A method and device for identifying car purchase concerns

Abstract

An embodiment of the present invention provides a method and device for identifying car purchase concerns. The method includes: obtaining text information corresponding to different car purchase information acquisition modes; the text information includes online and offline text information, and the online and offline text information has its own text characteristics; according to the information mining strategy corresponding to the text characteristics of each text information, identifying the car purchase concerns of users to summarize the total car purchase concerns for each text information. By mining the online and offline text information of registered users, based on integrating different data sources corresponding to different car purchase information acquisition modes, catering to different modes of users to obtain car purchase information, comprehensively capturing the car purchase concerns of users, while solving the problem of information loss of a single data source and improving the label coverage rate, so as to provide data support for subsequent sales activities.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicles, and particularly to a method for identifying car purchase concerns and a device for identifying car purchase concerns. Background Art

[0002] Nowadays, in order to gain a foothold in the increasingly competitive automotive consumer market, automobile manufacturers and dealers not only need to have a more accurate and in-depth understanding of the car purchase concerns of the target audience, but also have a certain grasp of the user consumption trend, so as to expect to carry out a more forward-looking market layout.

[0003] Currently, the main way to obtain users' car purchase concerns is to rely on the conversations between store sales staff and users to form manual annotation records. However, this method not only consumes a lot of manpower, but also due to the resistance of some users to marketing methods, the label coverage rate obtained through sales conversations is not high. Summary of the Invention

[0004] In view of the above problems, embodiments of the present invention are proposed to provide a method for identifying car purchase concerns and a corresponding device for identifying car purchase concerns that overcome the above problems or at least partially solve the above problems.

[0005] Embodiments of the present invention disclose a method for identifying car purchase concerns, and the method includes:

[0006] Obtain text information corresponding to different car purchase information acquisition modes; the text information includes online and offline text information, and the online and offline text information has its own text characteristics;

[0007] According to the information mining strategy corresponding to the text characteristics of each text information, identify the car purchase concerns of the user to summarize the total car purchase concerns for each text information.

[0008] Optionally, the online text information includes user post text information; the identifying the car purchase concerns of the user according to the information mining strategy corresponding to the text characteristics of each text information includes:

[0009] Obtain a pre-constructed car purchase sentiment label tagging model and a car purchase concern label tagging model; wherein, the car purchase sentiment tagging model is trained based on a theme dictionary for car purchase sentiment, and the car purchase concern label tagging model is trained based on a theme dictionary for different car purchase concerns;

[0010] Perform tagging operations on the user's posted text information for theme tags according to the pre-constructed car purchase sentiment tag marking model, and perform tagging operations on the user's posted text information for concern tags according to the car purchase concern tag marking model, to complete the identification of car purchase concerns for the user's posted text information.

[0011] Optionally, the operation of performing tagging operations on the user's posted text information for theme tags according to the pre-constructed car purchase sentiment tag marking model includes:

[0012] Extract the keywords of the user's posted text information and the corresponding frequencies of the extracted keywords, obtain a keyword set based on the keywords and the corresponding frequencies, and calculate the theme vectors of each theme in the theme dictionary;

[0013] Traverse the extracted keyword set, obtain at least one target keyword that meets the preset rule conditions for the target theme from the keyword set, and calculate the theme attribution degree of the at least one target keyword for the target theme; the theme attribution degree for the target theme is determined based on the frequency of the target keyword, the weight of the target keyword in the theme dictionary, and the similarity between the target keyword vector and the theme vector;

[0014] Accumulate the theme attribution degrees of at least one target keyword to obtain the total theme attribution degree. If the total theme attribution degree meets the tagging conditions, perform tagging operations on the user's posted text information for the target theme tags.

[0015] Optionally, the car purchase concern word set includes core words for different car purchase concerns; the operation of performing tagging operations on the user's posted text information for concern tags according to the car purchase concern tag marking model includes:

[0016] Divide the user's posted text information to obtain a set of posted sentences;

[0017] Traverse the divided set of posted sentences, and obtain at least one target sentence that meets the preset rule conditions for the concern and contains the core word from the set of posted sentences;

[0018] Use a preset sentiment analysis strategy and / or a preset matching strategy to identify the at least one target sentence, and obtain the number of sentence matches of the at least one target sentence for different concerns;

[0019] If the number of sentence matches for different concerns meets the threshold conditions for the target concern, perform tagging operations on the user's posted text information for the target concern tags.

