Method, device, electronic device and readable storage medium for recommending points of interest

Abstract

The present disclosure provides a method, device, electronic device and readable storage medium for recommending points of interest, which relate to the field of artificial intelligence technology, and in particular to the field of intelligent recommendation. The specific implementation scheme is: extracting a first feature vector of the question information to be recommended through the first representation layer in the preset matching model; determining the similarity between the first feature vector and at least one second feature vector, the second feature vector is a feature vector of the preset point of interest extracted through the second representation layer in the matching model; and determining the target point of interest of the question information to be recommended from the preset point of interest based on the similarity. Based on this scheme, the points of interest in the question can be effectively extracted, which provides a basis for question distribution based on the points of interest in the question, helps to improve the accuracy of question distribution, and improves the answering effect of the question.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to the field of intelligent recommendation technology. Specifically, the present disclosure relates to a method, device, electronic device, and readable storage medium for recommending points of interest. Background Art

[0002] With the rapid development of Internet technology, various question-and-answer platforms have emerged. In question-and-answer platforms, users who ask questions generally submit questions, and the question-and-answer platforms distribute the questions to answering users for answers.

[0003] Whether questions can be accurately distributed to appropriate answering users in the question-and-answer platform will directly affect the effectiveness of answering the questions. Therefore, how to improve the accuracy of question distribution is an important technical issue in the question-and-answer platform. Summary of the invention

[0004] In order to solve at least one of the above-mentioned defects, the present disclosure provides a method, device, electronic device and readable storage medium for recommending points of interest.

[0005] According to a first aspect of the present disclosure, a method for recommending points of interest is provided, the method comprising:

[0006] Extracting a first feature vector of the question information to be recommended through a first representation layer in a preset matching model;

[0007] Determine a similarity between the first feature vector and at least one second feature vector, where the second feature vector is a feature vector of a preset interest point extracted by a second representation layer in the matching model;

[0008] Based on the similarity, a target interest point of the question information to be recommended is determined from the preset interest points.

[0009] According to a second aspect of the present disclosure, a device for recommending points of interest is provided, the device comprising:

[0010] A vector extraction module, used for extracting a first feature vector of the question information to be recommended through a first representation layer in a preset matching model;

[0011] A vector similarity module, used to determine the similarity between the first feature vector and at least one second feature vector, where the second feature vector is a feature vector of a preset interest point extracted by a second representation layer in the matching model;

[0012] The target interest point determination module is used to determine the target interest point of the question information to be recommended from the preset interest points based on similarity.

[0013] According to a third aspect of the present disclosure, an electronic device is provided, the electronic device comprising:

[0014] at least one processor; and

[0015] A memory in communication with the at least one processor; wherein,

[0016] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for recommending points of interest.

[0017] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable a computer to execute the above-mentioned method for recommending points of interest.

[0018] According to a fifth aspect of the present disclosure, a computer program product is provided, including a computer program, which implements the above-mentioned method for recommending points of interest when executed by a processor.

[0019] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0021] Figure 1 is a flowchart of a method for recommending points of interest provided by an embodiment of the present disclosure;

[0022] Figure 2 is a schematic diagram of a process for determining a target point of interest in an optional implementation manner of the present disclosure;

[0023] Figure 3 is a schematic diagram of a process of marking a newly added point of interest in an optional implementation manner of the present disclosure;

[0024] Figure 4 is a flowchart of another method for recommending points of interest provided by an embodiment of the present disclosure;

[0025] Figure 5 is a flowchart of another method for recommending points of interest provided by an embodiment of the present disclosure;

[0026] Figure 6 is a structural schematic diagram of a device for recommending points of interest provided by an embodiment of the present disclosure;

[0027] Figure 7 is a schematic diagram of the structure of another device for recommending points of interest provided by an embodiment of the present disclosure;

[0028] Figure 8 is a structural schematic diagram of another device for recommending points of interest provided by an embodiment of the present disclosure;

[0029] Fig. 9 The block diagram is a block diagram of an electronic device for implementing the method for recommending points of interest according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0030] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0031] The questions submitted by the user who asked the question can reflect the interests of the user who asked the question, and the interests of the user who answered the question can often reflect the questions that he is good at answering. If the points of interest of the answering user can be extracted from the points of interest reflected in the question, then it can be considered that the points of interest of the answering user match the points of interest in the question, and the answering user's answer to the question is effective. It can be seen that if questions can be distributed based on the points of interest in the question, the accuracy of distribution can be improved, and the answering effect of the question can be improved. Therefore, how to effectively extract the points of interest in the question has become an important technical issue.

[0032] The method, device, electronic device and readable storage medium for recommending points of interest provided by the embodiments of the present disclosure are intended to solve at least one of the above technical problems in the prior art.

[0033] Figure 1 A flowchart of a method for recommending points of interest provided by an embodiment of the present disclosure is shown. Figure 1 As shown in , the method can mainly include:

[0034] Step S110: extracting a first feature vector of the question information to be recommended through a first representation layer in a preset matching model;

[0035] Step S120: determining a similarity between the first feature vector and at least one second feature vector, where the second feature vector is a feature vector of a preset interest point extracted by a second representation layer in the matching model;

[0036] Step S130: determining a target interest point for the question information to be recommended from the preset interest points based on the similarity.

