Method, device, equipment and storage medium for determining summary of search results

Abstract

The embodiment of the present application discloses a method, device, equipment and storage medium for determining the summary of search results. The method includes: obtaining a search request and search results, the search request includes a search keyword, the search results include a target document, and the target document includes multiple candidate document sentences; performing attention learning on the sentence features of each candidate document sentence in the multiple candidate document sentences and the search features of the search keyword to determine the correlation between each candidate document sentence and the search keyword; based on the correlation between each candidate document sentence and the search keyword, determining the target candidate document sentence from the multiple candidate document sentences, and using the target candidate document sentence as the summary of the search results, the correlation between the target candidate document sentence and the search keyword is greater than the correlation between other candidate document sentences and the search keyword; outputting the summary of the search results. The use of the embodiment of the present application can improve the efficiency of determining the summary of the search results.

Description

Technical Field

[0001] The present application relates to the field of computer application technology, and in particular to a method, device, equipment and storage medium for determining a summary of a search result. Background Art

[0002] With the popularization of computer technology, search engines play an important role in querying questions and retrieving information in daily life. When displaying search results, the document title and partial content of the document are generally displayed. Users judge whether the document is useful based on this displayed information and decide whether to click it. In order to provide a better search experience, more and more search engines are making "partial content of the document" query-aware, providing different "partial content of the document" displays for different queries.

[0003] The prior art, when determining the summary of query-aware search results, concatenates the query and each document sentence in the document to obtain each concatenated sentence, and then fuses the representation information of each concatenated sentence with the representation information of the query to determine whether each document sentence should be displayed as a summary, resulting in a long time spent on determining the summary of the search results, that is, the efficiency of determining the summary of the search results is low. Summary of the invention

[0004] The embodiments of the present application provide a method, apparatus, device and storage medium for determining a summary of a search result, which can improve the efficiency of determining the summary of the search result.

[0005] In one aspect, an embodiment of the present application provides a method for determining a summary of a search result, the method comprising:

[0006] Obtaining a search request and search results found in response to the search request; wherein the search request includes a search keyword, the search results include a target document, and the target document includes a plurality of candidate document sentences;

[0007] Performing attention learning on the sentence features of each candidate document sentence in the plurality of candidate document sentences and the search features of the search keyword to determine the degree of association between each candidate document sentence and the search keyword;

[0008] Based on the associations between the candidate document sentences and the search keyword, a target candidate document sentence is determined from the multiple candidate document sentences, and the target candidate document sentence is used as a summary of the search results; wherein the association between the target candidate document sentence and the search keyword is greater than the association between other candidate document sentences in the multiple candidate document sentences and the search keyword;

[0009] Output a summary of the search results.

[0010] In one embodiment, performing attention learning on the sentence features of each candidate document sentence in the plurality of candidate document sentences and the search features of the search keyword to determine the degree of association between each candidate document sentence and the search keyword includes:

[0011] Performing linear transformation on sentence features of each candidate document sentence in the plurality of candidate document sentences to obtain a query matrix, a key-value matrix, and a value matrix of each candidate document sentence;

[0012] Performing a linear transformation on the search features of the search keyword to obtain a key-value matrix and a value matrix of the search keyword;

[0013] The attention mechanism is adopted to determine the association between each candidate document sentence and the search keyword based on the query matrix, key-value matrix and value matrix of each candidate document sentence, and the key-value matrix and value matrix of the search keyword.

[0014] In one embodiment, the attention mechanism is used to determine the association between each candidate document sentence and the search keyword based on the query matrix, key-value matrix and value matrix of each candidate document sentence, and the key-value matrix and value matrix of the search keyword, including:

[0015] For any candidate document sentence among the multiple candidate document sentences, concatenate the key-value matrix of the candidate document sentence with the key-value matrix of the search keyword to obtain a fused key-value matrix of the candidate document sentence;

[0016] Concatenate the value matrix of any candidate document sentence with the value matrix of the search keyword to obtain a fused value matrix of any candidate document sentence;

[0017] Attention learning is performed on the query matrix of any candidate document sentence using the fused key-value matrix and the fused value matrix of any candidate document sentence to obtain the correlation between any candidate document sentence and the search information.