[0020] Optionally, the text information on the line includes user comment text information. Identifying the user's car purchase concerns according to the information mining strategy corresponding to the text characteristics of each text information includes:

[0021] Construct a theme dictionary containing different concern themes, and perform processing operations on the user comment text information;

[0022] Determine the matching degree between the processed user comment text information and each concern theme. If the matching degree with each concern theme meets the threshold condition of the target concern, perform a tagging operation of the target concern label on the user comment text information.

[0023] Optionally, the text information on the line includes user search text information. Identifying the user's car purchase concerns according to the information mining strategy corresponding to the text characteristics of each text information includes:

[0024] Construct a theme dictionary containing different concern themes, and perform word segmentation operations on the user search text information;

[0025] Determine the matching degree between the segmented search text words and each concern theme. If the matching degree with each concern theme meets the threshold condition of the target concern, perform a tagging operation of the target concern label on the user search text information.

[0026] Optionally, the text information offline includes sales note text information. Identifying the user's car purchase concerns according to the information mining strategy corresponding to the text characteristics of each text information includes:

[0027] Construct a theme sentence library for different concern themes, and perform sentence splitting operations on the sales note text information;

[0028] Filter the segmented sales text sentences based on the theme sentence library, calculate the similarity between the filtered sales text sentences and the theme sentences of each concern. If the similarity with each concern theme meets the threshold condition of the target concern, perform a tagging operation of the target concern label on the sales note text information.

[0029] An embodiment of the present invention also provides a device for identifying car purchase concerns. The device includes:

[0030] A text information acquisition module for acquiring text information corresponding to different car purchase information acquisition modes; the text information includes online and offline text information, and the online and offline text information has its own text characteristics;

[0031] A car purchase concern recognition module, which is used to recognize the car purchase concerns of users according to information mining strategies corresponding to the text characteristics of each text information, so as to summarize the total car purchase concerns for each text information.

[0032] Optionally, the online text information includes user post text information; the car purchase concern recognition module includes:

[0033] A tagging model acquisition sub-module, which is used to acquire a pre-constructed car purchase sentiment tag tagging model and a car purchase concern tag tagging model; among them, the car purchase sentiment tagging model is trained based on a theme dictionary for car purchase sentiment, and the car purchase concern tag tagging model is trained based on a theme dictionary for different car purchase concerns;

[0034] The first tag tagging sub-module is used to perform the tagging operation of the theme tag on the user post text information according to the pre-constructed car purchase sentiment tag tagging model;

[0035] The second tag tagging sub-module is used to perform the tagging operation of the concern tag on the user post text information according to the car purchase concern tag tagging model, and complete the recognition of the car purchase concerns for the user post text information.

[0036] Optionally, the first tag tagging sub-module includes:

[0037] A keyword set extraction unit, which is used to extract the keywords of the user post text information and the corresponding frequencies of the extracted keywords, obtain a keyword set based on the keywords and the corresponding frequencies, and calculate the theme vector of each theme in the theme dictionary;

[0038] A theme attribution calculation unit, which is used to traverse the extracted keyword set, obtain at least one target keyword that meets the preset rule conditions for the target theme from the keyword set, and calculate the theme attribution of the at least one target keyword for the target theme; the theme attribution for the target theme is determined based on the frequency of the target keyword, the weight of the target keyword in the theme dictionary, and the similarity between the target keyword vector and the theme vector;

[0039] The first tag tagging unit is used to accumulate the theme attributions of at least one target keyword to obtain the total theme attribution. If the total theme attribution meets the tagging condition, the tagging operation of the target theme tag is performed on the user post text information.

[0040] Optionally, the car purchase concern word set includes core words for different car purchase concerns; the second tag tagging sub-module includes:

[0041] A text division unit, which is used to divide the user post text information to obtain a set of post sentences;

[0042] A target statement extraction unit, configured to traverse the divided set of post statements, obtain from the set of post statements at least one target statement that meets the preset rule conditions for the concern points and includes the core word;

[0043] A target statement recognition unit, configured to recognize the at least one target statement by using a preset sentiment analysis strategy and / or a preset matching strategy, and obtain the number of statement matches of the at least one target statement for different concern points;

[0044] A second tag marking unit, configured to perform a tagging operation on the user post text information with a target concern point tag when the number of statement matches for different concern points meets the threshold condition of the target concern point.

