Hotspot information processing method and device, server and readable storage medium

Abstract

The application provides a hotspot information processing method and device, a server and a readable storage medium, and relates to the technical field of digital marketing. The method comprises the following steps: obtaining a hotspot item of a first website, determining application scenario information corresponding to the hotspot item, obtaining hotspot information from a second website, determining target hotspot information associated with the application scenario information in the hotspot information, and obtaining potential item demand according to the target hotspot information and the application scenario information. In the technical solution, a high-concern hotspot item in the first website is obtained, hotspot information is obtained from other network sites, target hotspots with high association degrees with the shopping scenarios of the hotspot items are found in the hotspot information, potential item demand is analyzed, manual real-time monitoring of hotspot information of other network sites and subjective selection of a hotspot item set in a shopping website are not required, and manual cost is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Internet, and particularly relates to a hotspot information processing method and device, a server and a readable storage medium. BACKGROUND

[0002] With the development of Internet technology, more and more users shop in online shopping malls. When the users browse the online shopping mall, the online shopping mall can push various items to the users. The online shopping mall has many types of items, and usually selects some items of interest to the users and pushes them to the users, for example, based on some hot topics / hot events on the current network, the online shopping mall recommends relevant items to the users.

[0003] In the prior art, when pushing items based on hot topics / hot events on the network, hot information on the network is obtained by manual pre-processing, a relevant item set is selected based on the hot information, and then items are selected from the item set and pushed to the users.

[0004] However, the prior art needs manual real-time monitoring of hot information and manual selection of items, and the whole process is time-consuming and laborious, and involves high labor costs. SUMMARY

[0005] The present application provides a hotspot information processing method and device, a server and a readable storage medium, which are used to solve the problem of high labor costs in the prior art item pushing process.

[0006] In a first aspect, an embodiment of the present application provides a hotspot information processing method, comprising:

[0007] obtaining a hot item of a first website, determining application scenario information corresponding to the hot item, and the hot item being an item in the first website with a user attention degree reaching a preset threshold;

[0008] obtaining hot information from a second website, and determining target hot information associated with the application scenario information in the hot information;

[0009] obtaining a potential item demand according to the target hot information and the application scenario information, and the potential item demand being used to indicate an item of interest to a user in the first website.

[0010] In a possible design of the first aspect, the obtaining of the hot item of the first website comprises:

[0011] obtaining a historical behavior of the user to the items in the first website, and the historical behavior comprising at least one of a browsing behavior, a search behavior, an ordering behavior and a collection behavior;

[0012] According to the number of the historical behaviors and the preset threshold, a hot item is selected from the items of the first website.

[0013] In a possible design of the first aspect, the determining of the application scenario information corresponding to the hot item comprises:

[0014] Obtaining item information of the hot item, the item information comprising title information and attribute information;

[0015] According to the title information and the attribute information, obtaining application scenario information of the hot item.

[0016] In a possible design of the first aspect, the obtaining of the application scenario information of the hot item according to the title information and the attribute information comprises:

[0017] Extracting a scene keyword from the title information, the scene keyword being a word describing an application scenario of the hot item;

[0018] Obtaining a preset attribute in the attribute information, to obtain an attribute value corresponding to the preset attribute;

[0019] Combining the scene keyword and the attribute value to obtain the application scenario information.

[0020] In a possible design of the first aspect, the determining of the target hot information associated with the application scenario information in the hot information comprises:

[0021] According to the hot information and the application scenario information, training a preset model to obtain a first target model;

[0022] According to the first target model, determining the target hot information associated with the application scenario information from the hot information.

[0023] In a possible design of the first aspect, the training of the preset model according to the hot information and the application scenario information to obtain a first target model comprises:

[0024] Vectorizing the hot information to obtain a representation vector of the hot information;

[0025] Vectorizing the application scenario information to obtain a representation vector of the application scenario information;

[0026] Obtaining a distance between the representation vector of the hot information and the representation vector of the application scenario information;

[0027] According to the distance, determining positive samples and negative samples of the preset model, the positive samples and the negative samples being used as training data of the preset model.

[0028] According to training data of the preset model, the preset model is trained to obtain a first target model.

[0029] In a further possible design of the first aspect, the vectorization of the hotspot information to obtain the representation vector of the hotspot information includes:

[0030] segmenting the hotspot information to obtain at least one hotspot segment;

[0031] vectorizing each hotspot segment to obtain a segment vector corresponding to each hotspot segment;

[0032] obtaining the representation vector of the hotspot information according to the segment vector corresponding to each hotspot segment.

[0033] In a further possible design of the first aspect, the vectorization of the application scenario information to obtain the representation vector of the application scenario information includes:

[0034] obtaining a scene name of the application scenario information and an item included in the application scenario information, the application scenario information including at least one item information;

[0035] vectorizing the scene name to obtain a representation vector of the scene name;

[0036] vectorizing the item information to obtain a representation vector of the item information;

[0037] obtaining the representation vector of the application scenario information according to the representation vector of the scene name and the representation vector of the item information.

[0038] In a further possible design of the first aspect, the determination of the positive sample and the negative sample of the preset model according to the distance includes:

[0039] obtaining hotspot information with a distance greater than a preset distance as the positive sample of the preset model;

[0040] selecting hotspot information with a distance less than or equal to the preset distance as the negative sample of the preset model.

[0041] In a further possible design of the first aspect, the training of the preset model according to the training data of the preset model to obtain a first target model includes:

[0042] training the preset model according to the positive sample and the negative sample of the preset model to obtain an initial model;

[0043] According to the initial model, scoring hot spot information with a distance less than or equal to a preset distance, to obtain a score corresponding to each hot spot information with a distance less than or equal to the preset distance;

[0044] From the hot spot information with a distance less than or equal to the preset distance, selecting hot spot information with a score greater than a preset score threshold and updating to the positive sample;

[0045] From the hot spot information with a distance less than or equal to the preset distance, selecting hot spot information with a score less than or equal to a preset score and updating to the negative sample;

[0046] According to the updated positive sample and negative sample, training the initial model to obtain the first target model.

