Parking reminder device and method

Abstract

The present invention provides a parking reminder method and device. The method includes determining a destination identifier; obtaining at least one message about the destination from at least one medium based on the destination identifier; generating parking lot information around the destination based on map data; and processing the message and the parking lot information to generate a parking convenience reminder about the destination.

Description

Technical Field

[0001] The present invention relates to a technology for obtaining parking lot information, and more particularly to a technology for predicting the parkable state at a destination. Background Art

[0002] With the increase in social vehicles, higher requirements are put forward for parking needs. In particular, parking at destinations such as popular scenic spots is a headache, especially during holidays. However, at the same time, it should also be noted that the problem of difficult parking in some parking lots is also dynamic, with temporality or contingency. Therefore, it is very valuable for users to know the parking state of the destination in advance, which can avoid the time spent looking for parking spaces and even avoid traffic jams. Conventionally, users can search through a map system to determine the parking lot information near the destination, but only providing the location information of the parking lot cannot effectively understand the actual parking state. Summary of the Invention

[0003] The present invention provides a solution that can help users timely understand the parking situation at the destination, thereby providing valuable reference information for users' travel.

[0004] According to one aspect of the present invention, there is provided a parking reminder method, including determining a destination identifier; based on the destination identifier, obtaining at least one message about the destination from at least one medium; generating parking lot information around the destination based on map data; processing the message and the parking lot information to generate a parking convenience reminder for the destination. In some embodiments of the present invention, the medium includes one or more electronic social medias and / or electronic publications, and the message includes one or more messages published by the at least one medium within a certain time range. In addition, in some embodiments, the parking convenience reminder for a certain destination is classified according to date characteristics, for example, it is divided into a holiday parking convenience reminder and a normal parking convenience reminder. Therefore, the obtained messages include messages published on holidays such as weekends and normal messages within a certain time range, such as within 1 month, and the corresponding holiday parking convenience reminder and normal parking convenience reminder are generated respectively based on the messages published on weekends or normal messages.

[0005] According to one aspect of the present invention, there is provided a parking reminder device, including a destination confirmation module for determining the destination identifier to be evaluated; a message scraping module configured to obtain at least one message about the destination from at least one medium based on the destination identifier; and a parking prediction module configured to process the at least one message and the parking lot information around the destination to generate a parking convenience reminder for the destination. Brief Description of the Drawings

[0006] Figure 1 A block diagram schematically showing a parking reminder device according to an example;

[0007] Figure 2 A block diagram schematically showing a parking prediction module according to an example;

[0008] Figure 3A A schematic illustration of an example of a message;

[0009] Figure 3B , 3C A schematic illustration of an example of a prediction result;

[0010] Figure 4 A block diagram schematically showing a situation awareness module;

[0011] Figure 5 A process flow diagram schematically showing a parking reminder method according to an example;

[0012] Figure 6 A process flow diagram schematically showing message processing. Detailed implementation manners

[0013] The methods and devices provided in the embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Although the preferred embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure will be more thorough and complete, and can fully convey the scope of the present disclosure to those skilled in the art.

[0014] The parking reminder device and method according to the present invention can predict the parking situation of a destination by learning the parking lots and social messages of the destination. In one implementation manner of the present invention, the parking situation of a destination such as a popular scenic spot can be judged based on historical messages published by one or more different media within a time period TimeFrame, and a parking situation database DB of the popular scenic spot can be established by storing the judged results. Thus, when a user queries the parking situation of a destination in the future, the pre-judged results can be directly provided from the database, so that valuable parking information can be quickly provided to the user. In another implementation manner of the present invention, based on the destination information requested by the user, the parking situation of the destination can be judged in real time by obtaining in real time the messages published by various media within the most recent time period TimeFrame, so as to provide a prediction of the real-time parking situation for the user. The following respectively describe different exemplary implementation manners of the present invention.

[0015]

Embodiment 1

[0016] Figure 1The block diagram of the parking status prompt device 100 is shown. The prompt device 100 includes a destination confirmation module, which is implemented as Figure 1 the input interface 200, the message scraping module 300, and the parking prediction module 400 shown. The destination confirmation module is used to identify or receive the identification TID of the destination to be evaluated. In this embodiment, the destination confirmation module serves as an input interface to receive user input, which can be text input or voice input, to specify the destination of the user's expected itinerary. For example, the input can be "Where is the parking lot near the Forbidden City", or the scenic spot name or the identification TID of other features that can uniquely identify the destination can be directly input. In the case of inputting a short text such as "Where is the parking lot near the Forbidden City", the destination confirmation module 200 can identify the destination TID "Forbidden City" from this phrase. In the case where the input interface 200 receives voice input, the voice recognition module inside the input interface 200 can be used to convert the user's voice input into text, and the destination TID input by the user can be identified from the text. In an example of the present invention, the prompt device 100 can be integrated with Figure 1 the in-vehicle map system 500 shown. Therefore, the destination confirmation module / input interface 200 can be the graphical user interface of the map system. As an example, the input interface 200 can provide corresponding destination list prompts during the process of the user inputting the destination identification, so as to facilitate the user to accurately select the desired destination from the list.

