Mining method and device for point-of-interest data

[0080] Method embodiment one

[0081] Reference figure 1 , Shows a step flow chart of Embodiment 1 of a method for mining interest point data of the present invention, which may specifically include the following steps:

[0082] Step 101: Obtain input content generated by the user in the preset platform of the smart terminal;

[0083] Step 102: dig out target points of interest from the input content;

[0084] Step 103: Obtain geographic location information of the target point of interest.

[0085] In the embodiment of the present invention, the aforementioned smart terminal specifically includes but is not limited to: smart phones, tablet computers, e-book readers, MP3 (Moving Picture Experts Group AudioLayer III) players, MP4 (Moving Picture Experts Group AudioLayer III) players, Video experts compress standard audio level 4, Moving Picture Experts Group Audio Layer IV) players, laptop portable computers, car computers, desktop computers, set-top boxes, smart TVs, wearable devices, etc.

[0086] In an optional embodiment of the present invention, input environment information corresponding to the aforementioned input content may also be obtained. The input environment information includes: at least one of context environment information, location environment information, time environment information, and platform environment information One kind. Wherein, for one or multiple input content, the context environment may include: the one or multiple input content corresponds to a series of complete input content of the user. For example, for input content generated by an instant messaging platform, the context environment corresponding to the input content may be the corresponding communication content of the user within the time period corresponding to the input content. For example, if the user wants to go or has gone to a POI, The content of the corresponding POI will be shared with one or more users. Therefore, the content of the communication between it and one or more users can be used as the context of a certain input. For another example, with the development of mobile Internet technology and communication technology, a user usually generates multiple inputs related to a certain POI. For example, when the user intends to go to a POI (that is, before going to a POI), it may use the navigation platform, The travel platform or instant messaging platform generates one or more input content related to the POI; further, after reaching the POI, the user may also generate one or more input content related to the POI through the instant messaging platform or social platform in order to Moments share the information of the POI; and, after the user leaves the POI, it is possible to generate travel content related to the POI through the travel platform, etc. Therefore, the user generates multiple inputs related to a certain POI within a preset time period The content is relevant, so it can be used as context for each other.

[0087] Optionally, the above-mentioned smart terminal may be a terminal with a positioning function. In this way, the position environment information corresponding to the above-mentioned input content may be obtained through the smart terminal. The smart terminal may use GPS (Global Positioning System, Global Positioning System), IP ( The Internet Protocol (Internet Protocol) and other technologies realize the positioning function.

[0088] In the embodiment of the present invention, the user may include an Internet user. In an optional embodiment of the present invention, when an Internet user generates input content in a preset platform, the preset platform can generate a corresponding log, and the embodiment of the present invention can collect logs from the foregoing preset platform to obtain The input generated by the user in the preset platform.

[0089] In practical applications, the privacy of the log and other factors cause some preset platforms to not open their query logs. In view of the above-mentioned problems, in another optional embodiment of the present invention, the user's status can be obtained from the input method log. The input content generated in the preset platform. Among them, as a boarding program, the input method program can be hosted in any host program, so the input method program can monitor the current host program environment where it is. If the current host program environment is the environment of the preset platform or the search of the preset platform In the box environment, the user's input in the search box can be recorded as input to the input method log. Optionally, the input method program filters the hosted host program, and filters applications whose occurrence probability of POI-related input content is lower than the probability threshold from the above host program to obtain preset applications. For example, the occurrence probability of POI-related input content in game applications or office applications is low, so it can be excluded from the preset applications. It is understood that those skilled in the art can determine the above probability according to actual application requirements For the value of the threshold, the embodiment of the present invention does not limit the specific probability threshold.

[0090] Optionally, the input method of the aforementioned input content may include a keyboard input method, a handwriting input method, or a voice input method, etc. That is, the input method program can capture the input content input by the user in any manner and record it.

[0091] The method for mining points of interest in the embodiment of the present invention may be executed by the client and/or the server. Among them, when the client executes the point of interest mining method of the embodiment of the present invention, the client can obtain the input content generated by the user in the preset platform of the smart terminal in real time, and perform step 102 and step 103 to mine and obtain target points of interest. Geographic location information. When the server executes the point-of-interest mining method of the embodiment of the present invention, the client can periodically or instantly send the input content generated by the user in the preset platform of the smart terminal to the server, and the server can send to the client by executing steps 102 and 103 The input content of the smart terminal is processed immediately, or the server can record the input content sent by the client to the log, and periodically obtain the input content generated by the user in the preset platform of the smart terminal from the log through step 101, and then perform step 102 And step 103 digs and obtains the geographic location information of the target point of interest. It can be understood that the embodiments of the present invention do not impose restrictions on specific execution subjects.