[0047] In another possible design of the first aspect, the obtaining of the potential item demand according to the target hot spot information and the application scenario information includes:

[0048] According to the application scenario information, obtaining a hot spot information set associated with the application scenario information from the second website;

[0049] According to the hot spot information set, training a preset classification model to obtain a second target model;

[0050] According to the target hot spot information, the application scenario information, and the second target model, obtaining the potential item demand.

[0051] In a second aspect, an embodiment of the present application provides a hot spot information processing apparatus, including:

[0052] An item obtaining module, configured to obtain a hot spot item of a first website, and determine application scenario information corresponding to the hot spot item, the hot spot item being an item in the first website with a user attention degree reaching a preset threshold;

[0053] An information association module, configured to obtain hot spot information from a second website, and determine target hot spot information associated with the application scenario information from the hot spot information;

[0054] An item demand obtaining module, configured to obtain a potential item demand according to the target hot spot information and the application scenario information, the potential item demand being used to indicate an item of interest to a user in the first website.

[0055] In a third aspect, an embodiment of the present application provides a server, including a processor and a memory in communication connection with the processor;

[0056] The memory stores computer execution instructions;

[0057] The processor executes computer-executed instructions stored in the memory to implement the method as described above.

[0058] In a fourth aspect, an embodiment of the present application provides a readable storage medium, wherein the readable storage medium stores computer instructions, and the computer instructions are executed by a processor to implement the method as described above.

[0059] In a fifth aspect, an embodiment of the present application provides a program product, comprising computer instructions, and the computer instructions are executed by a processor to implement the method as described above.

[0060] The method, device, server and readable storage medium provided by the embodiments of the present application can obtain hot items with high attention in a first website, and obtain hot information from other network sites, find target hot spots with high correlation with the shopping scene of the hot items in the hot information, and analyze potential item demand, without the need for manual real-time monitoring of hot information of other network sites and subjective selection of hot item sets in shopping websites, thereby reducing labor costs. BRIEF DESCRIPTION OF DRAWINGS

[0061] The accompanying drawings, which are incorporated herein and form a part of the specification, illustrate embodiments consistent with the present application and, together with the description, further serve to explain the principles of the present application;

[0062] Fig. 1 A scene schematic diagram of the method for processing hot information provided by the embodiments of the present application is shown;

[0063] Fig. 2 A flowchart of the first embodiment of the method for processing hot information provided by the embodiments of the present application is shown;

[0064] Fig. 3 A flowchart of the second embodiment of the method for processing hot information provided by the embodiments of the present application is shown;

[0065] Fig. 4 A structure schematic diagram of the device for processing hot information provided by the embodiments of the present application is shown;

[0066] Fig. 5 A structure schematic diagram of the server provided by the embodiments of the present application is shown.

[0067] The above-described drawings have shown the specific embodiments of the present application, and more detailed descriptions will be given hereinafter. These drawings and textual descriptions are not intended to limit the scope of the concept of the present application by any means, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0068] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative effort belong to the scope of protection of the present application.

[0069] First, the terms involved in the present application are explained:

[0070] Hotspot: refers to news or information that is more concerned by the general public or welcomed, or refers to a place or problem that attracts attention at a certain period of time. For example, "social hotspot" and "Mianshan becomes a hotspot for tourism".

[0071] Fig. 1 The scene schematic diagram of the hotspot information processing method provided by the embodiments of the present application. The application scenario of the present application can be the scenario in which the user shops on a website. As shown in the figure, Fig. 1 the user can log in to the website through a mobile terminal 10 (such as a personal computer), browse, click, collect and place an order for the items displayed on the website. The user will have different degrees of attention to different items according to his personal interests. For example, the user will pay attention to the current hot information in life, and if the user finds that there are some potential item needs in the hot information, he may place an order for these items on the website. In order to improve the user experience, the website can also actively search for the current hot information in life, and then find some items that the user may need from it and display them on the page.

[0072] However, in actual life applications, the maintenance personnel of the website needs to manually obtain hot information, and then screen some items that the user may need from the website according to subjective speculation, form an item set and upload it to the server 11 of the website, and then the server 11 distributes and pushes these items to each user. This way requires a lot of manual cost, and the response speed to hot information is not fast enough. In order to reduce the labor cost, there is another method, which mainly crawls public opinion information from other website servers 12 through an algorithm, and then inputs it into a prediction model to predict which public opinion belongs to hot public opinion and which belongs to non-hot public opinion. This method does not have explainability through algorithm mining of hot public opinion, is easy to lead to the inability to reuse historical experience, and can only determine the process of manually obtaining hot public opinion, but cannot solve the process of finding associated items from hot information.

[0073] To solve the above problems, the embodiment of the present application provides a hotspot information processing method, device, server and readable storage medium, which finds a hotspot item with high attention degree and application scenario information corresponding to the hotspot item by using the attention degree of users to different items on a website, then obtains hotspot information from other websites, determines which hotspot information is target hotspot information with high correlation degree with the application scenario information, and finally analyzes potential item demand of the user according to the target hotspot information and the application scenario information. The whole process can eliminate the workload of manually mining hotspot information, can realize automatic identification of target hotspot information, can reduce the cost of manually screening associated items, can enhance the interpretability of hotspot information, can finally infer the potential item demand of the user, and can improve the use experience of the user on the website.

[0074] The technical solutions of the present application will be described in detail through specific embodiments. It should be noted that the following specific embodiments can be combined with each other, and the same or similar concepts or processes can not be described in detail in some embodiments.