[0017] After receiving the destination identification TID from the input interface 200, the message scraping module 300 scrapes at least one message MSG about this destination published on one or more media from the network resources through the network interface 600 on the vehicle, so as to form a message set {MSG1, MSG2, …… MSG N}, where N represents the number of messages scraped. In one implementation, the message scraping module 300 can target the destination TID for extraction and retrieve messages related to the destination from relevant network resources, so as to obtain information about the destination. Such network resources can be the web content published by various media, or the network information resource library containing media published information provided by a third party. In the present invention, such media can be, for example, public accounts, Weibo, specific web pages (such as travel websites), forums or comment areas and other electronic social media and / or electronic publications. In an example, the message scraping module 300 can extract messages published within a certain time range TimeFrame from these media, and this time range is also variable. For example, during specific holidays, it can be the holiday period, while during normal times, it can be just the cycle of one day.

[0018] After receiving the destination identifier TID input by the user, the parking prediction module 400 obtains the surrounding parking lot information ParkInfo around the destination TID from the map system 500, and processes the message MSG provided by the message scraping module 300 and the parking lot information ParkInfo, so as to provide the user with a parking tip TIP for this destination. It is not difficult to understand that the traffic conditions such as congestion and other information included in the message MSG released by the media and the parking lot information near the destination will surely help to understand the difficulty of parking at the destination. Therefore, according to the present invention, the message MSG and the parking lot information ParkInfo are processed by the parking prediction module 400, so as to generate a tip on whether it is convenient to park. In the present invention, the parking tip can be, for example, 'easy to park' or 'difficult to park', etc., or further provide the parking lot information where it is convenient to park nearby.

[0019] Figure 2 The block diagram of the parking prediction module 400 according to an example of the present invention is shown. As shown in the figure, the parking prediction module 400 includes a parking lot acquisition module 401, a situation awareness module 402, and a parking evaluation module 403. After receiving the information related to the destination identifier TID input by the user from the input interface 200, the parking lot acquisition module 401 sends a query request to the map system 500, requesting to obtain the surrounding parking lot information around the destination TID, for example, obtaining the parking lot information ParkInfo within a radius of 500 m. Based on this parking lot information ParkInfo, the parking lot acquisition module 401 can preliminarily determine whether it is easy to park at the destination TID. According to one implementation manner, the parking lot acquisition module 401 can determine the attributes ParkAttr of the m parking lots included in the parking lot information ParkInfo. The attribute ParkAttr can be the geographical attributes of the destination or nearby parking lots, for example, whether it is a city or a suburb; the attribute ParkAttr can also be the parking space capacity of these m parking lots. The parking lot acquisition module 401 can determine whether a parking convenience tip TIP can be generated based on the attribute ParkAttr. For example, when the destination or nearby parking lot is located in the suburb, or when the number of parking spaces in a certain parking lot or several of the m parking lots is sufficient, for example, greater than a certain quantity threshold, the parking prediction module 400 can directly output a parking tip TIP of 'easy to park' at the destination TID.

[0020] In another implementation, the m parking lots included in the parking lot information ParkInfo can be matched with the list of parking lots previously marked as "easy to park" in the database, so as to identify whether the current parking lot information ParkInfo contains parking lots that are "easy to park". If so, the parking prediction module 400 can directly output a parking prompt TIP that it is "easy to park" at the destination TID. It should be noted here that the parking lots in the list of parking lots previously determined in the database are pre-determined based on characteristics such as the destination or parking lot attributes. For example, a large parking lot located in the suburbs can be marked as "easy to park"; it can also be determined through data mining and big data analysis and prior learning.

[0021] If the parking lot acquisition module 401 cannot directly determine the parking situation of the destination only based on the parking lot information ParkInfo, such as the parking lot attribute ParkAttr therein, the parking lot acquisition module 401 further acquires the parking lot information ParkInfo within a larger radius range, such as within 2 kilometers, and then divides it according to the distance dimension, so as to form multi-dimensional parking lot space features. For example, for the parking lots within a radius of 2 kilometers, the number of parking lots N1 to N4 within the ranges of 500 meters, 500 - 1000 meters, 1000 - 1500 meters, and 1500 - 2000 meters are respectively counted.

[0022] In another example of the present invention, the parking prediction module 400 can further provide the identifier PID of the parking lot to the message scraping module 300 based on the multi-dimensional parking lot information acquired by the parking lot acquisition module 401, so as to instruct the message scraping module 300 to directionally acquire messages MSG about the destination TID and the parking lot PID from social media, thereby forming a message set {MSG}.

[0023] The situation awareness module 402 receives the message set {MSG1, MSG2,... MSG N} scraped from different media from the message scraping module 300. As an example, Figure 3A are examples of messages about the destination "The Forbidden City" and the "National Grand Theatre Parking Lot" posted by users 'AAA' and 'BBB' respectively obtained from the Weibo media, where the "National Grand Theatre Parking Lot" is very close to the "The Forbidden City".