[0037] The information of the question to be recommended may be text information of the question to be recommended, and the question to be recommended may be submitted to the question-answering platform by the user asking the question.

[0038] The preset interest points may be existing interest points, and the interest points may be labels of the interest of the questioning user reflected in the question, and the preset interest points may also be interest labels of the answering user. When the interest points extracted for the question to be recommended are covered by the interest points of the answering user, it can be considered that the answering user is good at answering the question to be recommended.

[0039] The matching model may be a network model for predicting the matching degree between the question and the point of interest. The first representation layer in the matching model may include a feature vector extraction model for the question, for extracting the feature vector of the question. The second representation layer in the matching model may include a feature vector extraction model for the point of interest, for extracting the feature vector of the point of interest.

[0040] As an example, the matching model may be a dual-tower model, which may include a first representation layer, a second representation layer, and a matching layer. The matching layer may be used to match the feature vector of the question with the feature vector of the point of interest to obtain the matching degree between the question and the point of interest.

[0041] The first feature vector may be a semantic feature vector of the question information to be recommended. The question information to be recommended may be mapped into numbers and then input into a feature vector extraction model for the question in the first representation layer to obtain the first feature vector.

[0042] The second feature vector may be a semantic feature vector of a preset interest point. The preset interest point may be first mapped into a number and then input into a feature vector extraction model for the question in the second representation layer to obtain the second feature vector.

[0043] As an example, the feature vector extraction model for the problem in the first representation layer and the feature vector extraction model for the problem in the second representation layer can both be a miniaturized bidirectional encoder representation from transformers (A LiteBidirectional Encoder Representations from Transformers, Albert) model.

[0044] There are usually multiple preset interest points, and the second feature vector of each preset interest point can be extracted in advance based on the second representation layer, and a vector library can be constructed based on multiple second feature vectors. As an example, the vector library can be an Approximate Nearest Neighbor (ANN) vector library.

[0045] In the disclosed embodiment, after determining the first feature vector of the question information to be recommended, the similarity between the first feature vector and each second feature vector can be determined, thereby determining the target interest point based on the similarity between the first feature vector and each second feature vector, and the target interest point is the interest point of the question information to be recommended.

[0046] The method provided by the embodiment of the present disclosure extracts the first feature vector of the question information to be recommended through the first representation layer in the preset matching model, determines the similarity between the first feature vector and at least one second feature vector, the second feature vector is the feature vector of the preset interest point extracted through the second representation layer in the matching model, and determines the target interest point of the question information to be recommended from the preset interest points based on the similarity. Based on this solution, the interest points in the question can be effectively extracted, which provides a basis for question distribution based on the interest points in the question, helps to improve the accuracy of question distribution, and improves the answer effect of the question.

[0047] If the question information to be recommended is input into the matching model together with each preset interest point to obtain the matching degree, and the target interest point is determined based on the matching degree, then it is necessary to perform a model prediction process for the question information to be recommended and each preset interest point. Due to the large amount of data of the preset interest points, the calculation amount of this solution is too large and the calculation time is too long. In the embodiment of the present disclosure, the first feature vector of the question information to be recommended is extracted by using the representation layer of the matching model, and then the first feature vector is respectively calculated with the second feature vector of each preset interest point to determine the target interest point, which can greatly reduce the calculation amount and reduce the calculation time.

[0048] In an optional manner of the present disclosure, determining a target interest point for question information to be recommended from preset interest points based on similarity includes:

[0049] The preset interest point corresponding to the second feature vector whose similarity with the first feature vector is greater than a preset value is determined as the target interest point.

[0050] In the disclosed embodiment, a preset value of similarity can be set. It is considered that when the similarity between the first feature vector and the second feature vector is greater than the preset value, the similarity between the two is high, and the preset interest point corresponding to the second feature vector can be used as the target interest point; when the similarity between the first feature vector and the second feature vector is not greater than the preset value, the similarity between the two is low, and the preset interest point corresponding to the second feature vector cannot be used as the target interest point.

[0051] In the embodiment of the present disclosure, if there is a second feature vector whose similarity with the first feature vector is greater than a preset value, the preset interest point corresponding to the second feature vector is determined as the target interest point. If there is no second feature vector whose similarity with the first feature vector is greater than the preset value, it can be considered that the preset interest points do not contain the target interest point of the question information to be recommended.

[0052] As an example, a second feature vector having a similarity with the first feature vector greater than a preset value may be queried from a preset ANN vector library. For example, K second feature vectors with the highest similarity may be output.

[0053] In an optional manner of the present disclosure, after determining the target interest point for the question information to be recommended from the preset interest points based on the similarity, the method further includes:

[0054] Based on the association between the preset interest points and the answering users, the target answering users corresponding to the target interest points are determined, and the questions corresponding to the question information to be recommended are provided to the target answering users.

[0055] In the disclosed embodiment, an association relationship between the answering user and the preset interest points can be pre-established. Specifically, the interest points of the answering user can be marked by manual labeling, or the answering user's historical question and answer data can be obtained and data mining can be performed on the answering user's historical question and answer data to determine the answering user's preset interest points.

[0056] In an optional manner of the present disclosure, the matching model is trained based on the following method:

[0057] The sample question information and the positive sample interest points are used as positive samples, and the sample question information and the negative sample interest points are used as negative samples to construct a training set, wherein the positive sample interest points and the negative sample interest points are both preset interest points, the positive sample interest points are obtained by annotating the sample question information, and the negative sample interest points and the positive sample interest points are sub-level interest points under the same parent level interest points;

[0058] Train the matching model based on the training set.