[0075] Fig. 2 The flowchart of the hotspot information processing method provided by the embodiment of the present application is shown in the figure. The method can be applied to a local computer device, and can also be applied to a server in the cloud in actual application. As shown in the figure, the method can specifically include the following steps: Fig. 2

[0076] S201, obtaining a hotspot item of a first website and determining application scenario information corresponding to the hotspot item.

[0077] The hotspot item is an item with a user attention degree reaching a preset threshold in the first website.

[0078] In the embodiment, the first website can be a website for users to browse, search and order items, such as some existing shopping websites. These websites usually contain a large amount of item information. Due to the limitation of the display interface, the website generally selects only part of the items from the large amount of items and displays them on the display interface. If the items displayed on the display interface are not the items of interest to the user, the user still needs to browse, search and the like to find the items of interest.

[0079] When the user group of the first website is large, if different users pay attention to the same item, for example, all have the behaviors of clicking, searching and ordering the item, the attention degree of the item will exceed the preset threshold, and the item can be called a hotspot item. For example, the server can take a period of time (such as one month) as a time period, and count the attention degree of each item in each time period to determine which items in the time period are hotspot items. ​

[0080] In this embodiment, hot items usually carry item information to facilitate users to browse and understand the item, specifically, title information and attribute information, etc. The title information is usually a text description of the important features of the item, while the attribute information is a text description of the detailed attributes of the item.

[0081] For example, the item information of the hot items may be as shown in Table 1 below:

[0082]

[0083] Table 1

[0084] Among them, the attribute information includes the gross weight of the item, the origin of the item, the item identification, the category to which it belongs and the applicable event.

[0085] In this embodiment, the application scenario information can be extracted from the title information and attribute information. For example, taking Table 1 as an example, the application scenario information can be "student_outdoor_resumption of work, student_outdoor_school opening, student_outdoor_epidemic prevention, student_outdoor_first aid".

[0086] S202: Obtain hot spot information from the second website, and determine target hot spot information associated with the application scenario information in the hot spot information.

[0087] For example, the second website may be some network sites other than the first website, such as a social networking site, an information network site, a game forum, etc.

[0088] In this embodiment, hot topics are usually topics that have recently attracted widespread attention and discussion in society. Hot topics are usually composed of a simple sentence. For example, hot topics can be #The Strongest Cold Wave Is Coming#, #New Outdoor Travel#, and #Hawking Passes Away#.

[0089] Among them, some hotspot information can be associated with application scenario information, and these hotspot information will be used as target hotspot information. Specifically, the application scenario information and hotspot information can be vectorized separately, and then the vector distance between the application scenario information and the hotspot information can be obtained. Then, based on the distance, it can be determined which hotspot information has an association with the application scenario information.

[0090] For example, the hot information #The strongest cold wave is coming# is associated with the application scenario information "hydration_outdoors_keeping warm", while #Hawking's death# has no related application scenario information.

[0091] S203: Obtain potential item demand based on target hotspot information and application scenario information.

[0092] The potential item demand is used to indicate an item of interest of the user in the first website. For example, the item of interest of the user can be an outdoor backpack, a mask, or the like.

[0093] In this embodiment, the target hot information can be associated with the application scenario information, some application scenario information with strong relevance is found, and the potential item demand is inferred therefrom. For example, the application scenario information associated is "male student _ autumn and winter _ outdoor _ travel", and then the outdoor backpack, autumn and winter coat, and travel tent can be analyzed from the application scenario information.

[0094] Further, after the potential item demand indicating the item of interest of the user is obtained, the item of interest of the user can be displayed on the page of the first website for the user to check. In this way, the item related to the current hot information can be pushed to the user according to the current hot information, and the success rate of placing an order of the item is improved.

[0095] The embodiments of the present application obtain the hot item and the application scenario information corresponding to the hot item from the first website, and then screen the target hot information associated with the application scenario information from the second website, without manual mining and screening of the hot information, so that the labor cost is reduced. Meanwhile, the potential item demand can be inferred from the target hot information and the application scenario information, the item of interest of the user is accurately found, and finally the accurate pushing of the item is realized, and the purchase conversion rate of the item is improved.

[0096] In some embodiments, the step S201 of "obtaining the hot item of the first website" can be implemented by the following steps:

[0097] Obtaining the historical behavior of the user to the item in the first website;

[0098] Selecting the hot item from the item of the first website according to the number of historical behaviors and a preset threshold.

[0099] The historical behavior includes at least one of a browsing behavior, a search behavior, an order behavior, and a collection behavior.

[0100] In this embodiment, the historical behavior can be the behavior of the user in the first website in the recent period (for example, one month), and the hot item refers to the item with a number of historical behaviors reaching a certain frequency. The preset threshold can be an experience value.

[0101] For example, the item with a total browsing of more than one million times in the recent one month can be taken as the hot item.

[0102] The embodiment of the application mines which items in the website are hot items by using the historical behavior of the user on the website, avoids manual mining and subjective speculation of hot items, makes the hot items have interpretability, reduces the labor cost, and improves the mining efficiency of the hot items.

[0103] In some embodiments, the step of "determining the application scenario information corresponding to the hot item" in the above step S201 can be implemented through the following steps:

[0104] Obtaining item information of the hot item;

[0105] Obtaining the application scenario information of the hot item according to the title information and the attribute information.

[0106] The item information includes the title information and the attribute information.

[0107] In the embodiment, the attribute information can further include attributes such as applicable season, applicable crowd, and applicable event of the item, and different attributes correspond to different attribute values. For example, referring to Table 1, the attribute information in the above Table 1 includes the applicable event, and the attribute value corresponding to the applicable event is epidemic prevention and first aid.

[0108] The embodiment of the application splits the item information into the title information and the attribute information, obtains the application scenario information of the hot item through the title information and the attribute information, and can make the obtained application scenario information more accurate.

