tendency of potential user preference information can be mutually extracted, so that the first user preference information set and the second user preference information set can be enriched, and therefore the cold start problem can be solved by acquiring the data with the same preference information of other users under the condition that the data of the preference information sets of the users are insufficient.

(4) According to the method and the device, the reference information set which belongs to the first user preference information and does not belong to the second user preference information can be obtained, and a plurality of reference tourism schemes which are at least partially matched with the reference user preference information are generated according to the first text information and the second text information, so that the problem that tourism preferences of different tourists are predicted by tourists with similar behaviors can be avoided, and the personalized requirements of different tourists cannot be well met by the recommended tourism route.

Drawings

FIG. 1 is a block schematic diagram of a preferred embodiment of the system of the present invention; and

FIG. 2 is a schematic flow chart of example 2 of the present invention.

List of reference numerals

10: cloud service platform 20: number of computing devices of a first user

30: a number of computing devices of a second user

Detailed Description

The following detailed description is made with reference to fig. 1 and 2.

Example 1

The embodiment discloses a method for pushing travel text information, and under the condition of not causing conflict or contradiction, the whole and/or part of the contents of the preferred implementation modes of other embodiments can be used as a supplement of the embodiment. Preferably, the method may be implemented by the system of the present invention and/or other alternative modules. For example, the method of the present invention is implemented by using various modules in the system of the present invention.

A personalized travel text information pushing method based on semantic understanding comprises the following steps executed by acloud service platform 10. Preferably, thecloud service platform 10 may be a platform configured based on one or more servers, computers, and the like, and capable of providing functions of computing, searching, downloading, storing, managing, and the like, and the platform may be connected to other servers, computing terminals of users, and computing terminals of other content providers through the internet.

S100: at least one topic keyword comprising at least one sequence semantic is generated based on the information of the user's travel needs. Preferably, thecloud service platform 10 can receive the travel requirement input by the user through at least one computing device of the user. The at least one computing device of the user may be a cell phone, a notebook, a desktop, a tablet, etc. Preferably, the user-entered travel needs include at least fixed questionnaire-type content, such as departure location, departure time, arrival location, arrival time, vehicle, occupancy, attraction, return location, return time. But also personal basic information of the user, such as age, sex, number of travelers, etc. Potential preference requirements that the user actively enters, such as local food, celebrities, traffic information, length of journey, shopping information, movie theaters, etc., may also be included. Preferably, the user can input the requirement of special travel in a long sentence mode, for example, the user for travel is a disabled person, or an older user, or a child, or a user who needs to carry a pet, or a user suffering from a special disease, such as asthma, allergy, and the like.

Preferably, thecloud service platform 10 performs word segmentation on the travel demand of the user to obtain topic keywords. Preferably, no word segmentation process is required for the short words input by the user, such as departure point place prefecture, or Beijing. The special requirements for user input may be in the form of long sentences. For example, the user inputs that the elderly have hypertension and that the legs and feet are handicapped. This requires word segmentation processing for the long sentence. Preferably, the semantics are serialized, i.e., classified according to the information of the semantics, and a sequence number associated with the semantics is generated. For example, two words that differ in character, expressing the same semantic meaning, are attributed to the same sequence of semantic meanings.

Preferably, the special requirements input by the user can be extracted by the way of matching the existing hash table dictionary word by word. For example, the elderly have hypertension and handicaps in legs and feet, and keywords such as the elderly, hypertension, legs, feet, legs and feet, handicaps may be extracted. Preferably, the keywords may comprise keywords of a sequence semantic, e.g. legs, feet and legs and feet are semantics of a sequence. It is also possible that the hypertensive legs and feet have disabilities, such as keywords comprising three sequence semantics including hypertension, legs and feet and disabilities. Through the setting mode, keywords without a plurality of sequence semantics in a dictionary or a word bank can be identified, the tourism requirements in the long sentence form input by the user can be accurately obtained, and the problem that the user requirements cannot be accurately understood due to wrong word segmentation is avoided, so that the follow-up acquisition of the tourism text information is influenced.