[0059] In the disclosed embodiment, the preset interest points may be divided into levels to obtain a hierarchical structure of the preset interest points, where the parent-level interest points are the preset interest points of the previous level of the child-level interest points.

[0060] Compared with coarse-grained labels, the model's ability to distinguish fine-grained labels is worse. Therefore, in this example, positive and negative samples are constructed for fine-grained sub-level interest points to improve the model's ability to distinguish fine-grained labels.

[0061] Specifically, the sample question information is the text information of the question used as the training sample, and the sample question information can be manually labeled to mark the positive sample interest points from the preset interest points. Then, other sub-level interest points under the same parent level interest point as the positive sample interest point can be used as negative sample interest points, thereby completing the construction of the training set.

[0062] As an example, music training is a parent-level interest point, and piano training, violin training, and cello training are all sub-level interest points of music training. If the positive sample interest point of the sample question information is piano training, then the negative sample interest point can be randomly selected from other sub-level interest points of music training except piano training, such as violin training.

[0063] When training the matching model, the Albert model of the first representation layer and the Albert model of the second representation layer are trained together, the loss function is calculated through the matching layer, and then the parameters of the Albert model of the first representation layer and the Albert model of the second representation layer are updated.

[0064] In an optional manner of the present disclosure, after determining the similarity between the first feature vector and at least one second feature vector, the method further includes:

[0065] If there is no second feature vector whose similarity with the first feature vector is greater than a preset value, the question information to be recommended is provided to the annotation user to annotate the newly added interest point.

[0066] In the disclosed embodiment, if there is no second feature vector whose similarity with the first feature vector is greater than a preset value, it can be considered that the preset interest points do not contain the target interest points of the question information to be recommended. At this time, the question information to be recommended can be provided to the annotation user for manual annotation, and the manually marked interest points are the newly added interest points.

[0067] In actual use, the annotation user can also re-annotate the recommendation question information from the preset points of interest. Specifically, the K points of interest with the highest similarity can be provided to the annotation user. If the annotation user believes that there are points of interest that meet the annotation requirements among the K points of interest with the highest similarity, then the points of interest can be used as target points of interest. If the annotation user believes that there are no points of interest that meet the annotation requirements among the K points of interest with the highest similarity, then new points of interest can be annotated.

[0068] In an optional manner of the present disclosure, after the marking user completes marking of the newly added interest point, the method further includes:

[0069] Update preset POIs based on newly added POIs;

[0070] Update the training set based on the updated preset points of interest;

[0071] Fine-tune the matching model based on the updated training set.

[0072] In the disclosed embodiment, the preset points of interest are updated based on the newly added points of interest, that is, the newly added points of interest are added to the preset points of interest as known points of interest. After the newly added points of interest are added to the preset points of interest, the training set can be updated, and then the matching model is fine-tuned based on the updated training set.

[0073] In an optional manner of the present disclosure, after fine-tuning the matching model based on the updated training set, the method further includes:

[0074] The second feature vector is updated based on the fine-tuned matching model.

[0075] In the disclosed embodiment, after fine-tuning the matching model, the second feature vector can be updated based on the fine-tuned matching model, that is, the second feature vector is extracted for the preset interest points including the newly added interest points based on the updated Albert model of the second representation layer.

[0076] Since the content of questions asked by users on the question-answering platform is diverse, it is difficult to limit a complete set of POIs as preset POIs. In this example, the POI set can be expanded by adding new POIs, which can continuously improve the compatibility of the recommendation system with questions.

[0077] As an example, Figure 2 A schematic diagram of a process for determining a target point of interest in an optional implementation of the present disclosure is shown in FIG.

[0078] like Figure 2 As shown in , the training phase is the training phase of the matching model. Question input is Albert, that is, the sample question information is input into the Albert model of the first representation layer. Interest input is Albert, that is, the preset interest points are input into the Albert model of the second representation layer. Vector similarity, that is, the sample question information is matched with the preset interest points in the matching layer, and the matching degree is output.

[0079] The prediction stage is the stage of recommending points of interest for the question information to be recommended. The first feature vector of the question information to be recommended is extracted by the Albert model of the second representation layer in the matching model, and the second feature vector similar to the first feature vector is searched in the ANN vector library, the K second feature vectors with the highest similarity are determined, and the corresponding preset points of interest are output.

[0080] As an example, Figure 3 A schematic diagram of the process of marking new points of interest in an optional implementation manner of the present disclosure is shown in FIG.

[0081] like Figure 3As shown in , the training phase is the training phase of the matching model. Question input is Albert, that is, the sample question information is input into the Albert model of the first representation layer. Interest input is Albert, that is, the preset interest points are input into the Albert model of the second representation layer. Vector similarity, that is, the sample question information is matched with the preset interest points in the matching layer, and the matching degree is output.

[0082] The prediction stage is the stage of recommending points of interest for the question information to be recommended, and the subsequent addition of points of interest. The first feature vector of the question information to be recommended is extracted by matching the Albert model of the second representation layer in the model, and the second feature vector with similarity to the first feature vector is queried in the ANN vector library, the K second feature vectors with the highest similarity are determined, and the corresponding preset points of interest are output. If none of the K second feature vectors with the highest similarity has a similarity with the first feature vector greater than the preset value, it can be considered that the information to be annotated contains new points of interest, and the new points of interest need to be annotated manually.