[0109] Further, on the basis of the above embodiment, in some embodiments, the step of "obtaining the application scenario information of the hot item according to the title information and the attribute information" can be implemented through the following steps:

[0110] Extracting the scene segmentation from the title information;

[0111] Obtaining a preset attribute in the attribute information to obtain an attribute value corresponding to the preset attribute;

[0112] Combining the scene segmentation and the attribute value to obtain the application scenario information.

[0113] The scene segmentation is a word describing the application scenario of the hot item.

[0114] In this embodiment, for title information, scene segmentation in the title information can be identified based on a sequence labeling model. Exemplarily, the sequence labeling model includes but is not limited to a probabilistic graphical model and a deep learning model. The probabilistic graphical model can be a conditional random field algorithm (CRF) or a hidden Markov model (HMM), and the deep learning model can be a long short-term memory artificial neural network model (Bi-directional Long Short-Term Memory conditional random field algorithm, BiLSTM-CRF).

[0115] The sequence tagging model can be divided into four parts of speech: applicable events, applicable locations, applicable groups, and applicable times. Scene segmentation is extracted from the title information using these four parts of speech.

[0116] For example, taking the title "AL-NASR / Alnas epidemic prevention package, resumption of work epidemic prevention package, student back-to-school epidemic prevention package, outdoor sterilization package, portable portable epidemic prevention supplies, student disinfection package, portable epidemic prevention package (family package)" in Table 1 as an example, the extracted scene segmentation is shown in Table 2 below:

[0117] Scene word segmentation Word class Work resumption Applicable event Student Applicable population School opening Applicable event Outdoor Applicable location Epidemic prevention Applicable event

[0118] Table 2

[0119] In this embodiment, after extracting scene segmentation words and attribute values, each typically contains at least one word. For example, the scene segmentation words in Table 2 include five words. For example, the format of application scenario information can be defined as: Applicable People, Applicable Locations, Applicable Events. By entering the corresponding scene segmentation words and attribute values ​​according to the application scenario format, application scenario information can be obtained.

[0120] For example, taking the scene segmentation word corresponding to the applicable population as students, the scene segmentation word corresponding to the applicable location as outdoor, the scene segmentation word corresponding to the applicable event as resumption of work, opening of school, and epidemic prevention, and the attribute values ​​corresponding to the applicable events in Table 1 as epidemic prevention and first aid, the application scenario information obtained by permutation and combination includes: students_outdoors_resumption of work, students_outdoors_opening of school, students_outdoors_epidemic prevention, students_outdoors_first aid.

[0121] The embodiment of the present application can combine more accurate application scenario information by extracting scene segmentation words and attribute values ​​corresponding to preset attributes in the title information, thereby facilitating the subsequent accurate finding of potential item needs.

[0122] In some embodiments, the step of "determining target hotspot information associated with the application scenario information from the hotspot information" in step S202 can be implemented by the following steps:

[0123] training the preset model according to the hotspot information and the application scenario information to obtain a first target model;

[0124] determining target hotspot information associated with the application scenario information from the hotspot information according to the first target model.

[0125] In the present embodiment, the preset model can be a scene-based machine learning binary classification model, including but not limited to logistic regression, gbdt, naive Bayes, etc. The training data of the preset model can be the hotspot information obtained from the second website.

[0126] For example, in the process of obtaining the hotspot information, the hotspot information can also be data enhanced based on the obtained hotspot information. Specifically, based on the obtained hotspot information, the second website is continuously searched for related information of the hotspot information, where the related information can be a topic related to the hotspot information and discussion content under the topic. For example, taking # the strongest cold wave is coming # as the hotspot information, the related topic can be the discussion information related to the hotspot information, such as "# the strongest cold wave is coming # friends ~ come and share your cold-weather equipment" and "the strongest cold wave is coming! What to do if the skin is severely dehydrated?".

[0127] In the present embodiment, after obtaining the hotspot information, the hotspot information can be automatically labeled to find the hotspot information related to the application scenario information as a positive sample, and the remaining hotspot information as a negative sample, so as to train the subsequent first target model.

[0128] The present embodiment can find the potential item demand corresponding to the hotspot information by data enhancing the hotspot information, and can accumulate more training data for the subsequent training of the first target model. The prediction effect of the first target model is improved, the first target model can be used to automatically distinguish the hotspot information related to the item from any information, without manually crawling the hotspot information from the network site and determining which hotspot information is related to the item, thereby reducing the labor cost and improving the response speed of the hotspot information to avoid the hotspot information from being outdated.

[0129] Further, on the basis of the above-mentioned embodiments, in some embodiments, the step of "training the preset model according to the hotspot information and the related information to obtain a first target model" can be implemented by the following steps:

[0130] vectorizing the hotspot information to obtain a representation vector of the hotspot information;

[0131] vectorize the application scenario information to obtain a representation vector of the application scenario information;

[0132] obtain a distance between the representation vector of the hotspot information and the representation vector of the application scenario information;

[0133] determine, according to the distance, positive samples and negative samples of the preset model;

[0134] train the preset model according to the training data of the preset model to obtain a first target model.

[0135] The positive samples and the negative samples are used as the training data of the preset model.

[0136] In this embodiment, the hotspot information can be represented by a semantic vector. For example, a corresponding semantic vector of the hotspot information can be generated based on a fasttext vector tool, so as to represent the hotspot information. Similarly, the application scenario information can also be represented by a semantic vector.

[0137] In this embodiment, the vector distance between the representation vector of the hotspot information and the representation vector of the application scenario information can be calculated. For example, the vector distance includes but is not limited to a cosine distance, an Euclidean distance, etc. Specifically, taking the cosine distance as an example, a distance threshold (for example, 0.95) can be set, and the hotspot information with a vector distance greater than the distance threshold is selected as the hotspot information strongly related to the application scenario information, and the hotspot information with a vector distance less than or equal to the distance threshold is selected as the hotspot information weakly related to the application scenario information.

