Inter-community information fusion methods, devices, computer equipment and storage media

Abstract

This application relates to a method, apparatus, computer device, storage medium, and computer program product for information fusion between communities. The method includes: acquiring currently published text; extracting word vectors corresponding to the currently published text; obtaining a classification result based on the word vectors using a preset question classification model; wherein the preset question classification model is trained based on a preset question solution database, which includes historical information from all communities; the classification result includes a question and a solution; when the classification result is a question, matching the word vectors in the preset question solution database to obtain the current solution text. This allows users to obtain corresponding solutions immediately, addressing the problem of information asymmetry between different communities and the waste of public resources and manpower caused by users asking the same question, and reducing the situation where users ask questions but do not receive answers.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence natural language understanding technology, and in particular to a method, apparatus, computer device and storage medium for inter-community information fusion. Background Technology

[0002] With the development of the internet and big data technologies, and based on the deep integration of information technology and public life platforms, smart community apps provide residents with a convenient information window. Residents can use these apps for online repair requests, epidemic prevention registration, utility bill payments, service notifications, policy interpretations, and other convenient services. Furthermore, features similar to community forums facilitate communication between the community and its residents. Leveraging internet technology, these community apps offer people a better living experience and greater convenience in their work.

[0003] Currently, traditional community apps require residents to register and be verified by their respective communities. Users can only access forum information and services specific to their local community and cannot access information from other communities. However, users from different communities may have posted the same or similar repair requests or help requests; this data cannot be shared, leading to a significant waste of public resources and manpower. Furthermore, communities with low user activity may experience user churn due to limited information. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, device, computer equipment, and storage medium for inter-community information fusion that can quickly find solutions for users who ask questions based on the experience data of all communities on the community APP, in order to address the above-mentioned technical problems.

[0005] Firstly, this application provides a method for information fusion between communities, the method comprising:

[0006] Get the currently published text;

[0007] Based on the currently published text, extract the word vectors corresponding to the currently published text;

[0008] Based on the word vectors, a pre-defined question classification model is used to obtain the classification result corresponding to the word vectors; wherein, the pre-defined question classification model is trained based on a pre-defined question solution database, and the pre-defined question solution database includes historical information of all communities; the classification result includes questions and solutions;

[0009] When the classification result is a problem, the current solution text is obtained by matching the word vector in the preset problem solution database.

[0010] In one embodiment, extracting word vectors corresponding to the currently published text based on the currently published text includes:

[0011] Obtain the embedding vectors corresponding to each word segment of the currently published text, and concatenate them to obtain the overall embedding vector corresponding to the currently published text;

[0012] Calculate the term frequency-inverse file frequency value of each word segment of the currently published text;

[0013] After filtering the overall embedding vector based on the word frequency-inverse file frequency value of each word segment, the word vector corresponding to the currently published text is extracted.

[0014] In one embodiment, the process of establishing the preset problem-solving solution database includes:

[0015] Obtain a preset publishing text and annotate the preset publishing text with problems or solutions;

[0016] A correspondence is created between the preset published texts labeled as problems and the preset published texts labeled as solutions, thus obtaining the preset problem-solution database.

[0017] In one embodiment, after obtaining the preset published text and labeling the preset published text with problems or solutions, and before creating a correspondence between the preset published text labeled with problems and the preset published text labeled with solutions, the method further includes:

[0018] Create a sentiment dictionary and assign sentiment weights to the word segments in the sentiment dictionary;

[0019] If there are words in the preset published text that belong to the sentiment dictionary, sentiment tags will be added to the corresponding questions or solutions based on the sentiment weights of the words.

[0020] In one embodiment, the training process of the preset problem classification model includes:

[0021] Obtain a labeled training dataset, which includes training texts, each of which is obtained from the preset problem solution database;

[0022] The preset problem classification model is trained based on the labeled training dataset.

[0023] In one embodiment, after obtaining the current solution text by matching the word vector in the preset problem solution database when the classification result is a problem, the method further includes:

[0024] Obtain feedback parameters based on the current solution text;

[0025] Create a correspondence between the currently published text, the currently resolved text, and the feedback parameters, and store them in the preset problem-solving database.

[0026] Secondly, this application also provides an inter-community information fusion device, the device comprising:

[0027] The retrieval module is used to retrieve the currently published text.