S200: and acquiring at least one initial text message based on the at least one topic keyword associated with the user corpus content. Preferably, the user corpus content is content generated by a user in the internet, for example, the user shares his or her living status, tourism experience, or his or her shopping experience with different websites and forums through various mobile internet devices, or shares content on social media such as blogs and microblogs. And the system also comprises BBS, question and answer communities, wiki, video websites, barrages, comments and other contents. The user corpus content may be text, pictures, voice, video, etc. Preferably, thecloud service platform 10 performs a preliminary search based on the location and/or attraction information output by the user, so as to obtain a website of the internet. Preferably, thecloud service platform 10 may employ a web crawler technology to crawl relevant text information from relevant internet websites. Preferably, the related text information is roughly extracted. The rough extraction at least comprises the steps of removing meaningless elements such as pictures, symbols, blank lines and the like from related text information, and carrying out standardized conversion on full half angles, capital and small cases, traditional Chinese characters and the like in character punctuations, so that standard texts only retaining the contents of Chinese characters, English, numbers, punctuation marks, line feed marks and the like are generated. Preferably, thecloud service platform 10 stores each standard text and the original corresponding related text information in an associated manner. For example, thecloud service platform 10 builds a standard text library. The standard text library is used for storing the standard text generated by each piece of relevant text information and adding a storage address for storing the relevant text information. Through the setting mode, the relevant text information meeting the tourism requirements of the user can be completely acquired through the storage address, the subsequent relevant text information is conveniently processed to acquire the relevant complete sentences about the topic keywords, the defects of partial relevant information such as pictures, voice, videos and the like can be avoided, and the user can conveniently acquire and understand all information about the topic keywords provided by the relevant text information completely.

Preferably, thecloud service platform 10 performs word segmentation processing of single sequence semantics on the standard text by using a word bank or a dictionary. Preferably, the word stock or dictionary can adopt the word stock in the LJParser software library for word segmentation processing. Preferably, thecloud service platform 10 labels stop words in the standard text by using the stop word vocabulary. Preferably, the stop word list may be a stop word list within LJParser. Preferably, stop words may be labeled as "///////", and then with line breaks. By the setting mode, some words with low word formation rate can be deleted, the redundancy of calculation can be reduced, and the searching efficiency can be improved.

Preferably, the segmentation of the text is performed by using punctuation marks. The punctuation marks are generally used as the pause or the end of a segment of content expression in the text, so that the punctuation marks are used for segmenting the standard text, the word segmentation efficiency of words at least containing two, three or more sequence semantics can be improved, and the effect of searching the words with the plurality of sequence semantics can not be reduced.

Preferably, thecloud service platform 10 performs word segmentation processing on at least two sequences of semantic words based on the segmented standard text. Preferably, the processing comprises searching to generate words of the semantics of at least two sequences taking into account the frequency of occurrence of any two words immediately preceding and following, based on statistics of word frequency. For example, each portion of a standard text segmentation is considered a word vector consisting of a series of words of sequential semantics. Preferably, the number of sequence semantics in each part is counted to obtain the sequence semantics with the smallest statistical number, the maximum confidence between the sequence semantics and other sequence semantics is calculated based on the sequence semantics with the smallest statistical number, and when the maximum confidence exceeds a threshold value, the two sequence semantics are considered to form the semantics of a plurality of sequences. Preferably, the threshold value is in the range of 0-0.1. The threshold value may be set manually and evaluated to adjust the threshold value based on the result of the current threshold value. Preferably, the above processes are repeated to obtain a set of new words with two sequence semantics, and three sequence semantics new words are obtained based on repeating the above steps of segmentation, counting the sequence semantics with the least number, and calculating the maximum confidence. Preferably, the above process is repeated until a new word is not generated. Through the setting mode, the problem that new words formed by a plurality of sequence semantics appearing in the text are difficult to correctly segment when the random and irregular written text is processed can be avoided. Such as curry sirloin, signboard, sea park, etc., which are new words composed of two or more words with single semantic meaning.

Preferably, thecloud service platform 10 extracts at least one sequence semantic keyword of the standard text based on the above processing manner, so as to obtain initial text information about the standard text.