[0138] The hotspot information strongly related to the application scenario information is the positive sample, and the hotspot information weakly related to the application scenario information is the negative sample.

[0139] By using the vector distance, the positive sample and the negative sample are selected as the training data of the preset model, and the first target model is trained, so that the automatic identification of the sudden public opinion can be realized, and the reaction speed of the shopping website to the time-sensitive hotspot information is greatly improved. The user is pushed with the items related to the hotspot information, so as to improve the purchase conversion rate of the items.

[0140] Further, on the basis of the above embodiment, in some embodiments, the above step of "vectorizing the hotspot information to obtain a representation vector of the hotspot information" can be implemented by the following steps:

[0141] perform word segmentation on the hotspot information to obtain at least one hotspot word segment;

[0142] vectorize each hotspot word segment to obtain a word segment vector corresponding to each hotspot word segment;

[0143] obtain the representation vector of the hotspot information according to the word segment vector corresponding to each hotspot word segment.

[0144] In the embodiment, the hotspot information is a short sentence, and the hotspot information needs to be segmented into several hotspot words, and a word vector corresponding to each hotspot word is calculated, and finally the word vectors of each hotspot word are integrated to obtain a representation vector of the hotspot information.

[0145] For example, the hotspot information is # the strongest cold wave is coming #, and after data augmentation, the related information # the strongest cold wave is coming # friends ~ share your cold-weather equipment and # the strongest cold wave is coming! What should I do if my skin is severely dehydrated? # is obtained. The segmentation process can be seen from Table 3:

[0146]

[0147] Table 3

[0148] In the process of segmenting the hotspot information, the jieba segmentation tool can be used, and then a semantic vector corresponding to each segmented word is obtained, and finally a representation vector of the hotspot information is obtained.

[0149] For example, if a hotspot information T contains several hotspot words, i.e. T = {w1, w2, …, wn}, where wn is the nth hotspot word (n is a positive integer not less than 1). Then the word vector V can be obtained n n is the nth hotspot word corresponding to the word vector. The representation vector of the hotspot information is:

[0150]

[0151] In the above formula, Vec T is the representation vector of the hotspot information.

[0152] On the basis of the above embodiment, in some embodiments, the above step “vectorizing the application scenario information to obtain a representation vector of the application scenario information” can be realized by the following steps:

[0153] Obtaining the scene name of the application scenario information and the items contained in the application scenario information;

[0154] Vectorizing the scene name to obtain a representation vector of the scene name;

[0155] Vectorizing the item information to obtain a representation vector of the item information;

[0156] According to the representation vector of the scene name and the representation vector of the item information, obtaining a representation vector of the application scenario information.

[0157] ​​​The application scenario information includes at least one item information. For example, the application scenario information in Table 4 has a scene name of student_outdoor_prevention, and three items corresponding to the scene name can be obtained by querying the scene name of the application scenario information.

[0158]

[0159] Table 4

[0160] In this embodiment, the segmented word set of the scene name and the segmented word set of the item name can be obtained first, and then the segmented word sets of the scene name and the item name set are vectorized to obtain the representation vector of the scene name and the representation vector of the item information. Then, the representation vector of the application scenario information is obtained.

[0161] For example, if the scene name of the application scenario information is student_outdoor_prevention, the segmented word set of the scene name is {student, outdoor, prevention}.

[0162] For example, the segmented word set of the scene name is represented by S name , which includes segmented words S name ={C1, C2, …, C n}, C n represents the nth segmented word of the scene name. The segmented word vector set corresponding to each scene name can be obtained. The segmented word vector of the nth scene name is represented by S

[0163]

[0164] In the above formula, S represents the representation vector of the scene name.

[0165] When obtaining the representation vector of the item information, the segmented word set of any item information can be represented by S sku , S sku ={S1, S2, …, S n}, wherein S n is the nth segmented word of any item information. The segmented word vector set corresponding to any item information is represented by S For example, the segmented word set of any item information is derived from the title name of the item and the attribute information of the item. Thus, the vector representation of all item information is obtained as follows:

[0166]

[0167] In the above formula, k represents that the application scenario information includes k item information, S represents the segmented word vector set corresponding to the jth item information.dis i,j represents a vector corresponding to the i-th segmented word of the j-th commodity in the shopping scenario.

[0168] In this embodiment, the representation vector of the application scenario information can be obtained by combining the representation vector of the aforementioned scenario name and the representation vector of the item information.

[0169] On the basis of the above embodiment, in some embodiments, the above step "determining the positive samples and the negative samples of the preset model according to the distance" can be implemented by the following steps:

[0170] Obtaining the hotspot information with a distance greater than the preset distance as the positive sample of the preset model;

[0171] Selecting the hotspot information with a distance less than or equal to the preset distance as the negative sample of the preset model.

[0172] In this embodiment, the cosine distance can be taken as an example, and the calculation formula of the vector distance is as follows:

[0173]

[0174] In the above formula, dis represents the distance between the representation vector of the hotspot information and the representation vector of the application scenario information, Vec scene represents the representation vector of the application scenario information, Vec T represents the representation vector of the hotspot information.

[0175] For example, the preset distance can be set to 0.95, and when the representation vector of the hotspot information is greater than the preset distance from the representation vector of the application scenario information, the hotspot information will be taken as the positive sample. The remaining hotspot information will be taken as the negative sample.