[0092] Step 102 can dig out target points of interest for one or more input content obtained in step 101. Among them, the input content for one or more times of mining can be input content generated by the same user or input content generated by multiple users; the multiple users can be multiple users corresponding to instant messaging, or the same group or the same Multiple users associated with the topic.

[0093] In practical applications, step 102 uses any semantic understanding technology to dig out target points of interest from one or more input content. Optionally, the step of digging out target points of interest from the input content may include: identifying a target point of interest scene corresponding to the input content; and digging out from the input content according to the target point of interest scene Target points of interest. Since the aforementioned POI scenarios can be used to indicate scenarios where POI-related input content is likely to be generated, if the target POI scene corresponding to the input content cannot be identified, it can be explained that the input content is not related to the POI. Therefore, the embodiment of the present invention can pass the target The point-of-interest scene recognition filters out a large amount of input content that is not related to POI, and improves the efficiency of POI mining. In addition, since the input content in a specific POI scene usually has a specific law, that is, the input content in a specific POI scene is relatively convergent, so according to the law of the input content of the target POI scene, from the input content Extracting the POI of the target POI can improve the accuracy of POI mining.

[0094] The embodiment of the present invention may provide the following recognition methods for recognizing the target interest point scene corresponding to the input content:

[0095] Identification method 1. Determine the target point of interest scene corresponding to the input content according to the input environment information corresponding to the input content; and/or

[0096] Identification method 2: Classify the input content and the surrounding interest points corresponding to the location environment information to obtain the target interest point scene corresponding to the input content.

[0097] Among them, the recognition method 1 can analyze the input environment information corresponding to the input content to obtain the corresponding target POI scene. For example, the context information corresponding to the input content may be analyzed to determine which of the preset POI scenes the target POI scene corresponding to the input content belongs to. For another example, the target POI scene to which the input content of multiple users belongs under the location environment information corresponding to the current input content can be analyzed to obtain the target POI scene corresponding to the current input content.

[0098] Or, the identification method 1 can also search in the mapping relationship between the preset platform or the preset platform category and the preset POI scene according to the platform environment information to obtain the corresponding target POI scene.

[0099] In an optional embodiment of the present invention, the process of acquiring the foregoing preset POI scene may include: analyzing and obtaining the corresponding preset POI scene for a preset platform that may generate POI-related input content. For example, map application A, map application B, map application C, map application D, map application E and other map category preset platforms can be analyzed to obtain the corresponding navigation scene; another example can be targeted at e-commerce platform 1. Pre-installed platforms for shopping categories such as platform 2, analyze the corresponding shopping scenes, etc. Further, a mapping relationship between a preset platform or a preset platform category and a preset POI scene can also be established to facilitate the determination of the POI scene corresponding to the input content.

[0100] In another optional embodiment of the present invention, the acquisition process of the preset POI scene may include: acquiring input content samples and surrounding interest point information corresponding to its location environment information, and obtaining the preset according to the above surrounding interest point information POI scene. Wherein, the distance between the above-mentioned surrounding interest points and the location environment information corresponding to the input content sample may be less than a distance threshold. Optionally, the distance threshold may be 50m, 100m, 200m, etc. The embodiment of the present invention has no specific distance threshold. Be restricted.

[0101] The recognition method 2 can realize the recognition of the target interest point scene corresponding to the input content by the method of machine classification. Since the POI scene to which the input content of a specific peripheral interest point belongs usually has a specific rule, the above rule can be obtained through machine learning.

[0102] In an optional embodiment of the present invention, the input content sample and the surrounding interest point information corresponding to its location environment information can be obtained, and the scene of the input content sample can be manually marked according to the surrounding interest point information to obtain the preset POI Scene; and train the input content sample to train a scene classifier, and use the scene classifier to classify the input content and surrounding points of interest in real time to obtain the corresponding target POI scene.