[0045] Optionally, the online text information includes user review text information, and the vehicle purchase concern point recognition module includes:

[0046] A text processing sub-module, configured to construct a theme dictionary containing different concern point themes, and perform a processing operation on the user review text information;

[0047] A third tag marking sub-module, configured to determine the matching degree of the processed user review text information with each concern point theme. If the matching degree with each concern point theme meets the threshold condition of the target concern point, perform a tagging operation on the user review text information with a target concern point tag.

[0048] Optionally, the online text information includes user search text information, and the vehicle purchase concern point recognition module includes:

[0049] A word segmentation sub-module, configured to construct a theme dictionary containing different concern point themes, and perform a word segmentation operation on the user search text information;

[0050] A fourth tag marking sub-module, configured to determine the matching degree of the divided search text word segmentation with each concern point theme. If the matching degree with each concern point theme meets the threshold condition of the target concern point, perform a tagging operation on the user search text information with a target concern point tag.

[0051] Optionally, the offline text information includes sales note text information, and the vehicle purchase concern point recognition module includes:

[0052] A sentence splitting sub-module, configured to construct a theme sentence library for different concern point themes, and perform a sentence splitting operation on the sales note text information;

[0053] The fifth label marking sub-module is used to filter the clauses of the sales text after clause division based on the theme sentence library, calculate the similarity between the filtered clauses of the sales text and the theme sentences of each concern point. If the similarity with the theme of each concern point meets the threshold condition of the target concern point, a marking operation of the target concern point label is performed on the sales note text information.

[0054] An embodiment of the present invention also discloses a vehicle, including: the vehicle purchase concern point recognition device, a processor, a memory, and a computer program stored on the memory and capable of running on the processor. When the computer program is executed by the processor, the steps of any one of the vehicle purchase concern point recognition methods are implemented.

[0055] An embodiment of the present invention also discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the vehicle purchase concern point recognition methods are implemented.

[0056] The embodiments of the present invention include the following advantages:

[0057] In the embodiments of the present invention, by obtaining text information corresponding to different vehicle purchase information acquisition modes, the obtained text information includes online and offline text information from various different data sources. At this time, according to the text characteristics of each text information, the vehicle purchase concern points of the user can be identified to summarize the total vehicle purchase concern points for each text information. By mining the online and offline text information of registered users, based on integrating different data sources corresponding to different vehicle purchase information acquisition modes, catering to different modes of the user to obtain vehicle purchase information, comprehensively capturing the vehicle purchase concern points of the user, and providing corresponding information mining methods based on the text information characteristics under different modes, optimizing the marking effect, and comprehensively completing the identification of the user's vehicle purchase concern points under different text information, while solving the problem of information loss of a single data source and improving the label coverage rate, so as to provide data support for subsequent sales activities. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 is a flowchart of the steps of an embodiment of a vehicle purchase concern point recognition method of the present invention;

[0059] Figure 2 is a schematic diagram of vehicle purchase concern point recognition provided by an embodiment of the present invention;

[0060] Figure 3 is a schematic diagram of text information mining provided by an embodiment of the present invention;

[0061] Figure 4 is a structural block diagram of an embodiment of a vehicle purchase concern point recognition device of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0062] To make the above objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0063] Determining the user's concerns when purchasing a vehicle is conducive to grasping the user's consumption trends in advance. Currently, the main way to obtain the user's concerns when purchasing a vehicle is to rely on the conversation between the store salesperson and the user to form a manually marked record. The label coverage rate obtained by this method is not high.

[0064] One of the core ideas of the embodiments of the present invention is to automatically label the user's concerns through artificial intelligence algorithms, effectively improving efficiency. In addition, by combining the user's information acquisition habits in the current era of rapid development of the Internet and integrating the user's online and offline behavior data for mining the concerns when purchasing a vehicle, the problem of low label coverage rate in the traditional labeling method is solved.

[0065] Refer to Figure 1 , which shows the step flow chart of an embodiment of a method for identifying concerns when purchasing a vehicle according to the present invention. Specifically, it may include the following steps:

[0066] Step 101, obtain text information corresponding to different vehicle purchase information acquisition modes;

[0067] Mining the concerns when purchasing a vehicle by integrating the user's online and offline behavior data can be achieved by obtaining text information corresponding to different vehicle purchase information acquisition modes, so as to comprehensively capture the user's concerns when purchasing a vehicle based on integrating different data sources corresponding to different vehicle purchase information acquisition modes, catering to the different modes of the user to obtain vehicle purchase information, and being able to solve the problem of information loss of a single data source while improving the label coverage rate.