[0028] The extraction module is used to extract word vectors corresponding to the currently published text based on the currently published text.

[0029] The classification module is used to obtain the classification result corresponding to the word vector by using a preset question classification model; wherein, the preset question classification model is trained based on a preset question solution database, and the preset question solution database includes historical information of all communities; the classification result includes the question and the solution.

[0030] The matching module is used to obtain the current solution text by matching the word vector in the preset problem solution database when the classification result is a problem.

[0031] In one embodiment, the extraction module is used to obtain the embedding vectors corresponding to each word segment of the currently published text, and concatenate them to obtain the overall embedding vector corresponding to the currently published text;

[0032] It is also used to calculate the term frequency-inverse file frequency value of each segment of the currently published text;

[0033] It is also used to filter the overall embedding vector based on the word frequency-inverse file frequency value of each word segment, and then extract the word vector corresponding to the currently published text.

[0034] In one embodiment, the process of establishing the preset problem-solving method database in the classification module includes:

[0035] Obtain a preset publishing text and annotate the preset publishing text with problems or solutions;

[0036] A correspondence is created between the preset published texts labeled as problems and the preset published texts labeled as solutions, thus obtaining the preset problem-solution database.

[0037] In one embodiment, the process of establishing the preset problem-solving method database in the classification module further includes:

[0038] Create a sentiment dictionary and assign sentiment weights to the word segments in the sentiment dictionary;

[0039] If there are words in the preset published text that belong to the sentiment dictionary, sentiment tags will be added to the corresponding questions or solutions based on the sentiment weights of the words.

[0040] In one embodiment, the training process of the preset problem classification model in the classification module includes:

[0041] Obtain a labeled training dataset, which includes training texts, each of which is obtained from the preset problem solution database;

[0042] The preset problem classification model is trained based on the labeled training dataset.

[0043] In one embodiment, the device further includes:

[0044] The feedback module is used to obtain feedback parameters based on the current solution text, create a correspondence between the current published text, the current solution text, and the feedback parameters, and store them in the preset problem solution database.

[0045] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the method described above.

[0046] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps of the above-described method.

[0047] The aforementioned information fusion method, apparatus, computer equipment, and storage medium for inter-community communication can extract keyword vectors from the text currently posted by a user. Based on these keyword vectors, a pre-defined question classification model is used to obtain classification results, including both the question and the solution. Therefore, the classification results can be used to determine whether a user is asking a question. If a question is identified, the corresponding solution is matched against a pre-defined question-solution database based on the keyword vectors and recommended to the user, ensuring they receive the solution immediately. This solution not only addresses the problem of wasted public resources and manpower caused by information asymmetry between different communities in a community app, where technical staff repeatedly answer or resolve the same questions raised by users in different communities, but also ensures that solutions are provided even in less active community forums, preventing users from leaving the community without answers. Attached Figure Description

[0048] Figure 1 This is a diagram illustrating the application environment of an inter-community information fusion method in one embodiment.

[0049] Figure 2 This is a flowchart illustrating an inter-community information fusion method in one embodiment;

[0050] Figure 3 This is a flowchart illustrating the inter-community information fusion method in another embodiment;

[0051] Figure 4 This is a flowchart illustrating the inter-community information fusion method in another embodiment;

[0052] Figure 5 This is a flowchart illustrating the inter-community information fusion method in another embodiment;

[0053] Figure 6 This is a structural block diagram of an inter-community information fusion device in one embodiment;

[0054] Figure 7 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0056] The inter-community information fusion method provided in this application embodiment can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or placed on a cloud or other network server. Specifically, server 104 obtains the currently published text through terminal 102; extracts the word vectors corresponding to the currently published text; and obtains the classification result corresponding to the word vectors using a preset question classification model. The preset question classification model is trained based on a preset question solution database, which includes historical information from all communities. The classification result includes both the question and the solution. When the classification result is a question, the current solution text is obtained by matching the word vectors in the preset question solution database. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. Portable wearable devices can be smartwatches, smart bracelets, head-mounted devices, etc. Server 104 can be implemented using a standalone server or a server cluster composed of multiple servers.