[0018] In one embodiment, the method further comprises:

[0019] Inputting each candidate document sentence into an encoder corresponding to each candidate document sentence, and extracting features of each candidate document sentence through the encoder corresponding to each candidate document sentence to obtain sentence features of each candidate document sentence;

[0020] The search keyword is input into an encoder corresponding to the search keyword, and features of the search keyword are extracted by the encoder corresponding to the search keyword to obtain search features of the search keyword.

[0021] In one embodiment, the number of the plurality of candidate document sentences is K, where K is a positive integer;

[0022] The performing attention learning on the sentence features of each candidate document sentence in the plurality of candidate document sentences and the search features of the search keyword to determine the degree of association between each candidate document sentence and the search keyword includes:

[0023] Call K threads to perform attention learning on sentence features of corresponding candidate document sentences among the multiple candidate document sentences and search features of the search keyword, so as to obtain the correlation between the corresponding candidate document sentences and the search keyword; wherein one thread corresponds to one candidate document sentence.

[0024] In one embodiment, the method further comprises:

[0025] Acquire feature vectors of each document sentence included in the target document and feature vectors of the search keyword;

[0026] Based on the feature vectors of the respective document sentences and the feature vectors of the search keywords, obtaining the matching degree between the respective document sentences and the search keywords;

[0027] Based on the matching degree between each of the document sentences and the search keyword, searching for a preset number of document sentences from the document sentences included in the target document; wherein the matching degree between the found document sentences and the search keyword is greater than the matching degree between other document sentences included in the target document and the search keyword;

[0028] The found document sentences are used as the multiple candidate document sentences.

[0029] In one embodiment, the method further comprises:

[0030] Obtain at least one document published to the content publishing platform;

[0031] Performing feature extraction on each document sentence included in each document of the at least one document to obtain a feature vector of each document sentence included in each document;

[0032] Storing the feature vectors of each document sentence included in each document into a memory;

[0033] The step of obtaining a feature vector of each sentence in the target document includes:

[0034] The feature vector of each document sentence included in the target document is obtained from the memory.

[0035] On the other hand, an embodiment of the present application provides a device for determining a summary of a search result, the device for determining a summary of a search result comprising:

[0036] An acquisition unit, configured to acquire a search request and a search result found in response to the search request; wherein the search request includes a search keyword, the search result includes a target document, and the target document includes a plurality of candidate document sentences;

[0037] a relevance determination unit, configured to perform attention learning on sentence features of each candidate document sentence among the plurality of candidate document sentences and search features of the search keyword to determine a relevance between each candidate document sentence and the search keyword;

[0038] a document sentence determination unit, configured to determine a target candidate document sentence from the plurality of candidate document sentences based on the degrees of association between the respective candidate document sentences and the search keyword, and use the target candidate document sentence as a summary of the search result; wherein the degree of association between the target candidate document sentence and the search keyword is greater than the degrees of association between other candidate document sentences in the plurality of candidate document sentences and the search keyword;

[0039] An output unit is used to output a summary of the search result.

[0040] On the other hand, an embodiment of the present application provides a computer device, including a processor, a storage device, and a communication interface, wherein the processor, the storage device, and the communication interface are connected to each other, wherein the storage device is used to store a computer program that supports the computer device to execute the above method, the computer program includes program instructions, and the processor is configured to call the program instructions to perform the following steps:

[0041] Obtaining a search request and search results found in response to the search request; wherein the search request includes a search keyword, the search results include a target document, and the target document includes a plurality of candidate document sentences;

[0042] Performing attention learning on the sentence features of each candidate document sentence in the plurality of candidate document sentences and the search features of the search keyword to determine the degree of association between each candidate document sentence and the search keyword;

[0043] Based on the associations between the candidate document sentences and the search keyword, a target candidate document sentence is determined from the multiple candidate document sentences, and the target candidate document sentence is used as a summary of the search results; wherein the association between the target candidate document sentence and the search keyword is greater than the association between other candidate document sentences in the multiple candidate document sentences and the search keyword;

[0044] Output a summary of the search results.

[0045] On the other hand, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. The computer program includes program instructions. When the program instructions are executed by a processor, the processor executes the above-mentioned method for determining a summary of a search result.

[0046] On the other hand, an embodiment of the present application provides a computer program product, which includes a computer program, and the computer program is suitable for being loaded by a processor and executing the above-mentioned method for determining a summary of a search result.