S300: and dividing at least one piece of initial text information based on a limited domain formed by the topic keywords and the related intra-domain feature words thereof to remove irrelevant high-frequency words and extract corresponding travel text information meeting travel requirements. Preferably, the initial text information obtained based on S200 includes new words generated by standard text. And searching the words in the new words which are the same as the topic keywords through the topic keywords. Preferably, in the standard text, intra-domain feature words around the topic keyword are searched. Preferably, in writing the text, when topic keywords are mentioned, some important features are mentioned. The words used for describing important characteristics of a certain aspect of the topic keywords are intra-domain characteristic words. For example, if "Chengdu Taigu" is the topic keyword, then the "square" of "Chengdu Taigu" is the feature word in its domain. For example, when the term "the palace" is a topic keyword, the term "buy tickets" is a feature word in the domain. Preferably, in the standard text, the position between the topic keyword and at least one feature word in the domain thereof constitutes a defined domain. The initial text information is divided through the limited domain, the standard text can be divided again on the basis of symbol segmentation, and therefore irrelevant words appearing with high frequency can be excluded from the limited domain. Preferably, the content within the initial text information is extracted through the divided bounding fields. For example, the topic keyword is "back sea food", but more than one place in the text message will have "back sea food", and some irrelevant high-frequency words will appear between two places "back sea food", for example, people may have a lot of "back sea food" when reviewing the travel text message and mentioning "back sea food", people may have a lot of "back sea food" and go to the back sea in winter and slide, and as a result, people are too many and not slide, and people can find a deep fry in a street of back sea food near the back sea skating rink when getting back to home, and the intestines are dipped with garlic juice, and people eat a good water! "information of this text, wherein" sea skating "occurs with high frequency is irrelevant high frequency vocabulary. And the second 'back sea food' is associated with the intradomain characteristic words such as 'one street', 'fried sausage' and 'garlic juice', so that the limited domains from 'back sea food' to 'garlic juice' are formed to extract the text information of the comments, and the irrelevant words of 'back sea skating' which appears at high frequency can be effectively presented.

S400: thecloud service platform 10 generates a sentence of travel text information in a manner of extracting initial text information along the continuous direction of the semantics of the in-domain feature words of the limited domain edge with the in-domain feature words of the limited domain edge as a starting point. Preferably, the content within the initial text information is extracted through the divided limited field, which may be incomplete. Preferably, punctuation marks are searched along the continuous direction of the semantics of the feature words in the domain, starting from each feature word in the domain. Preferably, the continuous direction of the semantics of the feature words in the domain is the expansion direction of the feature words in the domain, for example, if the feature words in the domain are "entrance tickets", then the price information, time information, etc. of the "entrance tickets" are described as words describing the expansion direction of the "entrance tickets". Preferably, punctuation marks that the first expression statement encountered by the vocabulary in the expansion direction completes serve as cut-off points. Thecloud service platform 10 extracts the limited domain and the text content to the cut-off point, so as to generate the recommended travel text information for the user. Through the setting mode, the tourism text information can be guaranteed to be a sentence which is expressed and completed based on the content extracted from the initial text information, and related information is prevented from being omitted. For example, the above example about the food of the sea afternoon can obtain the final key information about "good eating" through the setting mode, so that the information which is required to be expressed by the user for understanding the content expected to be generated by the user is facilitated.

Example 2

The embodiment discloses a travel scheme generation method, and under the condition of not causing conflict or contradiction, the whole and/or part of the contents of the preferred implementation modes of other embodiments can be used as a supplement of the embodiment. Preferably, the method may be implemented by the system of the present invention and/or other alternative modules. For example, the method of the present invention is implemented by using various modules in the system of the present invention.

A travel scheme generation method based on user preferences comprises acloud service platform 10 receiving at least one travel requirement input by a first user through at least one of a plurality ofcomputing devices 20 of the first user. The method further includes thecloud service platform 10 performing the following steps as shown in fig. 2:

s100: in response to the at least one travel requirement, at least one first textual information related to the at least one travel requirement is pushed that is capable of being acquired and recorded by at least one of the number of computing devices. Preferably, the first text information is the travel text information recommended by the method in embodiment 1.