[0057] In one embodiment, such as Figure 2 As shown, a method for information fusion between communities is provided, which can be applied to... Figure 1 Taking server 104 as an example, the following steps are included:

[0058] Step 202: Obtain the currently published text.

[0059] The currently published text can be a sentence, a paragraph, or an entire document. Specifically, the currently published text originates from the community app running on terminal 102, and can be a user-submitted repair request, or a post or comment published on a forum or online community platform. The method by which server 104 obtains the currently published text from terminal 102 is not limited in the embodiments of this application.

[0060] Step 204: Extract the word vectors corresponding to the currently published text.

[0061] Since the currently published text uses everyday natural language, it may contain many common words. These words serve for sentence completeness but do not contribute to semantic expression. Therefore, to improve the efficiency of subsequent data processing, it is necessary to extract important words from each segment of the currently published text to determine the word vectors that correspond to the current published text and have scene-specific and discriminative capabilities. This can be understood as determining the scene corresponding to the currently published text based on the word vectors, or quickly finding similar scene expressions based on the word vectors.

[0062] In one embodiment, based on natural language segmentation theory, the currently published text is segmented to determine each word in the text. The segmentation method can be at least one of dictionary-based segmentation algorithms and statistical machine learning algorithms. Dictionary-based segmentation algorithms mainly include forward maximum matching, backward maximum matching, and bidirectional matching segmentation, while statistical machine learning algorithms mainly include support vector machines and deep learning. Embedding vectors can encode natural language text using low-dimensional vectors while preserving its meaning, making them suitable for segmenting the currently published text.

[0063] In one embodiment, the embedding vectors corresponding to each word segment of the currently published text are obtained and concatenated to obtain the overall embedding vector corresponding to the currently published text; the term frequency-inverse file frequency value of each word segment of the currently published text is calculated; the overall embedding vector is filtered according to the term frequency-inverse file frequency value of each word segment, and the word vector corresponding to the currently published text is extracted.

[0064] Step 206: Based on the word vectors, use a preset question classification model to obtain the classification results corresponding to the word vectors; wherein, the preset question classification model is trained based on a preset question solution database, which includes historical information of all communities; the classification results include questions and solutions.

[0065] Specifically, based on the word vectors corresponding to the currently published text calculated above, these vectors are input into a preset question classification model to obtain the classification result corresponding to the word vectors. The classification result includes the question and the solution, which are the two outputs of the preset question classification model.

[0066] In one embodiment, the preset problem classification model is trained based on a preset problem solution database. The training process includes obtaining a labeled training dataset from the preset problem solution database and then training the preset problem classification model based on the labeled training dataset.

[0067] In one embodiment, the preset problem-solving solution database includes historical information from all communities in the community app. Specifically, preset published texts are first obtained from the historical information of all communities, and these texts are labeled with either problems or solutions. A correspondence is then established between the preset published texts labeled as problems and those labeled as solutions, resulting in the preset problem-solving solution database. In one embodiment, the preset published texts labeled as problems and those labeled as solutions in the preset problem-solving solution database also include sentiment tags.

[0068] Step 208: When the classification result is a problem, the current solution text is obtained by matching the word vector in the preset problem solution database.

[0069] Specifically, when the currently published text is classified as a question according to the preset question classification model, it means that the user has asked a question in the community app. Then, based on the word vector corresponding to the currently published text, similar question word vectors will be obtained by matching the word frequency and inverse file frequency values ​​in the preset question solution database, and the solution text corresponding to the similar question word vectors will be pushed to the user.

[0070] The aforementioned information fusion method between communities can extract keyword vectors from the text currently posted by users, and use a preset question classification model based on these word vectors to determine whether the user is asking a question. If so, it matches the corresponding solution in a preset question solution database based on the word vectors and recommends it to the user, so that the user can get the corresponding solution as soon as possible. This solves the problem of information not being shared between different communities and users asking the same question, which leads to a waste of public resources and manpower, and also reduces the situation where users ask questions but do not get answers.

[0071] In one embodiment, such as Figure 3 As shown, step 204, which involves extracting word vectors corresponding to the currently published text, includes:

[0072] Step 302: Obtain the embedding vectors corresponding to each word segment of the currently published text, and concatenate them to obtain the overall embedding vector corresponding to the currently published text.