[0068] Specifically, as Figure 2 shown, the obtained text information may include online and offline text information. The mining of online text information includes the mining of the user's posted text information, the mining of the user's comment text information, and the mining of the user's search text information. The mining of offline text information includes the sales note text information of registered users when purchasing a vehicle. There is often a problem of low label coverage rate when mining the user's concerns when purchasing a vehicle. The reason is the lack of comprehensive consideration of the user's information acquisition channel habits. Exemplarily, in the embodiments of the present invention, various online behaviors of the user (such as app posting, app commenting, app searching) can be tracked, and the potential concerns of the user's store consultation can be recorded through the side text description of the salesperson, greatly enriching the data source to solve the problem of low label coverage rate.

[0069] Among them, the obtained online and offline text information has its own text characteristics. For example, in the mining of user post text information, the text information has the characteristics of diverse content and emotions; in the mining of user comment text information, the comment text has the characteristics of short length and profound influence of the comment mechanism; in the mining of user search text information, the search text has the characteristic of extremely short length; in the mining of sales note text information, it has the characteristic of serious colloquialism, etc.

[0070] Step 102: According to the information mining strategies corresponding to the text characteristics of each text information, identify the car purchase concerns of users, so as to summarize the total car purchase concerns for each text information.

[0071] After obtaining the text information of different data sources corresponding to different car purchase information acquisition modes, corresponding information mining methods can be provided based on the text information characteristics under different modes to optimize the tagging effect and comprehensively complete the identification of users' car purchase concerns under different text information.

[0072] Among them, mining strategies for user post text information, user comment text information, user search text information, and sales text note information can be formulated respectively based on the text characteristics of the obtained online and offline text information itself.

[0073] In one case, when mining user post text information, the text information has the characteristics of diverse content and emotions. At this time, tagging operations for emotional labels and concern labels can be performed on the user post text. This can be mainly achieved through a pre-constructed car purchase emotional label tagging model and a car purchase concern label tagging model.

[0074] Specifically, the tagging operation of theme labels can be performed on the user post text information according to the pre-constructed car purchase emotional label tagging model, and the tagging operation of concern labels can be performed on the user post text information according to the car purchase concern label tagging model to complete the identification of the car purchase concerns of the user post text information.

[0075] Among them, the car purchase emotion tagging model is trained based on a theme dictionary for car purchase emotions. In the process of constructing the car purchase emotion tagging model, a set of car purchase emotion words can be selected for the car purchase mental journey theme, including positive regular words, negative regular words, low-correlation words, medium-correlation words, and core words. The weights can be configured according to the relevance of the words to the car purchase mental journey theme, that is, the car purchase emotion tagging model can be obtained based on the relevance weight configuration between the set of car purchase emotion words and the car purchase emotion theme.

[0076] In the process of information mining using a pre - built purchase sentiment tagging model, the keywords of the user's post text information and the corresponding frequencies of the extracted keywords can be obtained, a keyword set based on the keywords and the corresponding frequencies can be obtained, and the topic vectors of each topic in the topic dictionary can be calculated. At this time, the extracted keyword set can be traversed, at least one target keyword that meets the preset rule conditions for the target topic can be obtained from the keyword set, and the topic membership degree of at least one target keyword for the target topic can be calculated. Then, the topic membership degrees of at least one target keyword can be accumulated to obtain the total topic membership degree. If the total topic membership degree meets the tagging conditions, a tagging operation for the target topic label can be performed on the user's post text information.

[0077] In practical applications, the topic vector of each topic calculated in the topic dictionary can be the sum of the w2v vectors of the low - related words and the core words in the topic. When it is necessary to tag a certain topic, after screening the target keywords that meet the positive and negative rule conditions in the keyword set, the topic membership degree of each screened target keyword is calculated, and its specific expression can be as follows:

[0078] word_topic_score[tagid]=max(cos,0.0)*freq[word]*weight

[0079] In the formula, word_topic_score[tagid] can be the topic membership degree for a certain target topic, freq[word] can be the frequency of a certain target keyword, weight can be the weight of the target keyword in the topic dictionary, and (cos,0.0) can refer to the target keyword vector, that is, the similarity between the keyword w2v vector and the topic vector. That is, the topic membership degree for the target topic can be determined based on the frequency of the target keyword, the weight of the target keyword in the topic dictionary, and the similarity between the target keyword vector and the topic vector. At this time, after the traversal of the extracted keyword set is completed, the topic membership degrees of all keywords can be accumulated to obtain the main text topic score, and the title topic score and the main text topic score are weighted and added. When the total topic membership degree meets the tagging conditions, such as an article whose total score reaches the threshold, a tagging operation for the target topic label can be performed on the user's post text information.