[0083] After the newly added points of interest are marked, the newly added points of interest can be added to the preset points of interest as known points of interest. After the newly added points of interest are added to the preset points of interest, the training set can be updated, and then the matching model can be fine-tuned based on the updated training set.

[0084] Figure 4 FIG. 4 is a flow chart showing another method for recommending points of interest provided by an embodiment of the present disclosure. Figure 4 As shown in , the method can mainly include:

[0085] Step S410: extracting a first feature vector of the question information to be recommended through a first representation layer in a preset matching model;

[0086] Step S420: determining a similarity between the first feature vector and at least one second feature vector, where the second feature vector is a feature vector of a preset interest point extracted by a second representation layer in the matching model;

[0087] Step S430: determining a preset interest point corresponding to a second feature vector having a similarity with the first feature vector greater than a preset value as a target interest point;

[0088] Step S440: Based on the association relationship between the preset interest points and the answering users, determine the target answering users corresponding to the target interest points, and provide the questions corresponding to the question information to be recommended to the target answering users.

[0089] The information of the question to be recommended may be text information of the question to be recommended, and the question to be recommended may be submitted to the question-answering platform by the user asking the question.

[0090] The preset interest points may be existing interest points, and the interest points may be labels of the interest of the questioning user reflected in the question, and the preset interest points may also be interest labels of the answering user. When the interest points extracted for the question to be recommended are covered by the interest points of the answering user, it can be considered that the answering user is good at answering the question to be recommended.

[0091] The matching model may be a network model for predicting the matching degree between the question and the point of interest. The first representation layer in the matching model may include a feature vector extraction model for the question, for extracting the feature vector of the question. The second representation layer in the matching model may include a feature vector extraction model for the point of interest, for extracting the feature vector of the point of interest.

[0092] As an example, the matching model may be a dual-tower model, which may include a first representation layer, a second representation layer, and a matching layer. The matching layer may be used to match the feature vector of the question with the feature vector of the point of interest to obtain the matching degree between the question and the point of interest.

[0093] The first feature vector may be a semantic feature vector of the question information to be recommended. The question information to be recommended may be mapped into numbers and then input into a feature vector extraction model for the question in the first representation layer to obtain the first feature vector.

[0094] The second feature vector may be a semantic feature vector of a preset interest point. The preset interest point may be first mapped into a number and then input into a feature vector extraction model for the question in the second representation layer to obtain the second feature vector.

[0095] As an example, the feature vector extraction model for the problem in the first representation layer and the feature vector extraction model for the problem in the second representation layer can both be a miniaturized bidirectional encoder representation from transformers (A LiteBidirectional Encoder Representations from Transformers, Albert) model.

[0096] There are usually multiple preset interest points, and the second feature vector of each preset interest point can be extracted in advance based on the second representation layer, and a vector library can be constructed based on multiple second feature vectors. As an example, the vector library can be an Approximate Nearest Neighbor (ANN) vector library.

[0097] In the disclosed embodiment, after determining the first feature vector of the question information to be recommended, the similarity between the first feature vector and each second feature vector can be determined, thereby determining the target interest point based on the similarity between the first feature vector and each second feature vector, and the target interest point is the interest point of the question information to be recommended.

[0098] In the disclosed embodiment, a preset value of similarity can be set. It is considered that when the similarity between the first feature vector and the second feature vector is greater than the preset value, the similarity between the two is high, and the preset interest point corresponding to the second feature vector can be used as the target interest point; when the similarity between the first feature vector and the second feature vector is not greater than the preset value, the similarity between the two is low, and the preset interest point corresponding to the second feature vector cannot be used as the target interest point.

[0099] In the embodiment of the present disclosure, if there is a second feature vector whose similarity with the first feature vector is greater than a preset value, the preset interest point corresponding to the second feature vector is determined as the target interest point. If there is no second feature vector whose similarity with the first feature vector is greater than the preset value, it can be considered that the preset interest points do not contain the target interest point of the question information to be recommended.

[0100] In the disclosed embodiment, an association relationship between the answering user and the preset interest points can be pre-established. Specifically, the interest points of the answering user can be marked by manual labeling, or the answering user's historical question and answer data can be obtained and data mining can be performed on the answering user's historical question and answer data to determine the answering user's preset interest points.

[0101] The method provided by the embodiment of the present disclosure extracts the first feature vector of the question information to be recommended through the first representation layer in the preset matching model, determines the similarity between the first feature vector and at least one second feature vector, the second feature vector is the feature vector of the preset interest point extracted through the second representation layer in the matching model, and determines the preset interest point corresponding to the second feature vector whose similarity with the first feature vector is greater than a preset value as the target interest point; based on the association between the preset interest point and the answering user, determines the target answering user corresponding to the target interest point, and provides the question corresponding to the question information to be recommended to the target answering user. Based on this solution, the interest points in the question can be effectively extracted, and then the question distribution is carried out based on the interest points in the question, which helps to improve the accuracy of question distribution and improve the answering effect of the question.