[0103] The embodiment of the present invention can provide the following mining solution for mining a target interest point from the input content according to the target interest point scene:

[0104] Mining plan 1,

[0105] In technical solution 1, the step of digging out a target interest point from the input content according to the target interest point scene may include: determining a target interest point sentence corresponding to the input content according to the target interest point scene Mode; according to the target interest point sentence pattern, extract the interest point feature words of the target interest point from the input content.

[0106] In practical applications, the input content generated by the user through the preset platform is diverse, including POI-related input content, and also a large number of POI-unrelated input content. Technical solution 1 can filter out the input content irrelevant to the POI from the input content according to the point of interest sentence pattern, and can obtain the input content related to the POI and the target point of interest sentence pattern corresponding to the input content related to the POI.

[0107] In the embodiment of the present invention, sentences are also called sentences, which are the basic unit of language use. Usually sentences are composed of words and phrases (phrases), which can express a complete meaning, such as telling someone something, asking a question, expressing a request or Stop, express a certain feeling, express the continuation or omission of a paragraph, etc. Sentence patterns can define the rules corresponding to characteristic information such as the part of speech, word order and position of the words contained in the sentence.

[0108] In the embodiment of the present invention, the point-of-interest sentence pattern can be used to represent the pattern of the sentence related to the POI and containing the characteristic words of the POI. In the embodiment of the present invention, the POI feature word can be used to identify the POI, which may specifically include, but is not limited to, features such as roads, buildings, and company names corresponding to the administrative division of the POI. For example, common points of interest sentence patterns can include: "I want to go to XXX", "I plan to go to XXX", "How to go to XXX", "How to go from *** to XXX", "XXX to open", "XXX has activities "?", "How to get to XXX", etc.; among them, "XXX" can be used to indicate the preset character string corresponding to the POI feature word, and "XXX" can also be replaced by character strings such as "###". It can be understood that the implementation of the present invention For example, there is no restriction on the specific preset character string corresponding to the POI feature word.

[0109] Since the point-of-interest sentence pattern of the embodiment of the present invention can be defined as: the word order, position and other characteristic information of the preset feature words, the embodiment of the present invention can extract the target interest from the input content according to the determined target point-of-interest sentence pattern Point of interest feature words. For example, if the input content is "Go to the headquarters of CCTV", the input content corresponds to the target point of interest sentence pattern "Go to XXX", so it can be based on the preset feature words contained in the target point of interest sentence pattern "Go to XXX" From the input content, extract the POI feature word "CCTV Headquarters Building".

[0110] The embodiment of the present invention may provide the following determination solution for determining the target interest point sentence pattern corresponding to the input content according to the target interest point scene:

[0111] Determine plan 1,

[0112] In the determination scheme 1, the step of determining the target point of interest sentence pattern corresponding to the input content may include: matching the sentence pattern corresponding to the input content with a preset point of interest sentence pattern corresponding to the target point of interest scene, The matched preset interest point sentence pattern is used as the target interest point sentence pattern corresponding to the input content.

[0113] In practical applications, the preset interest point sentence pattern can be obtained in advance. Corresponding pre-acquisition methods may include: setting preset POI scenes (such as navigation scenes, shopping scenes, travel scenes, business trip scenes, etc.) corresponding to POI-related input content that may generate POI-related input content, and checking the input content in the aforementioned preset POI scenes Perform analysis to get the corresponding preset interest point sentence pattern.

[0114] Since POI scenarios can be used to represent scenarios where POI-related input content is likely to be generated, the point-of-interest sentence patterns in specific POI scenarios usually have specific rules, that is, the sentence patterns of input content in specific POI scenarios are relatively convergent Therefore, the input content in a specific POI scene can be analyzed and summarized to obtain the corresponding preset interest point sentence pattern. For example, the preset point of interest sentence mode in the navigation scene may include: "I want to go to XXX", the preset point of interest sentence mode in the shopping scene may include: "I am shopping in XXX", the preset point of interest sentence in the travel scene The mode may include: "XXX N day trip", where N is a positive integer, and the preset point of interest sentence mode in the business trip scene may include: "to XXX business trip tomorrow" and so on.

[0115] The embodiment of the present invention uses the point-of-interest sentence pattern corresponding to the target POI scene as the preset point-of-interest sentence pattern corresponding to the aforementioned input content, so that the preset point-of-interest sentence pattern and matching accuracy can be realized.