[0080] In an embodiment of the present invention, the purchase concern label tagging model can be trained based on a topic dictionary for different purchase concerns. In the process of constructing the purchase concern label tagging model, a purchase concern word set (including core words, negative rule words, and positive rule words), a concern tagging method name, and the required number of matches can be selected for each purchase concern.

[0081] In the process of information mining using a pre - built model for tagging purchase - related concern labels, a text mining method as shown in Figure 3 can be adopted. The semantic representation of the text and the semantic representation of the concern labels are combined, and rule matching, sentiment analysis assistance, etc. are used to determine the similarity between the text semantics and the concern label semantics. For example, a similarity score is given. When the similarity score exceeds the threshold, a label is added.

[0082] Specifically, the text information of the user's post can be divided into a set of post sentences. The divided set of post sentences is traversed to obtain at least one target sentence that meets the preset rule conditions for the concern and contains the core word from the set of post sentences. A preset sentiment analysis strategy and / or a preset matching strategy is used to identify the at least one target sentence, and the number of sentence matches of the at least one target sentence for different concerns is obtained. At this time, if the number of sentence matches for different concerns meets the threshold conditions of the target concern, for example, the number of matches reaches the corresponding threshold requirement, then a tagging operation for the target concern label can be performed on the text information of the user's post.

[0083] In practical applications, when a tag needs to be added to a certain concern, after screening the post sentences that meet the positive and negative rule conditions and contain the core word in the set of post sentences, when determining the adopted identification strategy, corresponding sentiment analysis strategies and / or matching strategies can be adopted based on the different concerns to be tagged. As an example, there may be differences in the emotional expressions when users describe some considerations in purchasing a vehicle in their posts. For example, in the two sentences "I really like the interior of the vehicle" and "I don't care about the interior of the vehicle", the former sentence expresses that the user's concern is the interior, while the latter sentence does not. And there may be cases where the emotional color of the expression for some other concerns is weaker. At this time, different identification strategies need to be adopted based on the different concerns to be tagged. It should be noted that when using the sentiment analysis strategy, a Bayesian sentiment analysis model can be used to implement it. In this regard, the embodiments of the present invention do not impose any restrictions.

[0084] In step 102, in another case, when mining information from the user's comment text information, the comment text has the characteristics of being short in length and having a profound impact on the comment mechanism. At this time, a theme dictionary containing different concern topics can be constructed, and processing operations can be performed on the user's comment text information. The processing operations can include cleaning the comment text, synonym conversion, and word segmentation, etc. Then, a method as shown in Figure 3 can be adopted. Based on the semantic representation of the text and the semantic representation of the concern labels, combined with rule matching and the bm25 algorithm, the matching degree between the processed user comment text information and each concern topic is determined. If the matching degree with each concern topic meets the threshold conditions of the target concern, then a tagging operation for the target concern label can be performed on the user comment text information.

[0085] In practical applications, through machine learning methods, tags of user focus points can be labeled for the comments themselves. When no focus points can be mined from the comments themselves, tags of user focus points can also be labeled based on the parent comments and sub - comments of the current comment itself, and the obtained tags are used as supplements to the comment text.

[0086] In step 102, in another case, when mining information from the user's search text information, the search text has the characteristic of extremely short length. At this time, a theme dictionary containing different focus - point themes can be constructed, and the user's search text information can be segmented. Then, the matching degree between the segmented search text and each focus - point theme is determined. If the matching degree with each focus - point theme meets the threshold condition of the target focus point, a tagging operation of the target focus - point label can be performed on the user's search text information.

[0087] In another case, when mining information from the sales note text information, the sales note text information has the characteristic of being highly colloquial. At this time, a theme sentence library for different focus - point themes can be constructed, and the sales note text information can be segmented into sentences. Then, the segmented sales text sentences can be filtered using the rule - matching method based on the theme sentence library, and the similarity degree between the filtered sales text sentences and the theme sentences of each focus point is calculated. For example, the bert semantic vector similarity between the high - quality sentence set of each focus point and the sales text sentences is calculated. When the similarity degree with each focus - point theme meets the threshold condition of the target focus point, such as when the similarity reaches the threshold, a tagging operation of the target focus - point label is performed on the sales note text information.