[0073] The embedding vector is the encoding vector corresponding to each word in the currently published text. The embedding vector corresponding to each word in the currently published text can be obtained from a preset table of embedding vectors corresponding to word segments, or it can be obtained through a method recognized by those skilled in the art to obtain the embedding vector corresponding to the word segments. The specific process is not limited in the embodiments of this application.

[0074] Furthermore, after obtaining the embedding vectors corresponding to each word segment of the currently published text, the embedding vectors corresponding to each word segment are concatenated according to the order of text expression to obtain the overall embedding vector of the currently published text. It can be understood that each currently published text corresponds to a unique overall embedding vector.

[0075] Step 304: Calculate the term frequency-reverse file frequency value of each word segment of the currently published text.

[0076] The Term Frequency-Inverse Document Frequency (TF-IDF) is a weighted evaluation metric used in information retrieval and prospecting to assess the importance of a word segment to a document within a document set or corpus. The importance of a word segment increases proportionally to the number of times it appears in the document, but decreases inversely proportionally to its frequency in the corpus. The number of times a word segment appears in the document is the term frequency (TF), and the frequency of the word segment in the corpus is the inverse document frequency (IDF). The product of these two values ​​yields the TF-IDF value for the word segment.

[0077] Specifically, the term frequency-inverse file frequency value of each word in the currently published text is calculated.

[0078] Step 306: After filtering the overall embedding vector based on the word frequency-inverse file frequency value of each word segment, extract the word vector corresponding to the currently published text.

[0079] When filtering the overall embedded vector based on the term frequency-reverse file frequency values ​​of each word segment, the filtering method can be to use a preset term frequency-reverse file frequency threshold. This preset threshold can be set based on actual experience and is not limited. Alternatively, other methods acceptable to those skilled in the art can also be used for filtering.

[0080] Specifically, the embedding vectors corresponding to word segments that are lower than the preset word frequency-inverse file frequency threshold are filtered out, and the overall embedding vector obtained by arranging the embedding vectors corresponding to word segments that are higher than the preset word frequency-inverse file frequency threshold is used as the word vector corresponding to the currently published text.

[0081] In one embodiment, such as Figure 4 As shown, the process of establishing the preset problem-solving solution database in step 206 includes:

[0082] Step 401: Obtain the preset published text and annotate the preset published text with problems or solutions.

[0083] The preset published text and the currently published text are obtained in the same way and from the same source. The preset published text comes from the historical information of all communities in the community APP running on terminal 102. The aforementioned historical information can be user-submitted repair requests, or posts and comments published in forums or online communities.

[0084] Specifically, the system retrieves all pre-published texts from the community app and annotates them. The annotation types include questions and solutions. Annotation can be done manually, by machine, or using third-party data annotation services such as Amazon MechanicalTurk or the Huicui app. During the annotation process, corresponding question and solution dictionaries need to be created in advance. If a word segment in a pre-published text belongs to the question dictionary, the text is annotated as a question; if a word segment in a pre-published text belongs to the solution dictionary, the text is annotated as a solution. The methods for creating the question and solution dictionaries can be adopted in a manner recognized by those skilled in the art, and will not be elaborated upon in this embodiment.

[0085] Furthermore, for the pre-defined published texts obtained from the historical information of all communities, k problem labels and k solution labels can be collected. It's understandable that the value of k is not fixed; the larger the value of k in the collected historical data, the more preliminary labeled data is obtained, thus leading to more accurate recommendations for the scenario.

[0086] Step 404: Create a correspondence between the preset published text labeled as a problem and the preset published text labeled as a solution, and obtain a preset problem solution database.

[0087] In one embodiment, word vectors are extracted from preset published texts labeled as problems and preset published texts labeled as solutions. Specifically, the embedding vectors corresponding to each word segment of each preset published text are obtained and concatenated to obtain the overall embedding vector for each preset published text; the term frequency-inverse file frequency value of each word segment of each preset published text is calculated; the overall embedding vector of each preset published text is filtered based on the term frequency-inverse file frequency value of each word segment to extract the word vectors corresponding to each preset published text. The process of extracting the corresponding word vectors can be referred to the description above and will not be repeated here.