[0176] Further, in some embodiments, the above step "training the preset model to obtain the first target model according to the training data of the preset model" can be implemented by the following steps:

[0177] Training the preset model according to the positive samples and the negative samples of the preset model to obtain an initial model;

[0178] Scoring the hotspot information with a distance less than or equal to the preset distance according to the initial model to obtain a score corresponding to each hotspot information with a distance less than or equal to the preset distance;

[0179] Selecting the hotspot information with a score greater than a preset score threshold from the hotspot information with a distance less than or equal to the preset distance and updating it to the positive sample;

[0180] Selecting the hotspot information with a score less than or equal to the preset score from the hotspot information with a distance less than or equal to the preset distance and updating it to the negative sample;

[0181] According to the positive sample and the negative sample after the update, the initial model is trained to obtain a first target model.

[0182] In this embodiment, the initial model is obtained by training the first target model through the positive sample and the negative sample. The initial model can be used to predict the correlation scores from the original negative sample, and then the samples with higher scores are selected and added to the original positive sample to update the original positive sample to obtain the positive sample after the update, and the remaining samples are used as the negative sample to update the original negative sample. Then, the initial model is trained according to the positive sample and the negative sample after the update to obtain the first target model.

[0183] For example, the samples with scores higher than a preset score threshold are selected and added to the original positive sample. The correlation score refers to the prediction result of the initial model, which represents the degree of correlation between the hot information and the application scenario information.

[0184] For example, taking “the strongest cold wave is coming” and the related discussion “# the strongest cold wave is coming # friends ~ share your cold-weather equipment” and “the strongest cold wave is coming! What should I do if my skin is severely dehydrated?” as hot information, these hot information are related to the application scenario information “cold-weather _ hydration”, and can be added to the positive sample.

[0185] In this embodiment, the initial model can be iteratively trained n times, that is, after each iteration, the positive sample and the negative sample of the last iteration are updated, and after n iterations, a batch of annotated positive samples and negative samples are obtained. The value of n can be determined by an experience threshold. After obtaining the annotated positive and negative samples, the latest version of the model is retrained based on the model structure of the initial model, thereby obtaining the final first target model.

[0186] The present embodiment can enhance the generalization ability of the first target model by updating the positive sample and the negative sample after each iteration, so that the first target model can be applied to a wider range of data scenarios, and on the other hand, it can also be used to mine potential item demand of hot information.

[0187] In some embodiments, the above step S203 can be implemented by the following steps:

[0188] According to the application scenario information, a set of hot information associated with the application scenario information is obtained from a second website;

[0189] According to the set of hot information, a second target model is obtained by training a preset classification model;

[0190] According to the target hot information, the application scenario information and the second target model, a potential item demand is obtained.

[0191] In this embodiment, the search words can be used to search on the second website based on the application scenario information to obtain hot information related to each application scenario information. For example, the search word is "male autumn and winter outdoor travel" when the application scenario information is "male autumn and winter outdoor travel", and the hot information such as "Heilongjiang Shengge Outdoor Travel Network" and "My Travel Equipment" related to the application scenario information can be obtained on the second website as training data to train the preset classification model.

[0192] The training data, i.e., the hot information, needs to be vectorized to obtain a feature vector, which is then used as an input of the preset classification model to train a second target model. After obtaining the second target model, any hot information can be crawled from the second website to predict whether it contains potential item demand.

[0193] For example, the preset classification model can be a machine learning multi-classification model, such as a Naive Bayes model or a decision tree, or can be based on deep learning, such as a common double-tower model.

[0194] Fig. 3 The flowchart of the second embodiment of the hot information processing method provided by the embodiment of the present application is shown in Fig. 3 The method includes the following steps:

[0195] S301, hot event library construction.

[0196] S302, related hot spot identification.

[0197] S303, reasoning of potential item demand based on hot information.

[0198] In this embodiment, the application scenario information contained in the item library of the shopping website and the recent behavior log of the user can be used to construct a hot event library. The related hot spot can refer to the hot information related to the application scenario information. By constructing the hot event library, on the one hand, the workload of manually mining hot spots can be saved, and on the other hand, the hot spots can be explained. The related hot spot identification can identify the hot information related to the shopping website from the hot information of any data source, effectively reducing the cost of manual screening. The reasoning of potential item demand can realize the automatic association of hot information and items, accurately find the item demand behind the user, and improve the item click rate and conversion rate of the shopping website.

[0199] The following is an apparatus embodiment of the present application, which can be used to execute the method embodiments of the present application. For details not disclosed in the apparatus embodiments of the present application, please refer to the method embodiments of the present application.

[0200] Fig. 4A structural schematic diagram of a hotspot information processing apparatus provided by an embodiment of the present application is shown in FIG. 4. The hotspot information processing apparatus can be integrated on a server, or can be independent of the server and cooperates with the server to implement the present solution. As shown in FIG. 4, the hotspot information processing apparatus 40 includes an article obtaining module 41, an information association module 42, and a demand obtaining module 43. Fig. 4

[0201] The article obtaining module 41 is configured to obtain a hotspot article of a first website, and determine application scenario information corresponding to the hotspot article. The information association module 42 is configured to obtain hotspot information from a second website, and determine target hotspot information associated with the application scenario information from the hotspot information. The demand obtaining module 43 is configured to obtain potential article demand according to the target hotspot information and the application scenario information.

[0202] The hotspot article is an article in the first website whose user attention reaches a preset threshold, and the potential article demand is used to indicate an article that a user is interested in in the first website.

[0203] In some embodiments, the article obtaining module 41 can be specifically configured to:

[0204] obtain historical behaviors of a user on articles in the first website;

[0205] select the hotspot article from the articles in the first website according to a number of the historical behaviors and a preset threshold.

[0206] The historical behaviors include at least one of a browsing behavior, a search behavior, an ordering behavior, and a collection behavior.

[0207] In some embodiments, the article obtaining module 41 can be specifically configured to:

[0208] obtain article information of the hotspot article;

[0209] obtain application scenario information of the hotspot article according to title information and attribute information.

[0210] The article information includes the title information and the attribute information.