[0088] After identifying the car - buying focus points by formulating mining strategies for the user's post text information, the user's comment text information, the user's search text information, and the sales text note information respectively based on the text characteristics of the obtained online and offline text information, the total car - buying focus points for each text information can be summarized, providing data support for subsequent sales activities.

[0089] In an embodiment of the present invention, by obtaining text information corresponding to different car purchase information acquisition modes, the obtained text information includes online and offline text information from various different data sources. At this time, according to the text characteristics of each text information, the car purchase concerns of the user can be identified to summarize the overall car purchase concerns for each text information. By mining the online and offline text information of registered users, based on integrating different data sources corresponding to different car purchase information acquisition modes, catering to different modes of users' acquisition of car purchase information, comprehensively capturing the car purchase concerns of users, and providing corresponding information mining methods based on the text information characteristics under different modes to optimize the tagging effect, comprehensively completing the identification of users' car purchase concerns under different text information, while solving the problem of information loss in a single data source and improving the tag coverage rate, so as to provide data support for subsequent sales activities.

[0090] It should be noted that for the method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the embodiments of the present invention are not limited by the described action sequence, because according to the embodiments of the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily essential for the embodiments of the present invention.

[0091] Refer to Figure 4 , which shows a structural block diagram of an embodiment of a car purchase concern identification device of the present invention, and specifically may include the following modules:

[0092] A text information acquisition module 401, configured to acquire text information corresponding to different car purchase information acquisition modes; the text information includes online and offline text information, and the online and offline text information has its own text characteristics;

[0093] A car purchase concern identification module 402, configured to identify the car purchase concerns of the user according to the information mining strategy corresponding to the text characteristics of each text information, so as to summarize the overall car purchase concerns for each text information.

[0094] In an embodiment of the present invention, the online text information includes user post text information, and the car purchase concern identification module 402 may include the following sub-modules:

[0095] A tagging model acquisition sub-module, configured to acquire a pre-constructed car purchase emotion label tagging model and a car purchase concern label tagging model; wherein, the car purchase emotion tagging model is trained based on a theme dictionary for car purchase emotions, and the car purchase concern label tagging model is trained based on a theme dictionary for different car purchase concerns;

[0096] The first label marking sub-module is used to perform the label marking operation of the theme label on the user's posted text information according to the pre-constructed purchase sentiment label marking model;

[0097] The second label marking sub-module is used to perform the label marking operation of the concern label on the user's posted text information according to the purchase concern label marking model, and complete the identification of the purchase concerns of the user's posted text information.

[0098] In an embodiment of the present invention, the first label marking sub-module may include the following units:

[0099] The keyword set extraction unit is used to extract the keywords of the user's posted text information and the corresponding frequencies of the extracted keywords, obtain a keyword set based on the keywords and the corresponding frequencies, and calculate the theme vector of each theme in the theme dictionary;

[0100] The theme attribution degree calculation unit is used to traverse the extracted keyword set, obtain at least one target keyword that meets the preset rule conditions for the target theme from the keyword set, and calculate the theme attribution degree of the at least one target keyword for the target theme; the theme attribution degree for the target theme is determined based on the frequency of the target keyword, the weight of the target keyword in the theme dictionary, and the similarity between the target keyword vector and the theme vector;

[0101] The first label marking unit is used to accumulate the theme attribution degrees of at least one target keyword to obtain the total theme attribution degree. If the total theme attribution degree meets the marking conditions, the label marking operation of the target theme label is performed on the user's posted text information.

[0102] In an embodiment of the present invention, the purchase concern word set includes core words for different purchase concerns. The second label marking sub-module may include the following units:

[0103] The text division unit is used to divide the user's posted text information to obtain a set of posted sentences;

[0104] The target sentence extraction unit is used to traverse the divided set of posted sentences, and obtain at least one target sentence that meets the preset rule conditions for the concern and contains the core word from the set of posted sentences;

[0105] The target sentence recognition unit is used to identify the at least one target sentence by using a preset sentiment analysis strategy and / or a preset matching strategy, and obtain the sentence matching numbers of the at least one target sentence for different concerns;

[0106] The second tag marking unit is used to perform the tagging operation of the target concern point tag on the user post text information when the number of statement matches for different concern points meets the threshold condition of the target concern point.