[0047] In an embodiment of the present application, attention learning is performed on sentence features of each candidate document sentence among multiple candidate document sentences and search features of the search keyword to determine the correlation between each candidate document sentence and the search keyword. Compared with the existing method of concatenating the search keyword and each document sentence in the target document for processing, the length of the text to be processed can be reduced, and the efficiency of determining the summary of the search results can be effectively improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0049] Figure 1 is a schematic diagram of an interface of a search engine provided in an embodiment of the present application;

[0050] Figure 2 It is a flowchart of a method for determining a summary of a search result provided in an embodiment of the present application;

[0051] Figure 3 It is a schematic diagram of the architecture of a search result summary determination system provided in an embodiment of the present application;

[0052] Figure 4It is a structural schematic diagram of a method for determining the degree of association between candidate document sentences and search keywords provided in an embodiment of the present application;

[0053] Figure 5 It is a schematic diagram of the architecture of another search result summary determination system provided in an embodiment of the present application;

[0054] Figure 6 It is a flowchart of another method for determining a summary of a search result provided in an embodiment of the present application;

[0055] Figure 7 It is a structural schematic diagram of a device for determining a summary of a search result provided in an embodiment of the present application;

[0056] Figure 8 It is a structural diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0058] like Figure 1 Taking the interface diagram of the search engine shown as an example, the user can query questions or retrieve information through the search engine. For example, the user can enter a search request in the search engine, for example, the search request can include the search keyword "Is there a surname Xiao in the Hundred Family Names" to query questions; another example is that the search request can include the search keyword "Forbidden City Pictures" to retrieve information. The search engine can find at least one search result that matches the search keyword and output at least one search result found, and any search result includes a document content. Since each search result generally displays the document title and part of the document content (such as the abstract) when it is displayed, the user determines whether the document is useful based on the displayed information, and thus decides whether to click. If the determination efficiency of the summary of the search result is low, it will result in a long search time. Therefore, while ensuring that the summary of the search result is accurate, the time consumed in determining the summary of the search result should be shortened, that is, the determination efficiency of the summary of the search result should be improved.

[0059] Among them, the user can input the search request in the search engine by text, voice or image. When the user inputs the search request in the search engine by text, the user can enter the text information in the search box and click the "Search" button. The computer device responds to the user's operation and can generate a search request. The search keyword included in the search request is the text information entered by the user in the search box. When the user inputs the search request in the search engine by voice, the user can click the "Voice Input" button on the search interface, input the voice information through the microphone of the computer device, and click the "Search" button. The computer device responds to the user's operation and can generate a search request. The search keyword included in the search request is the text information converted from the voice information entered by the user. When the user inputs the search request in the search engine by image, the user can click the "Photo" button on the search interface, collect the image through the camera of the computer device or search the image in the image library of the computer device, and click the "Search" button. The computer device responds to the user's operation and can generate a search request. The search keyword included in the search request is the text information obtained by character recognition or content extraction of the image entered by the user.

[0060] Based on this, the embodiment of the present application provides a method for determining the summary of search results. After obtaining the search request and the search results found in response to the search request, the sentence features of each candidate document sentence in the target document included in the search results and the search features of the search keyword are subjected to attention learning to determine the correlation between each candidate document sentence and the search keyword. Based on the correlation between each candidate document sentence and the search keyword, the target candidate document sentence is determined from multiple candidate document sentences, and the target candidate document sentence is used as the summary of the search result. Compared with the existing method of splicing the search keyword with each document sentence in the document to obtain each spliced ​​sentence, and then fusing the representation information of each spliced ​​sentence with the representation information of the search keyword to determine whether each document sentence should be displayed as a summary, the present application does not need to splice the search keyword with each document sentence in the document, and the present application adopts the attention learning method to replace the existing splicing method, which can reduce the length of the text to be processed and effectively improve the efficiency of determining the summary of the search results.

[0061] The summary determination method of the search results provided in the embodiment of the present application can be applied in a search engine. The search engine can be installed or integrated in a content publishing platform or a browser. The content publishing platform or the browser can run in a computer device. The computer device can include a terminal device or a server, etc. The computer device includes but is not limited to a smart phone, a camera, a wearable device or a computer, etc.

[0062] See also Figure 2 , Figure 2 is a flowchart of a method for determining a summary of a search result provided in an embodiment of the present application. The method for determining a summary of a search result can be executed by a search engine, a content publishing platform, a browser, or a computer device; Figure 2 The summary determination scheme of the search results shown includes but is not limited to steps S201 to S204, wherein:

[0063] S201, obtaining a search request and search results found in response to the search request, the search request includes a search keyword, the search results include a target document, and the target document includes a plurality of candidate document sentences.