[0116] In practical applications, the sentence pattern corresponding to the input content can be matched with the preset point of interest sentence pattern corresponding to the target point of interest scene. If the matching is successful, it can be determined that the input content is related to the POI, otherwise, if it matches If it fails, it can be determined that the input content is not related to the POI. Optionally, the process of matching the sentence corresponding to the input content with the preset point of interest sentence pattern may include: respectively analyzing the corresponding sentence of the input content and the preset point of interest sentence pattern to obtain the corresponding first sentence structure and first sentence structure. Second sentence structure, for example, the first sentence structure or the second sentence structure can include: subject, predicate, object and other structural components, then the first sentence structure and the second sentence structure can be compared. If the comparison is successful, The vocabulary corresponding to each structural component can be compared, and if the comparison is successful, the matching can be considered successful. Of course, the above matching process is only an optional embodiment. In fact, those skilled in the art can adopt other matching processes according to actual application requirements, for example, the first string of the sentence corresponding to the input content and the preset query The second character string of the sentence pattern is compared. The embodiment of the present invention does not limit the specific process of matching the sentence corresponding to the historical query string with the preset query sentence pattern.

[0117] Optionally, in the process of analyzing the corresponding sentence structure of the input content to obtain the corresponding first sentence structure, a word segmentation dictionary may be used to segment the input content to obtain the corresponding first vocabulary, and perform word segmentation on the first vocabulary. The part-of-speech tagging is used to obtain the first part-of-speech of the first vocabulary, and further obtain the first sentence structure for the part-of-speech, word order, position and other information of the first vocabulary. As for the acquisition process of the second sentence structure, since it is similar to the acquisition process of the first sentence structure, it will not be repeated here, and cross-referencing is sufficient.

[0118] Determine plan 2,

[0119] In the determination scheme 2, the step of determining the target interest point sentence pattern corresponding to the input content may include: classifying the sentence pattern corresponding to the input content according to the target interest point scene, and using the classification result as the Describe the target point of interest sentence pattern corresponding to the input content.

[0120] Determining scheme 1 The target interest point pattern is determined based on the matching of sentence patterns. When the matching degree between the input content corresponding sentence pattern and the preset interest point sentence pattern is high, the matching is considered successful, and the corresponding target interest point sentence can be obtained. Pattern; and when the matching is unsuccessful, the corresponding target interest point sentence pattern will not be obtained. For example, suppose the preset point of interest sentence pattern 1 is "Go to XXX", and if the input content 1 is "Go to Wudaokou", it is considered that the input content 1 matches the preset point of interest sentence pattern 1 successfully; and if the input content 2 If it is "plan to go to Wudaokou", it is considered that the input content 2 matches the preset interest point sentence pattern 1 unsuccessfully. It can be seen that determining the success rate of obtaining the scheme 1 depends on the comprehensiveness of the preset interest point sentence pattern, so the success rate of obtaining the target interest point sentence pattern cannot be guaranteed.

[0121] In view of the problem that the determination plan 1 cannot guarantee the success rate of the target point of interest sentence pattern, the determination plan 2 can classify the sentence pattern corresponding to the input content, and determine the difference between the sentence pattern corresponding to the input content and the preset point of interest sentence pattern. When the matching degree is low, the classification method can also obtain the classification result of the sentence pattern corresponding to the input content based on the structure of the sentence pattern, so that the success rate of obtaining the sentence pattern of the target point of interest can be improved.

[0122] In practical applications, the sentence pattern classifier can be used to classify the sentence pattern corresponding to the input content, wherein the sentence pattern classifier can select one of multiple preset interest point sentence patterns corresponding to the target interest point scene As a result of classification, different target interest point scenes can have different sentence pattern classifiers. The training process of the sentence pattern classifier may include: obtaining input content samples, using the aforementioned preset POI scene to mark the input content samples with sentence patterns, and using machine learning methods to mark the sentence patterns according to the input content samples and sentence pattern labeling results The classifier is trained to obtain a sentence pattern classifier with the ability to recognize sentence patterns. The sentence pattern classifier can be used to describe the correspondence between the feature vector and the sentence pattern of the target point of interest, where the feature vector can include: the part of speech of the sentence , Word order, location and other characteristics. In the classification process, the feature information of the sentence corresponding to the input content can be used as the input vector and input to the sentence pattern classifier, and the sentence pattern classifier outputs the corresponding classification result, which can be used as the target interest point sentence pattern.