S200: a first set of user preference information for the first user is obtained in response to a particular manipulation by the first user with respect to the first textual information, and a number of first travel itineraries that at least partially match the first set of user preference information are generated for at least one of the first textual information. Preferably, the specific operation includes a series of feedback operations such as comment, like, inquiry, share and the like of the user on the first text information. Preferably, positive or negative feedback of the first text information by the first user can be obtained by the comment. For example, the comment fed back by the user is searched for keywords representing positive tendency and negative tendency to obtain the potential preference tendency of the first text information of the first user. Preferably, the positive tendency keywords may include "thank you", "good you", and the like. The negative-oriented keywords may include words such as "what is," less eaten, "and" less tasted ". Preferably, the user can know that the user potentially prefers the information contained in the first text information through feedback of the behavior of praise, wind direction and the like. Preferably, the user is interested in a certain keyword in the first text information by querying the behavior feedback, and the potential preference of the first user for the certain keyword is obtained by retrieving the keyword of the query content. Through this mode of setting, can not obtain user's potential preference information, can arrive new user's action information through the potential preference information that obtains and tourism demand moreover, can alleviate the cold start problem, also can avoid the appearance of cold start problem to a certain extent. Preferably, the topic keywords included in the at least one first text message and the feature words in the domain are retrieved according to the first user preference information set derived interest keywords. A plurality of first travel plans are generated according to the number of words which are the same with each other and the occurrence frequency. The number and frequency of vocabularies can be set according to prior information or manually. The generated first travel plans can be arranged in a multi-stage mode based on the number of the matching keywords of interest.

S300: and optimizing a first user preference information set based on the selection result of the first user on the plurality of first travel schemes, and storing the optimized first user preference information set in a cloud service platform in a mode that a second user can match and retrieve at least according to user preferences. Preferably, the second user refers to a user who requests the generation of the travel itinerary in addition to the first user. For example, the user may be a new user without historical data, or may be a user who has historical data requesting travel solutions at least twice. Preferably, the user preference information on the basis of the first user preference information set by which the first user preference information set is optimized can further be derived from the results of the first user's selection of the number of first travel itineraries. Preferably, the optimization is to modify the information of the user preferences, deleting potential user preference information within the first set of user preference information that may not be relevant. Through the setting mode, the preference information of the user can be further accurately obtained, and therefore the travel scheme with high user preference degree can be generated.

S400: thecloud service platform 10 obtains at least one second text message related to the travel requirement of the second user through the plurality ofcomputing devices 30 of the second user. Thecloud service platform 10 acquires a second user preference information set of the second user in response to a specific operation on the second text information from the second user. Thecloud service platform 10 at least partially obtains the user preference information matched with each other based on the similarity between the second user preference information set and the first user preference information set to update the second user preference information set and/or the first user preference information set. Preferably, under the condition that the travel demands of the first user and the second user are irrelevant, the first user preference information set and the second user preference information set are compared and analyzed, and data with tendencies of potential user preference information can be mutually extracted, so that the first user preference information set and the second user preference information set can be enriched. Therefore, the problem of cold start can be solved by acquiring data of other users having the same preference information in the case where the preference information set data of the user is insufficient.

S500: thecloud service platform 10 generates a plurality of second travel schemes at least partially matching the updated second user preference information set for the second text information. Thecloud service platform 10 optimizes the second user preference information set based on the selection result of the second user about the plurality of second travel schemes. The optimized second user preference information set is stored in thecloud service platform 10 in a manner that the first user and the third user can match and retrieve at least according to the user preferences. Preferably, the third user refers to a new user who has not requested the generation of a travel itinerary, without historical data. Preferably, the updated, more data-rich second set of user preference information is used to match the second textual information to generate a number of second travel solutions for the second user. Compared with a plurality of first tourism schemes of a first user, the plurality of second tourism schemes generated by the setting mode have richer user preference information data and can obtain the tourism scheme with higher user preference degree.

S600: several reference travel scenarios are generated. Preferably, thecloud service platform 10 obtains the second text information of the second user matching the travel requirement of the first user in response to the travel requirement of the first user. Preferably, the matched content comprises at least the basic information listed in embodiment 1, i.e. at least the destination, arrival time, gender, age of the user. Preferably, the tourism needs at least meeting the four basic information of the destination, the arrival time, the gender and the age of the user are the second user tourism needs to be matched. Preferably, the remote service platform obtains the reference user preference information sets that do not match each other and that include the first user preference information based at least in part on a similarity of the second user preference information set and the first user preference information set. Under the condition that the tourism needs are the same, the preference information of different users is different, through the setting mode, a reference information set which belongs to the first user preference information and does not belong to the second user preference information can be obtained, and a plurality of reference tourism schemes at least partially matched with the reference user preference information set are generated according to the first text information and the second text information, so that the tourism preferences of different tourists can be predicted by the tourists adopting similar behaviors, the problem that the recommended tourism route cannot well meet the individual needs of different tourists is solved, namely the tourism scheme which is exclusive to a certain user is found on the basis of the similar user preferences, the preferences of the user can be well conformed, and the individual needs of the user can also be met.