[0102] If the question information to be recommended is input into the matching model together with each preset interest point to obtain the matching degree, and the target interest point is determined based on the matching degree, then it is necessary to perform a model prediction process for the question information to be recommended and each preset interest point. Due to the large amount of data of the preset interest points, the calculation amount of this solution is too large and the calculation time is too long. In the embodiment of the present disclosure, the first feature vector of the question information to be recommended is extracted by using the representation layer of the matching model, and then the first feature vector is respectively calculated with the second feature vector of each preset interest point to determine the target interest point, which can greatly reduce the calculation amount and reduce the calculation time.

[0103] Figure 5 A flowchart of another method for recommending points of interest provided by an embodiment of the present disclosure is shown. Figure 5 As shown in , the method can mainly include:

[0104] Step S510: extracting a first feature vector of the question information to be recommended through a first representation layer in a preset matching model;

[0105] Step S520: determining a similarity between the first feature vector and at least one second feature vector, where the second feature vector is a feature vector of a preset interest point extracted by a second representation layer in the matching model;

[0106] Step S530: If there is no second feature vector whose similarity with the first feature vector is greater than a preset value, the question information to be recommended is provided to the annotating user to annotate a new point of interest;

[0107] Step S540: updating the preset points of interest based on the newly added points of interest, updating the training set based on the updated preset points of interest, and fine-tuning the matching model based on the updated training set.

[0108] The information of the question to be recommended may be text information of the question to be recommended, and the question to be recommended may be submitted to the question-answering platform by the user asking the question.

[0109] The preset interest points may be existing interest points, and the interest points may be labels of the interest of the questioning user reflected in the question, and the preset interest points may also be interest labels of the answering user. When the interest points extracted for the question to be recommended are covered by the interest points of the answering user, it can be considered that the answering user is good at answering the question to be recommended.

[0110] The matching model may be a network model for predicting the matching degree between the question and the point of interest. The first representation layer in the matching model may include a feature vector extraction model for the question, for extracting the feature vector of the question. The second representation layer in the matching model may include a feature vector extraction model for the point of interest, for extracting the feature vector of the point of interest.

[0111] As an example, the matching model may be a dual-tower model, which may include a first representation layer, a second representation layer, and a matching layer. The matching layer may be used to match the feature vector of the question with the feature vector of the point of interest to obtain the matching degree between the question and the point of interest.

[0112] The first feature vector may be a semantic feature vector of the question information to be recommended. The question information to be recommended may be mapped into numbers and then input into a feature vector extraction model for the question in the first representation layer to obtain the first feature vector.

[0113] The second feature vector may be a semantic feature vector of a preset interest point. The preset interest point may be first mapped into a number and then input into a feature vector extraction model for the question in the second representation layer to obtain the second feature vector.

[0114] As an example, the feature vector extraction model for the problem in the first representation layer and the feature vector extraction model for the problem in the second representation layer can both be a miniaturized bidirectional encoder representation from transformers (A LiteBidirectional Encoder Representations from Transformers, Albert) model.

[0115] There are usually multiple preset interest points, and the second feature vector of each preset interest point can be extracted in advance based on the second representation layer, and a vector library can be constructed based on multiple second feature vectors. As an example, the vector library can be an Approximate Nearest Neighbor (ANN) vector library.

[0116] In the disclosed embodiment, after determining the first feature vector of the question information to be recommended, the similarity between the first feature vector and each second feature vector can be determined, thereby determining the target interest point based on the similarity between the first feature vector and each second feature vector, and the target interest point is the interest point of the question information to be recommended.

[0117] In the disclosed embodiment, if there is no second feature vector whose similarity with the first feature vector is greater than a preset value, it can be considered that the preset interest points do not contain the target interest points of the question information to be recommended. At this time, the question information to be recommended can be provided to the annotation user for manual annotation, and the manually marked interest points are the newly added interest points.

[0118] In actual use, the annotation user can also re-annotate the recommendation question information from the preset points of interest. Specifically, the K points of interest with the highest similarity can be provided to the annotation user. If the annotation user believes that there are points of interest that meet the annotation requirements among the K points of interest with the highest similarity, then the points of interest can be used as target points of interest. If the annotation user believes that there are no points of interest that meet the annotation requirements among the K points of interest with the highest similarity, then new points of interest can be annotated.

[0119] In the disclosed embodiment, the preset points of interest are updated based on the newly added points of interest, that is, the newly added points of interest are added to the preset points of interest as known points of interest. After the newly added points of interest are added to the preset points of interest, the training set can be updated, and then the matching model is fine-tuned based on the updated training set.

[0120] In the disclosed embodiment, after fine-tuning the matching model, the second feature vector can be updated based on the fine-tuned matching model, that is, the second feature vector is extracted for the preset interest points including the newly added interest points based on the updated Albert model of the second representation layer.

[0121] Since the content of questions asked by users on the question-answering platform is diverse, it is difficult to limit a complete set of POIs as preset POIs. In this example, the POI set can be expanded by adding new POIs, which can continuously improve the compatibility of the recommendation system with questions.