[0123] Optionally, the type of sentence pattern classifier may include: SVM (Support Vector Machine, Support Vector Machine), Bayes, KNN (K-Nearest Neighbor algorithm), etc. The embodiment of the present invention is for The specific type of the intent classifier is not limited.

[0124] Mining plan 2,

[0125] The mining scheme 2 can mine the target interest point from the input content according to the semantic understanding rules in the target interest point scene.

[0126] Since the input content mode in a specific POI scene is relatively convergent, so that the mining of POI in a specific POI scene has a law to follow, so the semantic understanding rules in the target point of interest scene can be used to mine the input content Target points of interest to improve the accuracy of POI mining.

[0127] Optionally, the above step of digging out the target point of interest from the input content according to the target point of interest scene may include:

[0128] Extract the target interest point from the input content in the first target interest point scene; and/or

[0129] Determine whether the consecutive multiple input content within the preset time period in the second target point of interest scene meets the preset correction input condition, if so, select the target input content from the consecutive multiple input content, and select the target input content from the target Extract the target points of interest from the input content.

[0130] In a first target point of interest scenario such as a shopping scene or a business trip scene, the user's input content is in line with the user's input intention, so the target point of interest can be extracted from the first input content in the first target point of interest scenario. For example, if the input content is "I'm shopping in XXX" or "I'm here XXX" for a business trip, it means that the input content contains the POI that needs to be input, so "XXX" can be extracted from it as the target POI.

[0131] In a second target point of interest scene such as a navigation scene, the user may continuously input multiple input content within a preset time period to obtain the desired navigation result. For example, the user’s first input of content in the preset time period is "Hengdian Building", and after not getting the required navigation results, the user’s re-entering content is "Yard 4, Wangjing East Road, Chaoyang District, Beijing", and the result is The required navigation result, so no new input content is generated. Therefore, it can be considered that the first input content and the re-input content meet the preset correction input conditions. Therefore, the target input content can be selected from the continuous multiple input content and input from the target The content extracts the target POI "Yard No. 4 Wangjing East Road, Chaoyang District" or the building corresponding to "Yard No. 4 Wangjing East Road, Chaoyang District". For example, the building may include "Hengdian Building" and so on. It is understandable that any semantic understanding technology can be used to select the target input content from the multiple consecutive input content. For example, the target POI scene corresponding to a certain input content can be recognized first, and if the recognition is successful, it will be used as the target input If the recognition fails, the input content is discarded. The embodiment of the present invention does not limit the specific selection process of the target input content.

[0132] In a second target interest point scenario such as a navigation scene, it is also possible to extract corresponding candidate interest points from multiple input content within a preset time period in the second target interest point scene, and output the candidate interest points to the user Click the first prompt information, and determine the target point of interest according to the user's first response operation to the first prompt information. When the number of candidate POIs is 1, the corresponding first prompt message can be "Have you finally reached candidate POI1?"; when the number of candidate POIs is greater than or equal to 2, the corresponding first prompt message can be "You In the end is the candidate POI2 or the candidate POI3?” In this way, according to the user's first response operation, it can be determined whether the candidate POI1 is the POI required by the user, or which of the candidate POI2 and the candidate POI3 is the user The required POI can improve the accuracy of the target POI.

[0133] It can be understood that those skilled in the art can obtain any semantic understanding rule corresponding to POI mining in a specific POI scenario through analysis. It can be understood that the embodiment of the present invention is based on the semantic understanding rule in the target interest point scenario, from the input The specific process of digging out target points of interest in the content is not restricted.

[0134] The above has explained the process of digging out target points of interest from the input content according to the target point of interest scenario through the mining scheme 1-mining scheme 2. It is understandable that those skilled in the art can use mining according to actual application requirements. Scheme 1-Any one or combination of mining scheme 2, or other mining schemes that dig out target points of interest from the input content may also be used, and the embodiment of the present invention does not limit specific mining schemes.

[0135] Since geographic location information is important information of the POI, it can determine the location of a POI on the map. Therefore, after the target POI is obtained by mining in step 102, the geographic location information of the target POI needs to be obtained in step 103.

[0136] The geographic location information of the target POI can be obtained in multiple ways. For example, the geographic location information of the target POI can be obtained manually.