[0088] The preset problem-solution database is a database that finds corresponding solutions by indexing problems; that is, each problem corresponds to one solution. In one embodiment, a correspondence is created between preset published texts labeled as problems and preset published texts labeled as solutions to obtain the preset problem-solution database. The preset problem-solution database contains a one-to-one correspondence table between problems and solutions. Each row of the table includes the preset published text labeled as a problem and its word vector, and the preset published text labeled as a solution and its word vector. The number of rows in the table is k.

[0089] In one embodiment, such as Figure 4As shown, after step 401 and before step 404, it further includes:

[0090] Step 402: Create an emotion dictionary and assign emotion weights to the word segments corresponding in the emotion dictionary.

[0091] Specifically, the word segments in the emotion dictionary are all expressions of emotion-related words, which are obtained by pre-induction after data collection and collation. Among them, according to the writing habits, the word segments in the emotion dictionary also have degree distinctions and need to be assigned corresponding emotion weights to describe their emotional degrees. Among them, when the emotion weight is positive, it represents that the word segment is a positive expression; when the emotion weight is negative, it represents that the word segment is a negative expression. In addition, when the weight is larger, it represents a greater degree of bias; when the weight is smaller, it represents a smaller degree of bias. For example, the character "weird" is mostly a negative expression in most cases, but the sentence "The smell is weirdly good" belongs to a positive expression. Therefore, the emotion weight of the word "weird" will be relatively negative, about -0.75.

[0092] Step 403: If there are word segments in the preset published text that belong to the emotion dictionary, add emotion labels to the corresponding annotations of the problems or solutions according to the emotion weights corresponding to the word segments.

[0093] Specifically, if there are word segments in the preset published text that belong to the emotion dictionary, add emotion labels to the corresponding annotations of the problems or solutions in the preset problem-solving method database. Among them, the emotion labels include positive expressions and corresponding emotion weights, and negative expressions and corresponding emotion weights.

[0094] In one embodiment, if there are no word segments in the preset published text that belong to the emotion dictionary, no emotion labels are added to the preset published text.

[0095] In one embodiment, if there are more than two word segments in the preset published text that belong to the emotion dictionary, add emotion labels to the corresponding annotations of the problems or solutions in the preset problem-solving method database according to the product of the emotion weights corresponding to the word segments.

[0096] In one embodiment, as Figure 5 shown, the training process of the preset problem classification model in step 206 includes:

[0097] Step 502: Obtain a labeled training data set, and the training data set includes each training text, and each training text is obtained from the preset problem-solving method database.

[0098] Specifically, the training data set of the preset problem classification model comes from the preset problem-solving method database. The training text includes a set of corresponding preset published texts labeled as problems and preset published texts labeled as solutions. It can be understood that the training text can be its corresponding word vector.

[0099] Step 504: Train the preset problem classification model based on the labeled training dataset.

[0100] Specifically, based on the labeled training dataset, the classification of problems and solutions are used as training tasks to train the network model, resulting in a preset problem classification model. This preset problem classification model can be used to classify whether text belongs to a problem or a solution. In one embodiment, the training process divides the training dataset into a training set and a test set in a 3:7 ratio. A training accuracy > 90% indicates successful model training, resulting in the preset problem classification model.

[0101] In one embodiment, the training method for the model can be a Convolutional Neural Network (CNN) or a Recursive Neural Network (RNN), or it can be TextRCNN (Recurrent Convolutional Neural Networks for Text Classification). In this embodiment, the TextRCNN model algorithm is used for training because it is better at capturing longer sequence information. During the training process, TextRCNN first uses a bidirectional RNN to obtain the forward and backward context representations of each word segment. In this way, the representation of each word segment becomes a concatenation of word vectors and forward and backward context vectors. Finally, the same convolutional and pooling layers as TextCNN are added to complete the model structure.

[0102] In one embodiment, such as Figure 2 As shown, after step 208, the method further includes:

[0103] Step 210: Obtain feedback parameters based on the current solution text.

[0104] Specifically, after a user receives the recommended solution text, their feedback parameters are obtained. The method for obtaining these feedback parameters is not unique; it can be obtained through methods such as setting up likes or feedback in the corresponding location within the community app, or by following up to see if the problem has been resolved. There is no limitation on this method.

[0105] Step 212: Create a correspondence between the currently published text, the currently resolved text, and the feedback parameters, and store them in the preset problem-solving database.