[0122] The method provided by the embodiment of the present disclosure extracts the first feature vector of the question information to be recommended through the first representation layer in the preset matching model, determines the similarity between the first feature vector and at least one second feature vector, the second feature vector is the feature vector of the preset interest point extracted through the second representation layer in the matching model, and if there is no second feature vector whose similarity with the first feature vector is greater than a preset value, then the question information to be recommended is provided to the annotation user to annotate the newly added interest point; the preset interest point is updated based on the newly added interest point, the training set is updated based on the updated preset interest point, and the matching model is fine-tuned based on the updated training set. Based on this solution, the interest points in the question can be effectively extracted, and when the extracted interest points do not belong to the known interest points, the known interest points are expanded based on the newly added interest points, which can continuously improve the compatibility of the recommendation system with the question.

[0123] Based on Figure 1 The same principle as shown in the method, Figure 6 FIG. 4 shows a schematic diagram of a structure of a device for recommending points of interest provided by an embodiment of the present disclosure, such as Figure 6 As shown, the POI recommendation device 60 may include:

[0124] A vector extraction module 610, configured to extract a first feature vector of the question information to be recommended through a first representation layer in a preset matching model;

[0125] A vector similarity module 620, configured to determine a similarity between the first feature vector and at least one second feature vector, where the second feature vector is a feature vector of a preset interest point extracted by a second representation layer in the matching model;

[0126] The target interest point determination module 630 is used to determine the target interest point of the question information to be recommended from the preset interest points based on the similarity.

[0127] The device provided by the embodiment of the present disclosure extracts the first feature vector of the question information to be recommended through the first representation layer in the preset matching model, determines the similarity between the first feature vector and at least one second feature vector, the second feature vector is the feature vector of the preset interest point extracted through the second representation layer in the matching model, and determines the target interest point of the question information to be recommended from the preset interest points based on the similarity. Based on this solution, the interest points in the question can be effectively extracted, which provides a basis for question distribution based on the interest points in the question, helps to improve the accuracy of question distribution, and improves the answer effect of the question.

[0128] Optionally, the target interest point determination module is specifically used for:

[0129] The preset interest point corresponding to the second feature vector whose similarity with the first feature vector is greater than a preset value is determined as the target interest point.

[0130] Optionally, the above device further includes:

[0131] The answering user determination module is used to determine the target interest point of the question information to be recommended from the preset interest points based on similarity, determine the target answering user corresponding to the target interest point based on the association between the preset interest point and the answering user, and provide the question corresponding to the question information to be recommended to the target answering user.

[0132] Optionally, the matching model is trained based on:

[0133] The sample question information and the positive sample interest points are used as positive samples, and the sample question information and the negative sample interest points are used as negative samples to construct a training set, wherein the positive sample interest points and the negative sample interest points are both preset interest points, the positive sample interest points are obtained by annotating the sample question information, and the negative sample interest points and the positive sample interest points are sub-level interest points under the same parent level interest points;

[0134] Train the matching model based on the training set.

[0135] Optionally, the above device further includes:

[0136] A new interest point marking module is used to provide the recommended question information to the marking user for marking new interest points after determining the similarity between the first feature vector and at least one second feature vector, if there is no second feature vector whose similarity with the first feature vector is greater than a preset value.

[0137] Optionally, the device further comprises a model fine-tuning module, and the model fine-tuning module is used to:

[0138] After the annotating user completes the marking of the newly added POI, the preset POI is updated based on the newly added POI;

[0139] Update the training set based on the updated preset points of interest;

[0140] Fine-tune the matching model based on the updated training set.

[0141] Optionally, the device further includes a vector updating module, and the vector updating module is used to:

[0142] After fine-tuning the matching model based on the updated training set, the second feature vector is updated based on the fine-tuned matching model.

[0143] It can be understood that the above modules of the device for recommending points of interest in the embodiment of the present disclosure have the function of realizing Figure 1 The functions of the corresponding steps of the method for recommending points of interest in the embodiment shown in . The function can be implemented by hardware, or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions. The above modules can be software and / or hardware, and the above modules can be implemented separately or integrated with multiple modules. For the functional description of each module of the above-mentioned device for recommending points of interest, please refer to Figure 1 The corresponding description of the method for recommending points of interest in the embodiment shown in is not repeated here.

[0144] Based on Figure 4 The same principle as shown in the method, Figure 7 FIG. 4 is a schematic diagram showing a structure of another device for recommending points of interest provided by an embodiment of the present disclosure. Figure 7 As shown, the POI recommendation device 70 may include:

[0145] A vector extraction module 710, configured to extract a first feature vector of the question information to be recommended through a first representation layer in a preset matching model;

[0146] A vector similarity module 720, configured to determine a similarity between the first feature vector and at least one second feature vector, where the second feature vector is a feature vector of a preset interest point extracted by a second representation layer in the matching model;

[0147] A target interest point determination module 730, configured to determine a preset interest point corresponding to a second feature vector having a similarity with the first feature vector greater than a preset value as a target interest point;

[0148] The answering user determination module 740 is used to determine the target answering user corresponding to the target interest point based on the association relationship between the preset interest point and the answering user, and provide the question corresponding to the question information to be recommended to the target answering user.

[0149] The device provided by the embodiment of the present disclosure extracts the first feature vector of the question information to be recommended through the first representation layer in the preset matching model, determines the similarity between the first feature vector and at least one second feature vector, the second feature vector is the feature vector of the preset interest point extracted through the second representation layer in the matching model, and determines the preset interest point corresponding to the second feature vector whose similarity with the first feature vector is greater than a preset value as the target interest point; based on the association between the preset interest point and the answering user, determines the target answering user corresponding to the target interest point, and provides the question corresponding to the question information to be recommended to the target answering user. Based on this solution, the interest points in the question can be effectively extracted, and then the question distribution is carried out based on the interest points in the question, which helps to improve the accuracy of question distribution and improve the answering effect of the question.