[0024] The situation awareness module 402 processes each message MSG in the message set by using natural language processing (NLP) technology to generate sentiment expression features Sentiment_Feature and / or theme classification features Theme_Feature, where the theme classification feature Theme_Feature characterizes the comment theme mainly involved in the captured message and whether it is associated with parking, and the sentiment expression feature Sentiment_Feature represents the comment of the message publisher on the parking situation at the destination in the corresponding message. Still taking Figure 3A the message as an example, for the first message, the sentiment expression feature Sentiment_Feature obtained after being processed by the situation awareness module 300 is: 'For the Forbidden City, parking is not very convenient', and the theme classification feature Theme_Feature is: 'Parking is not allowed at the Forbidden City'; for the second message, the sentiment expression feature Sentiment_Feature obtained after being processed by the situation awareness module 300 is: 'For the National Center for the Performing Arts, parking is very convenient', and the theme classification feature Theme_Feature is: 'Parking is available at the National Center for the Performing Arts'. In this way, the situation awareness module 402 can process each message in the message set {MSG1, MSG2,... MSG N} to generate corresponding theme classification features and sentiment expression features (Theme_Feature, Sentiment_Feature). After determining the theme classification features and sentiment expression features, the parking evaluation module 403 can call the parking prediction model PTM to process the theme classification features, sentiment expression features (Theme_Feature, Sentiment_Feature) generated in all messages and multi-dimensional parking space features such as N1 to N4 to generate a parking convenience prompt TIP. In the present invention, the parking prediction model PTM can be implemented by using any model in the prior art, including but not limited to a logistic regression model, an SVM model, a random forest model, an XGBoost model or an artificial neural network model, etc., and these models are used to form the parking prediction model PTM by training a large number of data samples.

[0025] Figure 4 An exemplary configuration of the situation awareness module 402 is shown for explaining the processing of messages. Here, the first message shown in Figure 3A is taken as an example for illustration. As Figure 4 shown, the situation awareness module 402 includes a feature identification module 4021, a segmentation and filtering module 4022, a semantic analysis module 4023, and a statistics module 4024. The feature identification module 4021 is configured to identify the message set {MSG1, MSG2,... MSG Nwhether each message MSG in it contains the destination information TID. For example, when determining that the destination TID is 'The Forbidden City' based on the user's input of 'Where is the parking lot near the Forbidden City?' from the input interface 200, the feature identification module 4021 can determine Figure 3A that the message shown contains the Forbidden City identifier, i.e., TID = 'The Forbidden City'.

[0026] The segmentation and filtering module 3002 segments the message MSG and determines whether each segment contains statements, words, etc. related to parking, so as to filter out the parking-related text segments related to the destination 'The Forbidden City' from the message. Here, the parking-related text segments can be either the subjective comments of the message publisher, such as 'congested', 'parking is not convenient', etc., or the traffic information released by the traffic management department, such as 'martial law', 'no entry', etc. In this example, after the natural language processing of MSG by the segmentation and filtering module 4022, the following segments are obtained:

[0027] ['Celebrate the New Year joyfully', 'The lights of the Forbidden City look extremely elegant and grand', 'Unfortunately', 'Couldn't book tickets online', 'Parking is not very convenient either', 'Can only take a casual look around the city wall'].

[0028] Furthermore, after the filtering process of the segmentation and filtering module 4022, it can be determined that there is a parking-related text segment, i.e., 'Parking is not very convenient either'. Here, filtering techniques known in the prior art can be used for filtering, for example, for both foreign and Chinese words, various word segmenters in the prior art can be used for keyword filtering, which will not be elaborated here.

[0029] After the segmentation and filtering module 3002 determines that the message contains parking-related information, the semantic analysis module 4023 analyzes the text segment containing the parking comment in combination with the destination identifier TID determined by the feature identification module 4021, so as to obtain an accurate understanding of the parking situation of the destination. In this example, for example, the user sentiment expression feature Sentiment_Feature of 'The Forbidden City, parking is not very convenient' is obtained. In addition, the semantic analysis module 4023 determines the theme in the current message, for example, using Latent Dirichlet Allocation (LDA) for theme classification. Usually, the messages released by the media contain at least one theme, and the semantic analysis module 4023 is used to determine that there is a theme related to the parking situation in the entire message, and accordingly extracts or outputs this theme as the theme feature Theme_Feature; for other possible themes in the message, they can be ignored. For example, for Figure 3A the message shown, through the analysis of the entire message by the LDA technology, it can be determined that the theme of the current message involves 'Parking is not allowed at the Forbidden City'.

[0030] According to an example of the present invention, if the semantic analysis module 4023 determines that there are no parking-related features in the current message, as an example, the parking evaluation module may choose to delete the message, or in another example, the semantic analysis module 4023 continues to analyze the main theme in the current message based on the destination identifier TID. Usually, this theme characterizes the user's environmental comments on the destination in the corresponding media. For example, in a certain media, the semantic analysis module 4023 may determine that a certain message therein has the following theme: 'There are so many people in the Forbidden City'. Obviously, such a theme will also necessarily reflect heavy traffic, and thus may also affect the difficulty of parking. Therefore, this theme is also helpful for evaluating the destination parking lot, even though the problem of difficult parking may not be mentioned at all in the whole message.