Preferably, the plurality of reference travel solutions are recommended to the first user in a manner that is prioritized over a plurality of first travel solutions of the first user. For example, the first plurality of travel itineraries and the reference plurality of travel itineraries are numbered in such a manner that the priority of the reference plurality of travel itineraries is 01, 02, 03 … … above the priority of the first plurality of travel itineraries, wherein 01 corresponds to the highest priority.

Preferably, thecloud service platform 10 optimizes the first user's preference information set based on the first user's selection of the plurality of reference travel itineraries and the plurality of first travel itineraries. Through the setting mode, the preference information of the user can be obtained again according to the feedback result of the reference travel scheme and the first travel scheme of the user, and the preference information of the user can be dynamically corrected through optimizing the updated first user information preference set through the preference information.

According to a preferred embodiment, the method further comprises: thecloud service platform 10 constructs a map including at least a user, a set of user preference information, and a travel solution in an associated manner by at least one of the number ofcomputing devices 20 of the first user and the number ofcomputing devices 30 of the second user. The corresponding travel itineraries and the corresponding set of user preference information are presented on the map with associated annotations. The corresponding preference weight index is associatively labeled at the corresponding travel itinerary. Preferably, the preference weight index represents a preference weight of the user's selection of the corresponding travel itinerary. Thecloud service platform 10 provides the modification authority of the preference weight index for the user, so that the user can modify the preference weight index within a preset modification range. Preferably, the user is provided with a modification right to the preference weight index because the user selects a travel itinerary that deviates from the user's actual preferences, albeit sometimes for special reasons when selecting for a travel itinerary. For example, the user may make the selection due to operator error or under casual circumstances, which is not a real preference of the user under the travel plan. However, in this case, if thecloud service platform 10 performs the analysis, the set of user preference information selected by the travel plan may deviate from the actual set. At this point, the user should be provided with an opportunity to correct. However, when the user corrects the travel plan, if the data is only the form data, the first user may have difficulty in reproducing the current travel plan in mind. Preferably, the travel itinerary is composed of a number of first travel itineraries and a number of reference travel itineraries. Therefore, according to the method and the device, the first user can rapidly and accurately judge the relevance between the user preference information set and the corresponding travel scheme again by means of the user preference information set and the corresponding travel scheme marked on the mapping chart, and the preference weight index is artificially modified by combining the understanding of the preference of the first user, so that the generated travel scheme can more accurately accord with the preference of the user.

Example 3

The embodiment discloses a travel scenario generation system, which is suitable for executing the steps of the method described in the invention to achieve the expected technical effect. The preferred embodiments of the present invention are described in whole and/or in part in the context of other embodiments, which can supplement the present embodiment, without resulting in conflict or inconsistency.

A travel scheme generation system based on user preferences includes at least acloud service platform 10 and a number ofcomputing devices 20 of a first user in communication with each other. The module schematic diagram is shown in FIG. 1. Thecloud service platform 10 is configured to:

at least one travel demand input by a first user is received by at least one of the number ofcomputing devices 20 of the first user.

In response to the at least one travel need, at least one first textual information related to the at least one travel need is pushed that is capable of being acquired and recorded by at least one of the number ofcomputing devices 20 of the first user. Preferably, the first text message may be the travel text message obtained in embodiment 1.

A first set of user preference information of a first user is acquired in response to a specific operation of the first user with respect to first text information. Preferably, the specific operation includes a series of feedback operations such as comment, like, inquiry, share and the like of the user on the first text information. Preferably, positive or negative feedback of the first text information by the first user can be obtained by the comment. For example, the comment fed back by the user is searched for keywords representing positive tendency and negative tendency to obtain the potential preference tendency of the first text information of the first user. Preferably, the positive tendency keywords may include "thank you", "good you", and the like. The negative-oriented keywords may include words such as "what is," less eaten, "and" less tasted ". Preferably, the user can know that the user potentially prefers the information contained in the first text information through feedback of the behavior of praise, wind direction and the like. Preferably, the user is interested in a certain keyword in the first text information by querying the behavior feedback, and the potential preference of the first user for the certain keyword is obtained by retrieving the keyword of the query content. Through this mode of setting, can not obtain user's potential preference information, can arrive new user's action information through the potential preference information that obtains and tourism demand moreover, can alleviate the cold start problem, also can avoid the appearance of cold start problem to a certain extent.