[0211] Optionally, in some embodiments, the article obtaining module 41 can be specifically configured to:

[0212] extract scene segmentation from the title information;

[0213] obtain an attribute value corresponding to a preset attribute in the attribute information;

[0214] combine the scene segmentation and the attribute value to obtain the application scenario information.

[0215] ​The scene segmentation is a word describing an application scenario of a hot item.

[0216] In some embodiments, the information association module 42 can be specifically configured to:

[0217] training the preset model according to the hot information and the application scenario information to obtain a first target model;

[0218] determining target hot information associated with the application scenario information from the hot information according to the first target model.

[0219] Optionally, in some embodiments, the information association module 42 can be specifically configured to:

[0220] vectorizing the hot information to obtain a representation vector of the hot information;

[0221] vectorizing the application scenario information to obtain a representation vector of the application scenario information;

[0222] obtaining a distance between the representation vector of the hot information and the representation vector of the application scenario information;

[0223] determining positive samples and negative samples of the preset model according to the distance;

[0224] training the preset model according to training data of the preset model to obtain the first target model.

[0225] The positive samples and the negative samples are used as the training data of the preset model.

[0226] Optionally, in some embodiments, the information association module 42 can be specifically configured to:

[0227] segmenting the hot information to obtain at least one hot segment;

[0228] vectorizing each hot segment to obtain a segment vector corresponding to each hot segment;

[0229] obtaining the representation vector of the hot information according to the segment vector corresponding to each hot segment.

[0230] Optionally, in some embodiments, the information association module 42 can be specifically configured to:

[0231] obtaining a scene name of the application scenario information and an item contained in the application scenario information;

[0232] vectorizing the scene name to obtain a representation vector of the scene name;

[0233] vectorizing the item information to obtain a representation vector of the item information.

[0234] According to the scene name representation vector and the item information representation vector, a representation vector of application scene information is obtained.

[0235] The application scene information includes at least one item information.

[0236] Optionally, in some embodiments, the information association module 42 can be specifically configured to:

[0237] Obtain hotspot information with a distance greater than a preset distance as positive samples of a preset model;

[0238] Select hotspot information with a distance less than or equal to the preset distance as negative samples of the preset model.

[0239] Optionally, in some embodiments, the information association module 42 can be specifically configured to:

[0240] According to the positive samples and the negative samples of the preset model, an initial model is obtained by training the preset model;

[0241] According to the initial model, the hotspot information with a distance less than or equal to the preset distance is scored to obtain a score corresponding to each hotspot information with a distance less than or equal to the preset distance;

[0242] From the hotspot information with a distance less than or equal to the preset distance, hotspot information with a score greater than a preset score threshold is selected and updated to the positive samples;

[0243] From the hotspot information with a distance less than or equal to the preset distance, hotspot information with a score less than or equal to the preset score is selected and updated to the negative samples;

[0244] According to the updated positive samples and negative samples, the initial model is trained to obtain a first target model.

[0245] In some embodiments, the demand obtaining module 43 can be specifically configured to:

[0246] According to the application scene information, a set of hotspot information associated with the application scene information is obtained from a second website;

[0247] According to the set of hotspot information, a second target model is obtained by training a preset classification model;

[0248] According to the target hotspot information, the application scene information and the second target model, potential item demand is obtained.

[0249] The device provided by the embodiments of the present application can be used to execute the method in the above-mentioned embodiments, and has similar implementation principles and technical effects, which will not be described here.

[0250] It should be noted that the division of the various modules of the above apparatus is only a logical functional division, and all or part of the modules can be integrated into one physical entity or physically separated in actual implementation. The modules can all be implemented in the form of software invoked by a processing element, for example, the item acquisition module can be stored in the memory of the apparatus in the form of program code, and the functions of the item acquisition module are invoked and executed by a processing element of the apparatus. The implementation of other modules is similar.

[0251] In the above embodiments, all or part can be implemented by software, hardware, firmware, or any combination thereof. When implemented by software, all or part can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to the embodiments of the present application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another, for example, the computer instructions can be transferred from one website, computer, server or data center to another website, computer, server or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) mode. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. integrated with one or more available media. The available media can be magnetic media (such as floppy disk, hard disk, magnetic tape), optical media (such as DVD), or semiconductor media (such as solid state disk (SSD)), etc.

[0252] Fig. 5 The structure of the server provided by the embodiments of the present application is shown in the schematic diagram. As shown in the figure, the server 50 includes at least one processor 51, a memory 52, a bus 53 and a communication interface 54. Fig. 5

[0253] Among them: the processor 51, the communication interface 54 and the memory 52 complete the communication between each other through the bus 53.

[0254] The communication interface 54 is used for communication with other devices. For example, the communication interface 54 can communicate with the server of the second website to obtain the hot spot information from the second website.

[0255] The processor 51 is used for executing the computer execution instructions stored in the memory 52, and can execute the related steps in the method described in the above embodiments. ​

[0256] The processor can be a central processing unit. The one or more processors included in the server can be processors of the same type, such as one or more CPUs; or can be processors of different types, such as one or more CPUs and one or more ASICs.

[0257] The memory is used to store computer execution instructions. The memory can include a high-speed RAM memory, and can also include a non-volatile memory, such as at least one disk memory.

[0258] The embodiment also provides a readable storage medium, and the readable storage medium stores computer instructions. When at least one processor of the server executes the computer instructions, the server executes the hotspot information processing method provided by the various embodiments.

[0259] The embodiment also provides a program product, and the program product includes computer instructions stored in a readable storage medium. At least one processor of the server can read the computer instructions from the readable storage medium, and the at least one processor executes the computer instructions to enable the server to implement the hotspot information processing method provided by the various embodiments.