[0031] In this way, each message in the entire message set {MSG1, MSG2, …… MSG N} is analyzed to obtain the sentiment feature Sentiment_Feature and / or the theme classification feature Theme_Feature of each message. Thus, the parking evaluation module 403 uses the parking prediction model PTM to process the (Sentiment_Feature, Theme_Feature) of all messages and the parking space features of the destination such as N1 to N4 to generate a parking convenience prompt TIP.

[0032] According to an embodiment, the semantic analysis module 4023 also scores the user's emotional expression features to determine the impact of the comments in the message on parking. Here, a trained scoring model ScoreM can be used to process the user's emotional expression features, thereby generating a score indicating the user's rating on destination parking. For example, in this example, for 'The Forbidden City, parking is not very convenient', after being processed by the scoring model ScoreM, the calculated score Score is 0.1634. The larger this value, the easier it is to park, and the smaller the value, the more difficult it is to park. Here, the scoring model ScoreM can either be a trained machine learning model obtained through learning and training of words that can affect parking comments; in another implementation, the scoring model ScoreM can also include a look-up table of standard words and their corresponding scores, and the score Score is determined by matching the keywords included in Sentiment_Feature with the standard words in the look-up table.

[0033] The statistics module 4024 is used to analyze the sentiment expression feature Sentiment_Feature of each message to obtain statistical data corresponding to multiple metrics. For example, the metrics may include the number of positive comments comment_pos_num involved in the sentiment expression feature and the average score comment_pos_mean_score of positive comments, the number of negative comments comment_neg_num involved in the sentiment expression feature and the average score comment_neg_mean_score of negative comments, etc. The average score here is calculated based on the score Score of each corresponding sentiment expression feature. In addition, the statistics module 4024 is used to analyze the topic expression feature of each message to divide it into a positive topic comment_LDA_pos and a negative topic comment_LDA_neg, and determine the parking relevance of each theme Theme. The parking relevance here can indicate the importance of the positive or negative theme among the multiple themes involved in the corresponding message, and thus can reflect the attention of the publisher to the parking-related theme in the entire message MSG. As an example, the attention can be determined by using the appearance position and frequency of the parking-related theme Theme in the message. For example, for the parking-related theme (whether it is a positive theme or a negative theme), the earlier its appearance position and / or the more its frequency in the message, the greater value is assigned to comment_LDA_pos or comment_LDA_neg, otherwise a smaller value is assigned.

[0034] Subsequently, the parking evaluation module 403 processes the statistical data comment_pos_num, comment_pos_mean_score, comment_neg_num, comment_neg_mean_score, comment_LDA_pos, comment_LDA_neg and the multi-dimensional parking space features (N1 to N4) output by the statistics module 4024 to generate a prompt TIP on whether the parking at the destination is convenient. Figure 3B Schematically shows the predicted scores and the corresponding parking convenience prompts after being processed by the model PTM for the destinations "The Palace Museum" and "National Centre for the Performing Arts".

[0035] According to a further embodiment of the present invention, after the parking prompt device 100 gives a parking prompt about the destination based on the model processing result, a prompt for a recommended parking lot Opt_Park that is convenient for parking can be further given, regardless of whether the parking at the destination is "easy to park" or "difficult to park". For example, for the prompt of "difficult to park" for "The Palace Museum", the parking prediction module 400 is based on the information captured by the message capture module 300 such as Figure 3AThe message about the "National Center for the Performing Arts" shown can be determined to have an emotional expression feature of "Parking at the National Center for the Performing Arts is very convenient" after being processed by the semantic analysis module 4023. Thus, the parking prediction module 400 can prompt the user with the parking lot and parking charging information of the National Center for the Performing Arts (when the user needs it) and can provide distance information for the user's reference. Figure 3C Schematically shows that prompt information about nearby parking lots is given for the user's reference according to the destination "The Forbidden City", and they can be sorted according to the parking convenience level. The figure shows two parking lots, Opt_Park, namely the "National Center for the Performing Arts" and the "Zhongshan Park". In another example of the present invention, specific parking lot prompts can also be directly given based on the tags of the parking lots pre-classified in the server. For example, parking lot information in society is pre-stored in the server, and its list is shown in Table 1 below:

[0036] Table 1

[0037] Parking Lot Name Label / Tip National Centre for the Performing Arts Parking Lot Easy to Park Beijing Zoo Parking Lot Difficult to Park Ming Tombs Parking Lot Easy to Park … Zhongshan Park Parking Lot Average

[0038] Thus, after the parking lot acquisition module 401 obtains the parking lot information ParkInfo around the destination 'The Forbidden City' through spatial search, the parking prediction module 400 can give parking prompts TIP for the National Center for the Performing Arts parking lot and the Zhongshan Park parking lot under the condition of distance priority. In addition, the charging and / or geographical location information of each parking lot can also be registered in the database. Thus, the charging standard and distance information of the recommended parking lot can be further given in the parking prompt provided to the user, making it more convenient for the user to choose.