[0107] In an embodiment of the present invention, the text information on the line includes user comment text information, and the vehicle purchase concern point recognition module 402 may include the following sub-modules:

[0108] The text processing sub-module is used to construct a theme dictionary containing different concern point themes and perform processing operations on the user comment text information;

[0109] The third tag marking sub-module is used to determine the matching degree between the processed user comment text information and each concern point theme. If the matching degree with each concern point theme meets the threshold condition of the target concern point, the tagging operation of the target concern point tag is performed on the user comment text information.

[0110] In an embodiment of the present invention, the text information on the line includes user search text information, and the vehicle purchase concern point recognition module 402 may include the following sub-modules:

[0111] The word segmentation sub-module is used to construct a theme dictionary containing different concern point themes and perform word segmentation operations on the user search text information;

[0112] The fourth tag marking sub-module is used to determine the matching degree between the segmented search text words and each concern point theme. If the matching degree with each concern point theme meets the threshold condition of the target concern point, the tagging operation of the target concern point tag is performed on the user search text information.

[0113] In an embodiment of the present invention, the text information offline includes sales note text information, and the vehicle purchase concern point recognition module 402 may include the following sub-modules:

[0114] The sentence splitting sub-module is used to construct a theme sentence library for different concern point themes and perform sentence splitting operations on the sales note text information;

[0115] The fifth tag marking sub-module is used to filter the split sales text sentences based on the theme sentence library, calculate the similarity degree between the filtered sales text sentences and the theme sentences of each concern point. If the similarity degree with each concern point theme meets the threshold condition of the target concern point, the tagging operation of the target concern point tag is performed on the sales note text information.

[0116] For the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple. For related parts, refer to the partial description of the method embodiment.

[0117] An embodiment of the present invention further provides a vehicle, including:

[0118] It includes the above-mentioned vehicle purchase concern recognition device, a processor, a memory, and a computer program stored on the memory and capable of running on the processor. When the computer program is executed by the processor, it realizes each process of the above-mentioned embodiment of the vehicle purchase concern recognition method, and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0119] An embodiment of the present invention further provides a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is executed by the processor, it realizes each process of the above-mentioned embodiment of the vehicle purchase concern recognition method, and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0120] Each embodiment in this specification is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. For the same or similar parts among the embodiments, reference can be made to each other.

[0121] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a device, or a computer program product. Therefore, the embodiments of the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the embodiments of the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes.

[0122] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowcharts and / or block diagrams can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal devices generate a device for realizing the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0123] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured product including an instruction device, and the instruction device realizes the process inFigure 1 one process or multiple processes and / or boxes Figure 1 the functions specified in one box or multiple boxes.

[0124] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable terminal device provide steps for implementing the functions specified in one process Figure 1 one process or multiple processes and / or boxes Figure 1 or multiple processes and / or the functions specified in one box or multiple boxes.

[0125] Although the preferred embodiments of the embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments and all changes and modifications falling within the scope of the embodiments of the present invention.

[0126] Finally, it should also be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or terminal device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or elements inherent to such process, method, article or terminal device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or terminal device comprising the element.

[0127] The above has introduced in detail a method and a device for identifying car purchase concerns provided by the present invention. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A method for identifying car purchase concerns, characterized in that, the method includes: Obtaining text information corresponding to different car purchase information acquisition modes; the text information includes online and offline text information, and the online and offline text information has its own text characteristics; According to the information mining strategies corresponding to the text characteristics of each text information, identify the car purchase concerns of users to summarize the total car purchase concerns for each text information; Among them, the online text information includes user post text information; the identifying the car purchase concerns of users according to the information mining strategies corresponding to the text characteristics of each text information includes: Obtaining a pre-constructed car purchase sentiment label tagging model and a car purchase concern label tagging model; among them, the car purchase sentiment tagging model is trained based on a theme dictionary for car purchase sentiment, and the car purchase concern label tagging model is trained based on a theme dictionary for different car purchase concerns; Perform the tagging operation of theme labels on the user post text information according to the pre-constructed car purchase sentiment label tagging model, and perform the tagging operation of concern labels on the user post text information according to the car purchase concern label tagging model to complete the identification of car purchase concerns for the user post text information.