[0106] Specifically, after obtaining the feedback parameters, a new correspondence will be created between the currently published text, the currently solved text, and the feedback parameters, and stored in the preset problem-solving database. This database will serve as the training set for the preset problem classification model, thereby improving the classification accuracy of the preset problem classification model.

[0107] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0108] Based on the same inventive concept, this application also provides an inter-community information fusion device for implementing the inter-community information fusion method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more embodiments of the inter-community information fusion device provided below can be found in the limitations of the inter-community information fusion method described above, and will not be repeated here.

[0109] In one embodiment, such as Figure 6 As shown, an inter-community information fusion device is provided, comprising: an acquisition module 610, an extraction module 620, a classification module 630, and a matching module 640, wherein:

[0110] Module 610 is used to retrieve the currently published text.

[0111] The extraction module 620 is used to extract the word vectors corresponding to the currently published text.

[0112] The classification module 630 is used to obtain the classification result corresponding to the word vector by using a preset question classification model based on the word vector; wherein, the preset question classification model is trained based on a preset question solution database, which includes historical information of all communities; the classification result includes the question and the solution.

[0113] The matching module 640 is used to match the current solution text in a preset problem solution database based on word vectors when the classification result is a problem.

[0114] In one embodiment, the extraction module 610 is used to obtain the embedding vectors corresponding to each word segment of the currently published text, and concatenate them to obtain the overall embedding vector corresponding to the currently published text; the extraction module 610 is also used to calculate the word frequency-inverse file frequency value of each word segment of the currently published text; the extraction module 610 is also used to filter the overall embedding vector according to the word frequency-inverse file frequency value of each word segment, and then extract the word vector corresponding to the currently published text.

[0115] In one embodiment, the process of establishing a preset problem-solving method database in the classification module 630 includes:

[0116] Retrieve the preset publication text and annotate the preset publication text with problems or solutions;

[0117] Create a mapping between preset posting texts labeled as problems and preset posting texts labeled as solutions to obtain a database of preset problem solutions.

[0118] In one embodiment, the process of establishing the preset problem-solving method database in the classification module 630 further includes:

[0119] Create a sentiment dictionary and assign sentiment weights to the word segments in the sentiment dictionary;

[0120] If the pre-defined published text contains words belonging to the sentiment dictionary, sentiment tags will be added to the corresponding questions or solutions based on the sentiment weight of the words.

[0121] In one embodiment, the training process of the preset problem classification model in the classification module 630 includes:

[0122] Obtain the labeled training dataset, which includes each training text, obtained from a pre-defined problem-solving database;

[0123] Based on the labeled training dataset, a pre-defined problem classification model is trained.

[0124] In one embodiment, such as Figure 6 As shown, the inter-community information fusion device also includes a feedback module 650:

[0125] The feedback module 650 is used to obtain feedback parameters based on the current solution text, create a correspondence between the current published text, the current solution text, and the feedback parameters, and store them in the preset problem solution database.

[0126] Specific limitations regarding the inter-community information fusion device can be found in the limitations of the inter-community information fusion method described above, and will not be repeated here. Each module in the aforementioned inter-community information fusion device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0127] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 7 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores published text data from a pre-defined problem-solving database. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements an inter-community information fusion method.

[0128] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0129] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the above-described inter-community information fusion method.

[0130] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the inter-community information fusion method described above.

[0131] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0132] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0133] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0134] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for information fusion between communities, characterized in that, The method includes: Get the currently published text; Based on the currently published text, extract the word vectors corresponding to the currently published text; Based on the word vectors, a pre-defined question classification model is used to obtain the classification result corresponding to the word vectors; wherein, the pre-defined question classification model is trained based on a pre-defined question solution database, and the pre-defined question solution database includes historical information of all communities; the classification result includes questions and solutions; When the classification result is a problem, the current solution text is obtained by matching the word vector in the preset problem solution database; The feedback parameters returned after obtaining the text of the recommended current solution; Create a correspondence between the currently published text, the currently resolved text, and the feedback parameters, and store them in the preset problem-solving database; The process of establishing the preset problem-solving solution database includes: Retrieve all preset published texts from the community app and annotate the preset published texts with problems or solutions; Create a sentiment dictionary and assign sentiment weights to the word segments in the sentiment dictionary; If there are words in the preset published text that belong to the sentiment dictionary, sentiment tags are added to the corresponding questions or solutions according to the sentiment weights of each word. When the sentiment weight is positive, it means that the word is a positive expression; when the sentiment weight is negative, it means that the word is a negative expression. The larger the sentiment weight, the greater the bias; the smaller the sentiment weight, the less the bias. The sentiment tags include positive expressions and their corresponding sentiment weights, and negative expressions and their corresponding sentiment weights. A correspondence is created between the preset published texts labeled as problems and the preset published texts labeled as solutions, thus obtaining the preset problem-solution database.