[0039] It should be noted here that the implementation manner of the present invention is not limited to the above parking prompt prediction based on statistical data. As another example, the parking evaluation module 403 can process the overall parking space features, the emotional feature Sentiment_Feature and / or the theme classification feature Theme_Feature generated by the semantic analysis module 4023, and generate a convenience prompt through a trained model PTM'. For example, the parking evaluation module 403 uses a tokenizer to perform tokenization and selection processing on the sentences in Sentiment_Feature, so as to obtain each keyword in the sentence and form a keyword sequence w. Then, vector conversion is performed on each keyword in the keyword sequence. For example, the word embedding vector corresponding to each keyword is read out by looking up the word vector table, thereby generating a word vector sequence v. The word vector table used here can be obtained by training a large number of message corpora using various existing conventional training tools such as Word2Vec, FastText, and Glove. [[ID=!6]]

[0040] Subsequently, the parking evaluation module 403 further processes these word vector sequences v to extract the sequence features sf of the entire sequence. For example, a trained recurrent neural network (RNN) model can be used to achieve feature extraction. The advantage of the RNN lies in memorizing the previous information through the connection structure between nodes in each layer and using this information to affect the output of subsequent nodes. Therefore, it can fully exploit the temporal information and semantic information in the sequence data, and this information is very meaningful for accurately understanding the impact brought by the media message. As for how to design the structure of the RNN and each unit in each RNN, those skilled in the art can determine it according to the actual situation.

[0041] Similarly, the parking evaluation module 403 can generate the sequence features sf' of each theme classification feature Theme_Feature. Thus, the parking evaluation module 403 provides the sequence features sf of the sentiment feature, the sequence features sf' of the theme classification feature, and the parking space features (N1 to N4) to the model PTM', so as to output a prompt regarding the ease of parking at the destination. Here, the model PTM' can be trained based on the sequence features sf of the sentiment feature, the sequence features sf' of the theme classification feature, and the parking lot information of each destination for a large number of message samples for different destinations.

[0042] In another implementation manner of the present invention, the message scraping module 300 can further analyze the time feature of the user's travel, such as the current date when the user inputs the destination TID through the input interface 200. If the current date belongs to a holiday, such as the weekend, the message scraping module 300 can be configured to obtain only the messages MSG released by the media on each weekend within a certain time range TimeFrame from the network resources, so that the parking prediction module 400 can more accurately predict the parking situation at the destination on the current weekend based on the messages MSG with the same time feature.

[0043]

Embodiment 2

[0044] In the above embodiment, the parking situation at the destination can be predicted in real time based on the user's input; while in another embodiment of the invention, considering that the parking state at the destination is relatively stable within a certain period of time, the parking prompt device 100 of the present invention can be used to establish the analysis and prediction results of the parking situation in advance for a large number of destinations. This situation is especially applicable to popular scenic spots, and then the prediction results are stored in the server database. The server can be either a local server, a remote server, or a cloud. Thus, when the user hopes to obtain the parking situation of a certain scenic spot, the corresponding parking state can be quickly displayed to the user directly by matching the pre-established prediction results.

[0045] Therefore, when expecting to obtain parking conditions of multiple scenic spots in multiple cities, the parking prompt device 100 can use the destination confirmation module 200 to select one of the scenic spots, such as the scenic spot name, as the destination TID to be evaluated each time. Then, it is the same as the first implementation mode combined above Figure 1-4 The message scraping module 300 scrapes messages MSG related to the destination or directly related to the parking at the destination published by one or more different media within a certain time range, and then the parking prediction module 400 generates a parking prompt TIP and a parking lot recommendation information Opt_Park for the destination by processing the parking lots around the destination or the parking lots and the message MSG. The processing performed by the parking prediction module 400 is the same as that in the first implementation mode and will not be elaborated here. Subsequently, the parking prompt device 100 can store the destination identifier TID, its parking prompt TIP, and the parking lot recommendation information Opt_Park in the server in an associated manner. In this way, the parking prompt device 100 can pre-establish corresponding parking condition prediction results for multiple destinations. For example, Table 2 below schematically shows a storage example of the prediction results:

[0046] Table 2

[0047]

[0048] In another implementation mode, corresponding parking convenience prompts can also be established for the same destination according to time characteristics. For example, similar to the first implementation mode described above, the parking prompt device 100 establishes corresponding parking convenience prompts TIP according to 'holidays' and 'weekdays' respectively. Thus, more accurate matching can be performed according to the current date of the user's travel during the query.