[0260] In the present application, "at least one" means one or more, and "multiple" means two or more. The "and / or" describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent the following three cases: A exists alone, A and B exist together, and B exists alone, where A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects before and after it; in the formula, the character " / ", represents a "division" relationship between the associated objects before and after it. "At least one of the following" or similar expressions means any combination of these items, including any combination of single item or multiple items. For example, at least one of a, b, or c, can represent a, b, c, a-b, a-c, b-c, or a-b-c, where a, b, and c can be single or multiple.

[0261] It can be understood that the various numbers involved in the embodiments of the present application are only distinguished for the convenience of description, and are not used to limit the scope of the embodiments of the present application. In the embodiments of the present application, the size of the serial number of the above processes does not mean the order of execution, and the execution order of the processes should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0262] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for processing hotspot information, characterized in that: The method comprises the following steps: acquiring a hot item of a first website, determining application scenario information corresponding to the hot item, the hot item being an item in the first website whose user attention reaches a preset threshold; acquiring hot information from a second website, and determining target hot information associated with the application scenario information in the hot information; acquiring potential item demand according to the target hot information and the application scenario information, the potential item demand being used to indicate an item of interest to a user in the first website.

2. The method of claim 1, wherein, The acquiring of the hot item of the first website comprises the following steps: acquiring historical behavior of the user to items in the first website, the historical behavior comprising at least one of browsing behavior, search behavior, ordering behavior and collection behavior; selecting a hot item from items of the first website according to a number of the historical behavior and the preset threshold.

3. The method of claim 1, wherein, The determining of the application scenario information corresponding to the hot item comprises the following steps: acquiring item information of the hot item, the item information comprising title information and attribute information; acquiring application scenario information of the hot item according to the title information and the attribute information.

4. The method of claim 3, wherein, The acquiring of the application scenario information of the hot item according to the title information and the attribute information comprises the following steps: extracting scene segmentation from the title information, the scene segmentation being a word describing an application scenario of the hot item; acquiring a preset attribute in the attribute information to obtain an attribute value corresponding to the preset attribute; combining the scene segmentation and the attribute value to obtain the application scenario information.

5. The method of claim 1, wherein, The determining of the target hot information associated with the application scenario information in the hot information comprises the following steps: training a preset model according to the hot information and the application scenario information to obtain a first target model; determining the target hot information associated with the application scenario information from the hot information according to the first target model.

6. The method of claim 5, wherein, The training of the preset model according to the hot information and the application scenario information to obtain a first target model comprises the following steps: vectorizing the hot information to obtain a representation vector of the hot information; vectorizing the application scenario information to obtain a representation vector of the application scenario information; acquiring a distance between the representation vector of the hot information and the representation vector of the application scenario information; determining positive samples and negative samples of the preset model according to the distance, the positive samples and the negative samples being used as training data of the preset model; training the preset model according to the training data of the preset model to obtain a first target model.

7. The method of claim 6, wherein, The vectorizing of the hot information to obtain a representation vector of the hot information comprises the following steps: segmenting the hot information to obtain at least one hot segmentation; vectorizing each hot segmentation to obtain a segmentation vector corresponding to each hot segmentation; acquiring the representation vector of the hot information according to the segmentation vector corresponding to each hot segmentation.

8. The method of claim 6, wherein, The vectorizing of the application scenario information to obtain a representation vector of the application scenario information comprises the following steps: acquire a scene name of the application scenario information and an item included in the application scenario information, the application scenario information including at least one item information; vectorize the scene name to obtain a representation vector of the scene name; vectorize the item information to obtain a representation vector of the item information; obtain a representation vector of the application scenario information according to the representation vector of the scene name and the representation vector of the item information.

9. The method of claim 6, wherein, The determining the positive samples and the negative samples of the preset model according to the distance comprises: acquiring hotspot information with a distance greater than a preset distance as the positive sample of the preset model; selecting hotspot information with a distance less than or equal to a preset distance as the negative sample of the preset model.

10. The method of claim 9, wherein, The training the preset model according to the training data of the preset model to obtain a first target model comprises: training the preset model according to the positive sample and the negative sample of the preset model to obtain an initial model; scoring the hotspot information with a distance less than or equal to a preset distance according to the initial model to obtain a score corresponding to each hotspot information with a distance less than or equal to a preset distance; selecting hotspot information with a score greater than a preset score threshold from the hotspot information with a distance less than or equal to a preset distance and updating the hotspot information to the positive sample; selecting hotspot information with a score less than or equal to a preset score from the hotspot information with a distance less than or equal to a preset distance and updating the hotspot information to the negative sample; training the initial model according to the updated positive sample and negative sample to obtain the first target model.

11. The method of claim 1, wherein, The acquiring the potential item demand according to the target hotspot information and the application scenario information comprises: acquiring a set of hotspot information associated with the application scenario information from the second website according to the application scenario information; training a preset classification model to obtain a second target model according to the set of hotspot information; acquiring the potential item demand according to the target hotspot information, the application scenario information and the second target model.

12. A processing apparatus of a hot spot information, characterized by, comprise: an item acquisition module configured to acquire a hotspot item of a first website, determine application scenario information corresponding to the hotspot item, and determine the hotspot item as an item with a user attention degree reaching a preset threshold in the first website; an information association module configured to acquire hotspot information from a second website, and determine target hotspot information associated with the application scenario information from the hotspot information; a demand acquisition module configured to acquire a potential item demand according to the target hotspot information and the application scenario information, and use the potential item demand to indicate an item of interest of a user in the first website.

13. A server, characterized by comprise: a processor and a memory connected with the processor in communication; the memory stores computer execution instructions; the processor executes the computer execution instructions stored in the memory to implement the method in any one of claims 1-11.

14. A readable storage medium, characterized by, The readable storage medium stores computer instructions, and the computer instructions are executed by the processor to implement the method in any one of claims 1-11.

15. A program product comprising computer instructions, characterized in that, The computer instructions, when executed by the processor, implement the method of any one of claims 1-11.

Images