2. The method according to claim 1, characterized in that, the performing the tagging operation of theme labels on the user post text information according to the pre-constructed car purchase sentiment label tagging model includes: Extracting the keywords of the user post text information and the corresponding frequencies of the extracted keywords to obtain a keyword set based on the keywords and the corresponding frequencies, and calculating the theme vectors of each theme in the theme dictionary; Traverse the extracted keyword set, obtain at least one target keyword that satisfies the preset rule conditions for the target theme from the keyword set, and calculate the theme attribution degree of the at least one target keyword for the target theme; the theme attribution degree for the target theme is determined based on the frequency of the target keyword, the weight of the target keyword in the theme dictionary, and the similarity between the target keyword vector and the theme vector; Accumulate the theme attribution degrees of at least one target keyword to obtain the total theme attribution degree. If the total theme attribution degree meets the tagging conditions, perform the tagging operation of the target theme label on the user post text information.

3. The method according to claim 1, characterized in that, the theme dictionary of the car purchase concerns includes core words for different car purchase concerns; the performing the tagging operation of concern labels on the user post text information according to the car purchase concern label tagging model includes: Dividing the user post text information to obtain a set of post sentences; Traverse the divided set of post sentences, and obtain at least one target sentence that satisfies the preset rule conditions for the concern and contains the core word from the set of post sentences; Use a preset sentiment analysis strategy and / or a preset matching strategy to identify the at least one target sentence, and obtain the sentence matching numbers of the at least one target sentence for different concerns; If the number of statement matches for different focus points meets the threshold condition of the target focus point, then perform the tagging operation of the target focus point label on the user's posted text information.

4. The method according to claim 1, wherein, the online text information includes user comment text information, and identifying the user's car purchase focus points according to the information mining strategy corresponding to the text characteristics of each text information includes: constructing a theme dictionary containing themes for different focus points, and performing processing operations on the user comment text information; determining the matching degree between the processed user comment text information and each focus point theme. If the matching degree with each focus point theme meets the threshold condition of the target focus point, then perform the tagging operation of the target focus point label on the user comment text information.

5. The method according to claim 1, wherein, the online text information includes user search text information, and identifying the user's car purchase focus points according to the information mining strategy corresponding to the text characteristics of each text information includes: constructing a theme dictionary containing themes for different focus points, and performing word segmentation operations on the user search text information; determining the matching degree between the segmented search text and each focus point theme. If the matching degree with each focus point theme meets the threshold condition of the target focus point, then perform the tagging operation of the target focus point label on the user search text information.

6. The method according to claim 1, wherein, the offline text information includes sales note text information, and identifying the user's car purchase focus points according to the information mining strategy corresponding to the text characteristics of each text information includes: constructing a theme sentence library for different focus point themes, and performing sentence splitting operations on the sales note text information; filtering the split sales text sentences based on the theme sentence library, calculating the similarity degree between the filtered sales text sentences and the theme sentences of each focus point. If the similarity degree with each focus point theme meets the threshold condition of the target focus point, then perform the tagging operation of the target focus point label on the sales note text information.

7. A car purchase focus point recognition device, wherein, the device includes: a text information acquisition module, configured to acquire text information corresponding to different car purchase information acquisition modes; the text information includes online and offline text information, and the online and offline text information has its own text characteristics; a car purchase focus point recognition module, configured to identify the user's car purchase focus points according to the information mining strategy corresponding to the text characteristics of each text information, so as to summarize the total car purchase focus points for each text information; wherein, the online text information includes user posted text information, and the car purchase focus point recognition module includes: a tagging model acquisition sub-module, configured to acquire a pre-constructed car purchase sentiment label tagging model and a car purchase focus point label tagging model; wherein, the car purchase sentiment tagging model is trained based on a theme dictionary for car purchase sentiment, and the car purchase focus point label tagging model is trained based on a theme dictionary for different car purchase focus points; The first label marking sub-module is used to perform the marking operation of theme labels on the user's post text information according to the pre-constructed car purchase sentiment label marking model; The second label marking sub-module is used to perform the marking operation of concern labels on the user's post text information according to the car purchase concern label marking model, and complete the identification of car purchase concerns for the user's post text information.

8. A vehicle, Characterized in that, Comprising: The car purchase concern identification device, processor, memory and computer program stored on the memory and capable of running on the processor as described in claim 7, and when the computer program is executed by the processor, the steps of the car purchase concern identification method as described in any one of claims 1-6 are realized.

9. A computer-readable storage medium, Characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the car purchase concern identification method as described in any one of claims 1-6 are realized.

Images