[0150] It can be understood that the above modules of the device for recommending points of interest in the embodiment of the present disclosure have the function of realizing Figure 4 The functions of the corresponding steps of the method for recommending points of interest in the embodiment shown in . The function can be implemented by hardware, or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions. The above modules can be software and / or hardware, and the above modules can be implemented separately or integrated with multiple modules. For the functional description of each module of the above-mentioned device for recommending points of interest, please refer to Figure 4 The corresponding description of the method for recommending points of interest in the embodiment shown in is not repeated here.

[0151] Based on Figure 5 The same principle as shown in the method, Figure 8 FIG. 4 shows a schematic diagram of the structure of another device for recommending points of interest provided by an embodiment of the present disclosure, such as Figure 8 As shown, the POI recommendation device 80 may include:

[0152] A vector extraction module 810, configured to extract a first feature vector of the question information to be recommended through a first representation layer in a preset matching model;

[0153] A vector similarity module 820, configured to determine a similarity between the first feature vector and at least one second feature vector, where the second feature vector is a feature vector of a preset interest point extracted by a second representation layer in the matching model;

[0154] A new interest point marking module 830 is used to provide the to-be-recommended question information to the marking user for marking the new interest point when there is no second feature vector having a similarity with the first feature vector greater than a preset value;

[0155] The model fine-tuning module 840 is used to update the preset interest points based on the newly added interest points, update the training set based on the updated preset interest points, and fine-tune the matching model based on the updated training set.

[0156] The device provided by the disclosed embodiment extracts the first feature vector of the question information to be recommended through the first representation layer in the preset matching model, determines the similarity between the first feature vector and at least one second feature vector, the second feature vector is the feature vector of the preset interest point extracted through the second representation layer in the matching model, and if there is no second feature vector whose similarity with the first feature vector is greater than a preset value, then the question information to be recommended is provided to the annotation user for annotation of the newly added interest point; the preset interest points are updated based on the newly added interest points, the training set is updated based on the updated preset interest points, and the matching model is fine-tuned based on the updated training set. Based on this solution, the interest points in the question can be effectively extracted, and when the extracted interest points do not belong to the known interest points, the known interest points are expanded based on the newly added interest points, which can continuously improve the compatibility of the recommendation system with the question.

[0157] It can be understood that the above modules of the device for recommending points of interest in the embodiment of the present disclosure have the function of realizing Figure 5 The functions of the corresponding steps of the method for recommending points of interest in the embodiment shown in . The function can be implemented by hardware, or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions. The above modules can be software and / or hardware, and the above modules can be implemented separately or integrated with multiple modules. For the functional description of each module of the above-mentioned device for recommending points of interest, please refer to Figure 5 The corresponding description of the method for recommending points of interest in the embodiment shown in is not repeated here.

[0158] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0159] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0160] The electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for recommending points of interest as provided in the embodiments of the present disclosure.

[0161] Compared with the prior art, the electronic device extracts the first feature vector of the question information to be recommended through the first representation layer in the preset matching model, determines the similarity between the first feature vector and at least one second feature vector, the second feature vector is the feature vector of the preset interest point extracted through the second representation layer in the matching model, and determines the target interest point of the question information to be recommended from the preset interest points based on the similarity. Based on this solution, the interest points in the question can be effectively extracted, which provides a basis for question distribution based on the interest points in the question, helps to improve the accuracy of question distribution, and improves the answer effect of the question.

[0162] The readable storage medium is a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute the method for recommending points of interest provided by the embodiment of the present disclosure.

[0163] Compared with the prior art, the readable storage medium extracts the first feature vector of the question information to be recommended through the first representation layer in the preset matching model, determines the similarity between the first feature vector and at least one second feature vector, the second feature vector is the feature vector of the preset interest point extracted through the second representation layer in the matching model, and determines the target interest point of the question information to be recommended from the preset interest points based on the similarity. Based on this solution, the interest points in the question can be effectively extracted, which provides a basis for question distribution based on the interest points in the question, helps to improve the accuracy of question distribution, and improves the answer effect of the question.

[0164] The computer program product includes a computer program, and when the computer program is executed by a processor, the method for recommending points of interest provided in the embodiment of the present disclosure is implemented.

[0165] Compared with the prior art, the computer program product extracts the first feature vector of the question information to be recommended through the first representation layer in the preset matching model, determines the similarity between the first feature vector and at least one second feature vector, the second feature vector is the feature vector of the preset interest point extracted through the second representation layer in the matching model, and determines the target interest point of the question information to be recommended from the preset interest points based on the similarity. Based on this solution, the interest points in the question can be effectively extracted, which provides a basis for question distribution based on the interest points in the question, helps to improve the accuracy of question distribution, and improves the answer effect of the question.

[0166] Fig. 9A schematic block diagram of an example electronic device 90 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0167] like Fig. 9 As shown, the electronic device 90 includes a computing unit 910, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 920 or a computer program loaded from a storage unit 980 into a random access memory (RAM) 930. In the RAM 930, various programs and data required for the operation of the device 90 can also be stored. The computing unit 910, the ROM 920, and the RAM 930 are connected to each other via a bus 940. An input / output (I / O) interface 950 is also connected to the bus 940.