[0049] Figure 5Schematically shown is a parking prompt processing flow chart implemented by a parking prompt device according to an example. As shown in the figure, at step 501, the identifier TID of the destination to be predicted is determined. For example, a text or voice message containing the identifier information TID about the destination is received from the user input and the desired destination TID of the user is identified therefrom; or, for establishing a destination parking status library, one scenic spot is selected from multiple destinations to be predicted, such as popular scenic spots. At step 503, based on the determined destination identifier TID, parking lot information ParkInfo within a certain radius around the destination TID is obtained from the map system 500. At step 505, based on this parking lot information ParkInfo, the attributes ParkAttr of one or m parking lots included therein are determined, including the geographical attributes of the destination or the parking lot and the parking space capacity of the parking lot, etc. Whether a parking convenience prompt TIP can be generated only based on the attributes is determined by analyzing the attributes ParkAttr. For example, when the destination or a nearby parking lot is located in the suburbs, or when the number of parking spaces in a certain parking lot or several of the m parking lots is sufficient, for example, greater than a certain quantity threshold, step 513 is entered, and a parking prompt TIP that "parking is easy" at the destination TID is directly output.

[0050] If it is determined at step 505 that a parking convenience prompt cannot be generated only based on the attributes ParkAttr, step 507 is entered, and the parking lot information ParkInfo is processed to generate multi-dimensional parking space features with the distance range as the dimension. For example, the parking lots within this radius or within a selected larger radius are divided according to the distance dimension, so as to be divided into multi-dimensional parking lot space features. For example, for the parking lots within a radius of 2 kilometers, the number of parking lots N1 to N4 within the ranges of 500 meters, 500 - 1000 meters, 1000 - 1500 meters, and 1500 - 2000 meters are respectively counted.

[0051] At step 509, one or more messages {MSG1, MSG2,... MSG N} about the destination TID published by one or more media within a certain time range TimeFrame are obtained based on the destination identifier TID. At step 511, the message set {MSG1, MSG2,... MSG NEach message MSG in} is used to generate the sentiment expression feature Sentiment_Feature and the theme classification feature Theme_Feature included in the message MSG. Subsequently, in step 512, the parking prediction model PTM processes Sentiment_Feature, Theme_Feature, and the multi-dimensional parking lot space features N1 to N4 to generate a parking tip TIP for the destination, and outputs the parking tip TIP in step 513.

[0052] In addition, the method of the present invention may further include a step of giving a prompt for the recommended parking lot Opt_Park. Whether the parking at the destination is "easy to park" or "difficult to park", here, candidate parking lots can be found either by spatially searching with the destination as the center by the parking tip device 100 or by social search to obtain candidate parking lots for the destination. Thus, the parking tip device 100 can select a recommended parking lot from these candidate parking lots based on the foregoing processing.

[0053] According to the present invention, in the case of real-time prediction of the parking situation at the destination, the parking tip device 100 directly presents the generated parking tip TIP and the recommended parking lot Opt_Park to the user as a response to the user input, for example, presents them to the user through a map navigation system.

[0054] In the case of establishing a parking situation database for multiple destinations, the generated parking tip TIP and the recommended parking lot Opt_Park can be registered in the database DB in association with the destination TID to form a record of the destination TID. Thus, when a query input request for the user to understand the parking situation at the destination is received through the input interface 200, the parking tip device 100 matches the target identifier TID' of the desired destination included in the input request with the identifier TID of the destination stored in the database DB, and extracts the convenience tip TIP and / or the recommended parking lot Opt_Park corresponding to the matched destination identifier TID. Subsequently, the parking tip device 100 can present the extracted convenience tip TIP and / or the recommended parking lot Opt_Park to the user.

[0055] Figure 6 Schematically shows a flowchart of the process of generating message features in step 511. In step 601, it is identified whether each message MSG contains destination information TID. For example, taking Figure 3A the shown message as an example, when it is determined that the user has input the destination "The Palace Museum" from the input interface, the feature identification module 4021 can determine Figure 3A that the shown message contains the identifier 'The Palace Museum'.

[0056] In step 603, natural language processing technology is used to perform natural segmentation on each message MSG. Then, in step 605, it is determined whether each segment contains comments such as statements or words related to parking, so as to filter out the parking-related segments related to the destination 'The Forbidden City' from the messages. For example, after filtering, it can be determined that there is a parking-related segment, that is, "Parking is not very convenient". If it is determined in step 605 that there is a segment containing a parking comment, then in step 607, the segment containing the parking comment is analyzed to obtain an accurate understanding of the parking situation of the destination. In this example, for instance, the user sentiment expression feature Sentiment_Feature of 'The Forbidden City, parking is not very convenient' is obtained, and in step 609, the theme in the current message MSG is determined. For example, the Latent Dirichlet Allocation (LDA) is used for topic classification. Usually, there may be more than one theme in a message. In the case of containing a parking comment, in step 609, it can be determined that there is a theme related to the parking situation in the whole message, and the theme is extracted or output accordingly as the theme feature Theme_Feature; while for other possible themes in the message, they are not considered. For example, for Figure 3A the message shown, through the analysis of the whole message by the LDA technology, it can be determined that the theme of the current message involves "Parking is not allowed at The Forbidden City".