2. The inter-community information fusion method according to claim 1, characterized in that, The step of extracting word vectors corresponding to the currently published text includes: Obtain the embedding vectors corresponding to each word segment of the currently published text, and concatenate them to obtain the overall embedding vector corresponding to the currently published text; Calculate the term frequency-inverse file frequency value of each word segment of the currently published text; After filtering the overall embedding vector based on the word frequency-inverse file frequency value of each word segment, the word vector corresponding to the currently published text is extracted.

3. The inter-community information fusion method according to claim 1, characterized in that, The preset problem-solving database is a one-to-one correspondence table containing problems and solutions. Each row of the correspondence table includes a preset published text labeled as a problem and its word vector, and a preset published text labeled as a solution and its word vector.

4. The inter-community information fusion method according to claim 3, characterized in that, If the preset published text contains words belonging to the sentiment dictionary, sentiment tags are added to the corresponding problem or solution annotations based on the sentiment weight of the word, including: If there are two or more word segments in the preset published text that belong to the sentiment dictionary, the corresponding problem or solution will be labeled with a sentiment tag based on the product of the sentiment weights of the word segments.

5. The inter-community information fusion method according to claim 3, characterized in that, The training process of the preset problem classification model includes: Obtain a labeled training dataset, which includes training texts, each of which is obtained from the preset problem solution database; The preset problem classification model is trained based on the labeled training dataset.

6. A community-based information fusion device, characterized in that, The device includes: The retrieval module is used to retrieve the currently published text. The extraction module is used to extract word vectors corresponding to the currently published text based on the currently published text. The classification module is used to obtain the classification result corresponding to the word vector by using a preset question classification model; wherein, the preset question classification model is trained based on a preset question solution database, and the preset question solution database includes historical information of all communities; the classification result includes the question and the solution. The process of establishing the preset problem-solving solution database in the classification module includes: obtaining all preset published texts in the community APP and labeling the preset published texts with problems or solutions; creating a sentiment dictionary and assigning sentiment weights to the words in the sentiment dictionary; if there are words in the preset published text that belong to the sentiment dictionary, adding sentiment tags to the corresponding problem or solution labels according to the sentiment weights of each word, wherein a positive sentiment weight represents a positive expression of the word, a negative sentiment weight represents a negative expression of the word, and the larger the sentiment weight, the greater the bias, and the smaller the sentiment weight, the smaller the bias; the sentiment tags include positive expressions and corresponding sentiment weights, and negative expressions and corresponding sentiment weights; creating a correspondence between the preset published texts labeled as problems and the preset published texts labeled as solutions to obtain the preset problem-solving solution database; The matching module is used to obtain the current solution text by matching the word vector in the preset problem solution database when the classification result is a problem; The feedback module is used to obtain the feedback parameters returned after recommending the current solution text, and to create a correspondence between the current published text, the current solution text, and the feedback parameters, and store them in the preset problem solution database.

7. The inter-community information fusion device according to claim 6, characterized in that, The extraction module is used to obtain the embedding vectors corresponding to each word segment of the currently published text, and concatenate them to obtain the overall embedding vector corresponding to the currently published text. It is also used to calculate the term frequency-inverse file frequency value of each segment of the currently published text; It is also used to filter the overall embedding vector based on the word frequency-inverse file frequency value of each word segment, and then extract the word vector corresponding to the currently published text.

8. The inter-community information fusion device according to claim 6, characterized in that, The training process of the preset problem classification model in the classification module includes: Obtain a labeled training dataset, which includes training texts, each of which is obtained from the preset problem solution database; The preset problem classification model is trained based on the labeled training dataset.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images