[0137] In an optional embodiment of the present invention, the location environment information corresponding to the input content may be obtained in advance; then, the step of obtaining geographic location information of the target point of interest may include: The geographic location information of the target point of interest is excavated from the location environment information corresponding to the relevant input content.

[0138] With the development of terminal technology, there are more and more smart terminals with positioning functions. Among them, smart terminals can use GPS, IP and other technologies to achieve positioning functions. Because when the user generates POI-related input content, his smart terminal may be in the geographic location environment corresponding to the POI, so the embodiment of the present invention digs out the location environment information corresponding to the input content related to the target point of interest. The geographic location information of the target point of interest can save the labor cost required for obtaining the geographic location information of the target POI, and improve the efficiency of obtaining the geographic location information of the target POI.

[0139] In practical applications, the step of digging out the geographic location information of the target interest point from the location environment information corresponding to the input content related to the target interest point may include: extracting the input content related to the target interest point From the corresponding location environment information, the location environment information matching the target point of interest is extracted as the location information of the target POI.

[0140] Optionally, the foregoing process of extracting geographic location environment information matching the target point of interest from the location environment information corresponding to the input content related to the target point of interest may include: obtaining a rough location range of the target POI, Determine whether the location environment information corresponding to the input content of the target POI matches the rough location range. For example, if the target POI is "China World Central Television Headquarters Building", the location range of "China World Trade Center" can be used as the rough location range first, and then the location environment information corresponding to the input content of the target POI can be matched with the rough location range. For example, user A has generated input content related to the target POI twice. The first input content "I want to go to China World Trade Center Central Television Headquarters" was generated on the map platform in Haidian District, and the second input content "I went to CCTV "Headquarters building" is generated through the instant messaging platform in Chaoyang District, you can match the location environment information corresponding to each input with the last rough location range, and get the second input that matches the rough location range The geographic location environment information corresponding to the content is taken as the geographic location information of the target POI.

[0141] Optionally, the above process of extracting geographic location environment information that matches the target interest point from the location environment information corresponding to the input content related to the target interest point may include: input content related to the target interest point Perform semantic analysis to determine whether the geographic location environment information corresponding to the input content is the geographic location information of the target POI, that is, determine whether the user is in the geographic location of the POI when the input content is generated. For example, if the input content is "I'm in XXX", "I'm in XXX", it means that the user is in the geographic location of POI "XXX" when the input content is generated; for another example, if the input content is "I want to go to XXX", " I plan to go to XXX" and "I go to XXX tomorrow", it means that the geographic location of POI "XXX" is not the place when the input is generated, so the accurate geographic location information of POI "XXX" can be extracted.

[0142] In an optional embodiment of the present invention, the method of the embodiment of the present invention may further include: increasing the credibility of the target interest point when multiple input content hits the target interest point; Reliability is used as a basis for whether to adopt the target point of interest. Here, the multiple input content may be input content generated by one user or multiple users. For example, if a target POI with the same location in the same city is mined in M ​​input content, the credibility of the target POI is +M. In the embodiment of the present invention, the credibility can be used as a basis for whether to adopt the target POI. For example, if the credibility of the target POI is greater than the credibility threshold, it can be accepted and added to the POI database. The database can be used to store information such as the name of the POI. Optionally, the POI in the POI database can be displayed on a map. For example, the POI corresponding to the target location can be obtained from the POI database according to the target location of the user or the target location corresponding to the user’s search term. And display the POI corresponding to the target location on the map.

[0143] In another optional embodiment of the present invention, the method of the embodiment of the present invention may further include: outputting second prompt information of the target point of interest to the user; and according to the user's second response to the second prompt information Operate to determine the authenticity of the data of the target point of interest. Wherein, the second prompt information may include information such as the name, coordinates, address, and category of the target point of interest. If the user is satisfied with the second prompt information, the generated second response operation may be a confirmation operation. In this case Below, the authenticity of the data of the target point of interest is True. On the contrary, if the user is not satisfied with the second prompt information, the second response operation generated can be a denial operation. In this case, the The authenticity of the data of the target interest point is False.