[0057] If it is determined in step 605 that there are no parking-related statements in the current message, then proceed to step 609 to analyze the main theme in the current message based on the destination identifier. Usually, this theme represents the environmental comments of users in the corresponding media on the destination, such as "There are so many people at The Forbidden City".

[0058] In step 611, the user's emotional expression features are scored to determine the impact of the comments in the message on parking. Based on this score, the emotional expression feature Sentiment_Feature of each message is analyzed to obtain statistical data corresponding to multiple metrics. The emotional expression feature involves the number of positive comments comment_pos_num and the average score of positive comments comment_pos_mean_score, the number of negative comments comment_neg_num involved in the emotional expression feature, and the average score of negative comments comment_neg_mean_score, etc. Here, the average score is calculated based on the score of each corresponding emotional expression feature. In addition, in step 611, the topic classification feature of each message is also analyzed to divide it into a positive topic comment_LDA_pos and a negative topic comment_LDA_neg. As mentioned above, the topic classification feature here characterizes the topic to be expressed by the corresponding message and its relevance to parking. The parking relevance here can indicate the importance of the positive or negative topic among the multiple topics involved in the corresponding message, and thus can reflect the attention of the entire message publisher to the parking-related topics. The attention is determined by using the appearance position and frequency of the parking-related topic Theme in the message. For example, for the parking-related topic (whether it is a positive topic or a negative topic), the earlier its appearance position and / or the more its frequency in the message, the greater value is assigned to comment_LDA_pos or comment_LDA_neg, otherwise a smaller value is assigned.

[0059] Thus, the statistical data comment_pos_num, comment_pos_mean_score, comment_neg_num, comment_neg_mean_score, comment_LDA_pos, comment_LDA_neg determined in step 611 are provided to the prediction model PTM, so that in step 512, the prediction model PTM processes this statistical data and the multi-dimensional parking space features (N1 to N4) to generate a prompt TIP on whether the destination parking is convenient.

[0060] The general solution of the present invention has been described in combination with the respective embodiments. It should be noted that the steps and modules in the above embodiments can be implemented by hardware, software, firmware, or a combination thereof. For example, the method disclosed herein can be implemented by a processor executing machine-readable programs or instructions stored in a memory. For example, in an application example of the present invention, the method disclosed herein can be implemented by a computing device, which includes a memory storing a computer-readable program and a processor, wherein the processor executes the readable program to implement the method for prompting the target parking state proposed by the present invention. In the implementation of the present invention, the memory can be any type of storage medium, such as a hard disk, a solid-state disk, an optical storage medium, etc.

[0061] The present invention has been shown and described in detail above through the accompanying drawings and preferred embodiments. However, the present invention is not limited to these disclosed embodiments. Based on the above-mentioned multiple embodiments, those skilled in the art can know that different technical features in the above different embodiments can be combined to obtain more embodiments of the present invention, and these embodiments are also within the protection scope of the present invention.

Claims

1. A parking reminder method, comprising: Determining a destination identifier; Based on the destination identifier, obtaining at least one message about the destination from at least one medium; Generating parking lot information around the destination based on map data; Processing the at least one message and the parking lot information to generate a parking convenience reminder for the destination, including: Processing the parking lot information to generate multi-dimensional parking space features with a distance range as a dimension; Processing the at least one message to generate an emotional expression feature and a theme classification feature, wherein the theme classification feature characterizes the theme expressed by the corresponding message and its relevance to parking, and the emotional expression feature characterizes the comments on the parking conditions in the corresponding message; and Using a parking prediction model to process the theme classification feature, the emotional expression feature, and the multi-dimensional parking space feature to generate the parking convenience reminder.

2. The method according to claim 1, wherein the at least one medium includes one or more electronic social media and / or electronic publications, and the message includes one or more messages published by the at least one medium within a certain time range.

3. The method according to claim 1, wherein using natural language processing technology to process each message in the at least one message to generate an emotional expression feature and a theme classification feature includes: Identifying the destination identifier included in the message; Segmenting the message and identifying whether it contains a parking-related passage; When there is a parking-related passage, performing semantic analysis on the parking-related passage and the scenic spot identifier to generate a first theme classification feature and an emotional expression feature; When there is no parking-related passage, Deleting the message, or Performing semantic analysis on the parking-related passage to generate a second theme classification feature, and the second theme classification feature characterizes the environmental review of the destination in the message.

4. The method according to claim 3, wherein using a parking prediction model to process the theme classification feature, the emotional expression feature, and the parking space feature generated in all messages includes: Analyzing the emotional expression feature in the at least one message to obtain first statistical data corresponding to at least one first index, wherein the first index includes the number of positive comments, the average score of positive comments, the number of negative comments, and the average score of negative comments; Analyzing the theme classification feature in the at least one message to obtain second statistical data corresponding to at least a second index, wherein the second index includes the parking relevance of positive themes and the parking relevance of negative themes, and the parking relevance indicates the importance of the positive theme or negative theme among multiple themes involved in the corresponding message; Using the parking prediction model to process the first statistical data, the second statistical data, and the multi-dimensional parking space feature to generate the parking convenience reminder.