[0168] A number of components in the device 90 are connected to the I / O interface 950, including: an input unit 960, such as a keyboard, a mouse, etc.; an output unit 970, such as various types of displays, speakers, etc.; a storage unit 980, such as a disk, an optical disk, etc.; and a communication unit 990, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 990 allows the device 90 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0169] The computing unit 910 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 910 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 910 performs the recommendation method of the points of interest provided in the embodiments of the present disclosure. For example, in some embodiments, the method for recommending points of interest provided in the embodiments of the present disclosure may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 980. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 90 via the ROM 920 and / or the communication unit 990. When the computer program is loaded into the RAM 930 and executed by the computing unit 910, one or more steps of the method for recommending points of interest provided in the embodiments of the present disclosure may be performed. Alternatively, in other embodiments, the computing unit 910 may be configured to perform the method for recommending points of interest provided in the embodiments of the present disclosure in any other appropriate manner (e.g., by means of firmware).

[0170] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0171] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0172] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0173] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0174] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0175] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0176] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0177] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for recommending points of interest, comprising: Extracting a first feature vector of the question information to be recommended through a first representation layer in a preset matching model; Determining a similarity between the first feature vector and at least one second feature vector, where the second feature vector is a feature vector of a preset interest point extracted by a second representation layer in the matching model; Determining a target interest point of the question information to be recommended from the preset interest points based on the similarity; The matching model is trained based on the following method: The sample question information and the positive sample interest points are taken as positive samples, and the sample question information and the negative sample interest points are taken as negative samples to construct a training set, wherein the positive sample interest points and the negative sample interest points both belong to the preset interest points, the positive sample interest points are obtained by annotating the sample question information, and the negative sample interest points and the positive sample interest points are sub-level interest points under the same parent-level interest points; the matching model is trained based on the training set.

2. The method according to claim 1, wherein: The determining the target interest point of the to-be-recommended question information from the preset interest points based on the similarity includes: A preset interest point corresponding to a second feature vector having a similarity with the first feature vector greater than a preset value is determined as a target interest point.

3. The method according to claim 1, wherein: After determining the target interest point of the to-be-recommended question information from the preset interest points based on the similarity, the method further includes: Based on the association relationship between the preset interest points and the question answering users, the target question answering users corresponding to the target interest points are determined, and the questions corresponding to the question information to be recommended are provided to the target question answering users.

4. The method according to any one of claims 1 to 3, wherein: After determining the similarity between the first feature vector and at least one second feature vector, the method further includes: If there is no second feature vector whose similarity with the first feature vector is greater than a preset value, the question information to be recommended is provided to the annotation user to annotate a new point of interest.

5. The method according to claim 4, wherein: After the marking user completes marking the newly added point of interest, the method further includes: Updating the preset points of interest based on the newly added points of interest; Updating the training set based on the updated preset points of interest; The matching model is fine-tuned based on the updated training set.

6. The method according to claim 5, after fine-tuning the matching model based on the updated training set, the method further comprises: The second feature vector is updated based on the fine-tuned matching model.

7. A device for recommending points of interest, comprising: A vector extraction module, used for extracting a first feature vector of the question information to be recommended through a first representation layer in a preset matching model; a vector similarity module, configured to determine a similarity between the first feature vector and at least one second feature vector, wherein the second feature vector is a feature vector of a preset interest point extracted by a second representation layer in the matching model; A target interest point determination module, used to determine the target interest point of the question information to be recommended from the preset interest points based on the similarity; The matching model is trained based on the following method: The sample question information and the positive sample interest points are taken as positive samples, and the sample question information and the negative sample interest points are taken as negative samples to construct a training set, wherein the positive sample interest points and the negative sample interest points both belong to the preset interest points, the positive sample interest points are obtained by annotating the sample question information, and the negative sample interest points and the positive sample interest points are sub-level interest points under the same parent-level interest points; the matching model is trained based on the training set.

8. The device according to claim 7, wherein: The target interest point determination module is specifically used for: A preset interest point corresponding to a second feature vector having a similarity with the first feature vector greater than a preset value is determined as a target interest point.

9. The device according to claim 7, wherein: The device also includes: The answering user determination module is used to determine the target interest point of the question information to be recommended from the preset interest points based on the similarity, determine the target answering user corresponding to the target interest point based on the association relationship between the preset interest point and the answering user, and provide the question corresponding to the question information to be recommended to the target answering user.

10. The device according to any one of claims 7 to 9, further comprising: A new interest point marking module is used to provide the recommended question information to the marking user for marking new interest points after determining the similarity between the first feature vector and at least one second feature vector, if there is no second feature vector whose similarity with the first feature vector is greater than a preset value.

11. The device according to claim 10, further comprising a model fine-tuning module The model fine-tuning module is used to: After the marking user completes marking of the newly added interest point, updating the preset interest point based on the newly added interest point; Updating the training set based on the updated preset points of interest; The matching model is fine-tuned based on the updated training set.

12. The device according to claim 11, further comprising a vector updating module, wherein the vector updating module is configured to: After fine-tuning the matching model based on the updated training set, the second feature vector is updated based on the fine-tuned matching model.

13. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 6.

14. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-6.

15. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 6.

Images