[0144] It should be noted that the first prompt information or the second prompt information in the embodiment of the present invention may be provided by a preset platform that generates input content corresponding to the POI, or may be provided by other platforms. For example, if the input generated by the user through the instant messaging platform hits the POI, the instant messaging platform can provide the first prompt information or the second prompt information, or other platforms such as the input method platform can provide the first prompt information or the second prompt information. Prompt information. In addition, the above-mentioned first prompt information or second prompt information may be output to the user through a pop-up window, a floating layer, etc. The embodiment of the present invention does not limit the specific output mode of the first prompt information or the second prompt information.

[0145] In yet another alternative embodiment of the present invention, in order to avoid the repetition of POIs in the POI database, the target POI obtained in step 102 may be compared with the existing POIs in the POI database to determine whether the target POI is new If the discovered POIs are consistent, it means that the target POI is not a newly discovered POI, and the name and geographic location of the known POI are updated based on the target POI data; if they are inconsistent, the target POI is a new discovery And add it to the POI database.

[0146] In summary, the method for mining point-of-interest data in the embodiment of the present invention can obtain input content generated by a user on a preset platform of a smart terminal, and mine the target POI from the input content, and then obtain the information of the target POI. Geographical location information. Since the embodiment of the present invention can automatically mine the target POI from the user's input content, it can save labor costs and improve the efficiency of POI mining. In addition, the embodiment of the present invention can complete POI mining without the feedback of the client user, and therefore can save the promotion cost corresponding to the feedback interface.

[0147] In addition, in the process of mining the target POI from the aforementioned input content, the embodiment of the present invention may first identify the target interest point scene corresponding to the input content, and then mine out the input content according to the target interest point scene Target points of interest, etc.; since the aforementioned point of interest scenes can be used to indicate scenarios that may produce POI-related input content, if the target point of interest scenes corresponding to the input content cannot be identified, it can indicate that the input content is not related to the POI. The embodiment of the invention can filter out a large amount of input content irrelevant to the POI through the recognition of the target interest point scene, and improve the efficiency of POI mining. In addition, since the input content in a specific POI scene usually has a specific law, that is, the input content in a specific POI scene is relatively convergent, so according to the law of the input content of the target POI scene, from the input content Extracting the POI of the target POI can improve the accuracy of POI mining.

[0148] Method embodiment two

[0149] Reference figure 2 , Shows a step flow chart of the second embodiment of a method for mining interest point data of the present invention, which may specifically include the following steps:

[0150] Step 201: The server obtains the input content generated by the user in the preset platform of the smart terminal and the input environment information corresponding to the input content through the input method log;

[0151] Optionally, the input environment information may include: location environment information, platform environment information, context environment information, time environment information, and so on.

[0152] Step 202: The server identifies the target POI scene corresponding to the input content;

[0153] Step 203: The server determines the target POI sentence pattern corresponding to the input content according to the target POI scenario;

[0154] Step 204: The server extracts the POI feature words of the target POI from the input content according to the target POI sentence pattern;

[0155] Step 205: The server digs out the geographic location information of the target POI from the location environment information corresponding to the input content related to the target POI.

[0156] Step 206: The server generates corresponding second prompt information according to the POI feature words and geographic location information of the target POI, and sends the second prompt information to the client;

[0157] Step 207: The client outputs the second prompt information of the target POI to the user, and receives and sends the user's second response operation to the second prompt information to the server.

[0158] Step 208: The server judges the authenticity of the target POI data according to the second response operation of the user to the second prompt information.

[0159] Optionally, when the authenticity of the target POI is true, the target POI is compared with an existing POI in the POI database, and if they are inconsistent, the target POI is added to the POI database.

[0160] In summary, in the method for mining point of interest data in the embodiment of the present invention, after identifying the target POI scene corresponding to the input content, the target POI sentence pattern corresponding to the input content is obtained according to the target POI scene; due to the point of interest sentence in the specific POI scene The pattern usually has a specific rule, that is, the sentence pattern of the input content in a specific POI scene is relatively convergent, so the target POI sentence pattern corresponding to the input content is obtained according to the target POI scene, and the target POI is extracted from the input content. The POI feature words can improve the accuracy of POI mining.

[0161] In addition, since the aforementioned POI scene can be used to indicate a scene where POI-related input content is likely to be generated, if the target POI scene corresponding to the input content cannot be identified, it can be explained that the input content is not related to the POI, so the embodiment of the present invention can Through the identification of the target interest point scene, a large amount of input content that is not related to POI is filtered out, and the efficiency of POI mining is improved.