5. The method according to any one of claims 1-4, wherein the parking prediction model includes one of the following models: Logistic regression model; SVM model; Random forest model; XGBoost model; or Artificial neural network model.

6. The method according to claim 2, wherein the parking convenience prompt is related to a specific date, and obtaining the at least one message includes obtaining messages of a plurality of dates having the same date characteristics as the specific date within the certain time range.

7. The method according to any one of claims 1-4, comprising: Storing the convenience prompt based on the destination identifier; Receiving a user input, the input including a target identifier of a desired destination; Matching the target identifier with the identifiers of the stored destinations to extract the convenience prompt of the matched destination identifier; Presenting the extracted convenience prompt to the user.

8. The method according to any one of claims 1-4, wherein determining the destination identifier comprises: Receiving a user input, the input including a target identifier of a desired destination; Wherein, obtaining at least one message about the destination from at least one medium includes: using the target identifier as the destination identifier, and searching for at least one message related to parking at the destination from the at least one medium through a communication network; The method further comprises: presenting the generated parking convenience prompt to the user.

9. A parking prompt device, comprising: A destination confirmation module for determining a destination identifier to be evaluated; A message scraping module configured to obtain at least one message about the destination from at least one medium based on the destination identifier; A parking prediction module configured to process the at least one message and parking lot information around the destination to generate a parking convenience prompt for the destination, including: a parking lot acquisition module configured to: process the parking lot information to generate multi-dimensional parking space features with a distance range as a dimension; A situation awareness module configured to process the at least one message to generate an emotional expression feature and a theme classification feature, wherein the theme classification feature characterizes the theme expressed by the corresponding message and its relevance to parking, and the emotional expression feature characterizes the comments on the parking conditions in the corresponding message regarding the parking lot; and A parking evaluation module configured to: use a parking prediction model to process the theme classification feature, the emotional expression feature, and the multi-dimensional parking space features to generate the parking convenience prompt.

10. The parking prompt device according to claim 9, wherein the medium includes one or more electronic social medias and / or electronic publications, and the message scraping module is further configured to scrape messages published by the medium within a certain time range.

11. The parking prompt device according to claim 9, wherein the situation awareness module further comprises: A feature identification module configured to identify the identifier of the destination included in the message; A segmentation and filtering module configured to segment the message and identify whether it contains a parking-related paragraph; A semantic analysis module configured to: When there is a parking-related paragraph, perform semantic analysis on the parking-related paragraph and the scenic spot identifier to generate a first theme classification feature and an emotional expression feature; When there is no parking-related paragraph, Delete the message, or Perform semantic analysis on the message to generate a second topic classification feature, which characterizes the environmental commentary of the corresponding media on the destination.

12. The parking reminder device according to claim 11, wherein the situation awareness module further comprises: A statistics module configured to: Analyze the sentiment expression features in the at least one message to obtain first statistical data corresponding to at least one first metric, where the first metric includes the number of positive comments, the average score of positive comments, the number of negative comments, and the average score of negative comments; Analyze the topic classification features in the at least one message to obtain second statistical data corresponding to at least a second metric, where the second metric includes the parking relevance of positive topics and the parking relevance of negative topics, and the parking relevance indicates the importance of the positive topic or negative topic among multiple topics involved in the corresponding message; Wherein the parking evaluation module processes the first statistical data, the second statistical data, and the multi-dimensional parking space features using the parking prediction model to generate the parking convenience reminder.

13. The parking reminder device according to any one of claims 9-12, wherein the parking prediction model includes one of the following models: Logistic regression model; SVM model; Random forest model; XGBoost model; or Artificial neural network model.

14. The parking reminder device according to claim 10, wherein the parking convenience reminder is related to a specific date, and obtaining the at least one message includes obtaining messages of multiple dates with the same date characteristics as the specific date within the certain time range.

15. The parking reminder device according to any one of claims 9-12, wherein the destination identifier is stored in a database in association with the corresponding convenience reminder; The destination confirmation module includes an input interface configured to receive user input, and the input includes a target identifier of the desired destination; Wherein the parking prediction module is further configured to: Match the target identifier with the identifiers of the stored destinations to extract the convenience reminder of the matched destination identifier; Present the extracted convenience reminder to the user.

16. The parking reminder device according to claim 15, wherein the database is stored on a medium accessible to the parking reminder device, the medium comprising: For the parking reminder device is a local memory or a remote server.

17. The parking reminder device according to any one of claims 9-12, wherein the destination confirmation module includes an input interface configured to receive user input, and the input includes a target identifier of the desired destination; Wherein the message scraping module is further configured to use the target identifier as the destination identifier to search for at least one message related to destination parking from the at least one media through a communication network; The parking prediction module is further configured to: present the generated parking convenience reminder to the user.

18. A computing device, comprising: A memory in which a computer-readable program is stored, A processor for executing the readable program to implement the method according to any one of claims 1-8.

19. A computer-readable storage medium having computer-readable instructions stored thereon, which when executed by a processor cause the processor to perform the method of any one of claims 1-8.

Images