When the first user preference information set data is sparse, user preference information matching each other is at least partially obtained based on the similarity of the preference information set of the at least one second user and the first user preference information set to update the first user preference information set. Preferably, the second user is a user other than the first user who requested the generation of the travel itinerary. The second user may be a user that has historical data and has requested that a travel itinerary be generated. Preferably, under the condition that the first user is irrelevant to the travel demands of other users, the first user preference information set and the other user preference information sets are compared and analyzed, and data with tendencies of potential user preference information can be mutually extracted, so that the first user preference information set and the other user preference information sets can be enriched. Therefore, the data sparseness problem can be solved by acquiring data of other users having the same preference information in the case where the preference information set data of the user is insufficient.

A number of first travel itineraries are generated for the at least one first textual information that at least partially match the updated first set of user preference information. Preferably, the topic keywords included in the at least one first text message and the feature words in the domain are retrieved according to the first user preference information set derived interest keywords. A plurality of first travel plans are generated according to the number of words which are the same with each other and the occurrence frequency. The number and frequency of vocabularies can be set according to prior information or manually. The generated first plurality of travel solutions may be arranged in a multi-from-at-least-place manner based on the number of matches of the keyword of interest.

According to a preferred embodiment, thecloud service platform 10 is further configured to: the text information of at least one second user matching the travel needs and the second user preference information set are obtained in response to the travel needs of the first user.

Reference user preference information that does not match each other and includes the first user preference information is obtained at least partially based on a similarity of the first user preference information set and the other user preference information sets. Under the condition that the tourism needs are the same, the preference information of different users is different, through the setting mode, a reference information set which belongs to the first user preference information and does not belong to the second user preference information can be obtained, and a plurality of reference tourism schemes at least partially matched with the reference user preference information set are generated according to the first text information and the second text information, so that the tourism preferences of different tourists can be predicted by the tourists adopting similar behaviors, the problem that the recommended tourism route cannot well meet the individual needs of different tourists is solved, namely the tourism scheme which is exclusive to a certain user is found on the basis of the similar user preferences, the preferences of the user can be well conformed, and the individual needs of the user can also be met.

And generating a plurality of reference travel schemes at least partially matched with the preference information of the reference users aiming at the first text information and the second text information of other users. Through the setting mode, the reference information set which belongs to the first user preference information and does not belong to the second user preference information can be obtained, and a plurality of reference tourism schemes which are at least partially matched with the reference user preference information are generated according to the first text information and the second text information, so that the tourism preferences of different tourists can be predicted by the tourists adopting similar behaviors, the problem that the recommended tourism routes cannot well meet the personalized requirements of different tourists is solved, namely, the tourism schemes which are exclusive to a certain user are found on the basis of the similar user preferences, the preferences of the user can be well met, and the personalized requirements of the user can also be met.

It should be noted that the above-mentioned embodiments are exemplary, and that those skilled in the art, having benefit of the present disclosure, may devise various arrangements that are within the scope of the present disclosure and that fall within the scope of the invention. It should be understood by those skilled in the art that the present specification and figures are illustrative only and are not limiting upon the claims. The scope of the invention is defined by the claims and their equivalents.

Claims (10)

1. A travel itinerary generation method based on user preferences, the method comprising a cloud service platform (10) receiving at least one travel requirement input by a first user via at least one of a number of computing devices (20) of the first user, the method further comprising the cloud service platform (10) performing the steps of:

in response to at least one of the travel needs, pushing at least one first textual information relating to at least one of the travel needs that is capable of being retrieved and recorded by at least one of the number of computing devices;

acquiring a first user preference information set of the first user in response to a specific operation of the first user on the first text information, and generating a plurality of first travel schemes at least partially matching the first user preference information set for at least one first text information;

optimizing a first user preference information set based on the selection result of the first user about the plurality of first travel schemes, and storing the optimized first user preference information set in the cloud service platform in a mode that a second user can match and retrieve at least according to user preferences.

2. The travel itinerary generation method according to claim 1, wherein the cloud service platform (10) obtains, through several computing devices (30) of the second user, at least one second text message related to the travel needs of the second user;

the cloud service platform (10) acquires a second user preference information set of the second user in response to a specific operation of the second user on the second text information, and at least partially obtains user preference information matched with each other based on the similarity of the second user preference information set and the first user preference information set so as to update the second user preference information set and/or the first user preference information set.

3. The travel itinerary generation method according to claim 1 or 2, characterized in that the cloud service platform (10) generates, for the second text information, a number of second travel itineraries matching at least partially the updated second set of user preference information;

the cloud service platform (10) optimizes a second user preference information set based on the selection result of the second user on the plurality of second travel schemes, and the optimized second user preference information set is stored in the cloud service platform (10) according to a mode that the first user and a third user can match and retrieve at least according to user preferences.

4. A travel itinerary generating method according to one of the preceding claims, wherein the cloud service platform (10) further performs the following steps:

obtaining second text information and a second user preference information set of a second user matching the travel needs of the first user in response to the travel needs;

obtaining, based at least in part on similarity of a first set of user preference information of the first user and the second set of user preference information, a reference set of user preference information that do not match each other and that includes the first user preference information;

generating a number of reference travel solutions for the first and second textual information that at least partially match the set of reference user preference information.

5. The travel itinerary generation method according to one of the preceding claims, wherein the reference travel itineraries are recommended to the first user with a priority greater than the first travel itineraries of the first user, and the cloud service platform (10) optimizes the first user's preference information set based on the first user's selection of the reference travel itineraries and the first travel itineraries.

6. The travel itinerary generating method according to one of the preceding claims, further comprising:

the cloud service platform (10) constructs a mapping chart at least comprising users, user preference information sets and travel schemes in an associated mode through at least one of the computing devices (20) of the first user and the computing devices (30) of the second user, corresponding travel schemes and corresponding user preference information sets are labeled on the mapping chart in an associated mode, corresponding preference weight indexes are labeled at the corresponding travel schemes in an associated mode and represent preference weights selected by the users for the corresponding travel schemes, and the cloud service platform (10) provides modification permission of the preference weight indexes for the users so that the users can modify the preference weight indexes within a preset modification range.

7. A personalized travel text information pushing method based on semantic understanding is characterized by comprising the following steps of:

generating at least one topic keyword comprising at least one sequence semantic based on information of the user's travel needs;

acquiring at least one initial text message based on at least one topic keyword associated user corpus content;

and dividing at least one piece of initial text information based on a limited domain formed by each topic keyword and the related intra-domain feature words thereof to extract corresponding travel text information meeting travel requirements in a mode of removing irrelevant high-frequency words and recommend the travel text information to users.

8. The method for pushing travel text information according to claim 7, further comprising:

the cloud service platform (10) generates the sentence of the travel text information in a mode of extracting the initial text information along the continuous direction of the semantics of the in-domain feature words of the limited domain edge by taking the in-domain feature words of the limited domain edge as a starting point.

9. A travel itinerary generation system based on user preferences, the system comprising at least a cloud service platform (10) and a number of computing devices (20) of a first user in communication with each other, characterized in that the cloud service platform (10) is configured to:

receiving, by at least one of a number of computing devices (20) of a first user, at least one travel demand input by the first user;

in response to at least one of said travel needs, pushing at least one first textual information relating to at least one of said travel needs that is capable of being retrieved and recorded by at least one of said first user's number of computing devices (20);

acquiring a first user preference information set of the first user in response to a specific operation of the first user with respect to the first text information;

when the first user preference information set data is sparse, at least partially obtaining user preference information matched with each other based on the similarity of the preference information set of at least one second user and the first user preference information set so as to update the first user preference information set;

a number of first travel itineraries are generated for at least one of the first textual information that at least partially match the updated first set of user preference information.

10. The travel itinerary generation system of claim 9, wherein the cloud service platform (10) is further configured to:

obtaining second text information and a second user preference information set of at least one second user matching the travel needs of the first user in response to the travel needs;

obtaining a reference user preference information set that does not match each other and that includes the first user preference information based at least in part on similarities of the first user preference information set and a second user preference information set;

generating a number of reference travel solutions for the first and second textual information that at least partially match the set of reference user preference information.

Images