[0064] For example, a user may input a search request through a search engine, so after detecting the search request, the search request may be responded to by finding at least one search result that matches the search keyword in the search request. Any search result may include a target document, and the target document may include multiple document sentences. Optionally, some or all of the multiple document sentences may be used as candidate document sentences.

[0065] S202, performing attention learning on the sentence features of each candidate document sentence in the plurality of candidate document sentences and the search features of the search keyword to determine the correlation between each candidate document sentence and the search keyword.

[0066] In the specific implementation, the Cross Attention method can be used for feature fusion, that is, the sentence features of each candidate document sentence in multiple candidate document sentences and the search features of the search keyword are learned by attention to determine the correlation between each candidate document sentence and the search keyword.

[0067] Among them, Cross Attention only changes the input of Self Attention, and Cross Attention is used in the decoder of Transformer. The decoder of Transformer has three inputs, namely input1, input2 and input3. Input1 refers to the output of the decoder in the past. Input2 refers to the position encoding, which is used to represent the position information. Input3 refers to the output of the encoder. Specifically, the decoder first recursively inputs input1, adds it to input2, and then performs Cross Attention with input3 after masked multi-head attention. Therefore, CrossAttention is usually used as a decoder module and used together with Self Attention as an encoder.

[0068] Transformer is a model that uses the attention mechanism to increase the speed of model training. Transformer can include an encoder and a decoder. The Query input to Cross Attention comes from the output of the encoder (Self Attention), while the Key and Value come from the initial input, that is, the input of the encoder. The essence of Cross Attention is to assign a weight coefficient to each element in the input sequence, which can also be understood as soft addressing. If each element in the sequence is stored in the form of (K, V), then Cross Attention completes the addressing by calculating the similarity between Q and K. The similarity calculated between Q and K reflects the importance of the retrieved V value, that is, the weight, and then the weighted sum is used to obtain the value of Cross Attention, that is, the correlation between each candidate document sentence and the search keyword.

[0069] In one implementation, the sentence features of each candidate document sentence in multiple candidate document sentences can be linearly transformed to obtain the query matrix (Query), key-value matrix (Key) and value matrix (Value) of each candidate document sentence, and then the search features of the search keyword can be linearly transformed to obtain the key-value matrix and value matrix of the search keyword, and then the attention mechanism is adopted to determine the correlation between each candidate document sentence and the search keyword based on the query matrix, key-value matrix and value matrix of each candidate document sentence, as well as the key-value matrix and value matrix of the search keyword.

[0070] Specifically, the first linear transformation matrix WQ can be used to linearly transform the sentence features of any candidate document sentence to obtain the query matrix of the candidate document sentence. The second linear transformation matrix WK can be used to linearly transform the sentence features of any candidate document sentence to obtain the key-value matrix of the candidate document sentence. The third linear transformation matrix WV can be used to linearly transform the sentence features of any candidate document sentence to obtain the value matrix of the candidate document sentence. Similarly, the method of linearly transforming the search features of the search keyword to obtain the key-value matrix of the search keyword is the same as the method of linearly transforming the sentence features of the candidate document sentence to obtain the key-value matrix of the candidate document sentence; the method of linearly transforming the search features of the search keyword to obtain the value matrix of the search keyword is the same as the method of linearly transforming the sentence features of the candidate document sentence to obtain the value matrix of the candidate document sentence, and the embodiments of the present application will not be repeated.

[0071] In one implementation, each candidate document sentence can be input into an encoder corresponding to each candidate document sentence, and features of each candidate document sentence can be extracted by the encoder corresponding to each candidate document sentence to obtain sentence features of each candidate document sentence. The search keyword can be input into an encoder corresponding to the search keyword, and features of the search keyword can be extracted by the encoder corresponding to the search keyword to obtain search features of the search keyword.

[0072] by Figure 3Taking the schematic diagram of the architecture of the summary determination system of the search results shown in FIG. 1 as an example, after obtaining the search keyword (query), the search keyword can be input into the encoder (query encoder) corresponding to the search keyword, and the encoder corresponding to the search keyword can be used to extract features of the search keyword to obtain the search features of the search keyword, and then the search features of the search keyword can be linearly transformed to obtain the key-value matrix and value matrix of the search keyword. In addition, after obtaining multiple candidate document sentences, each candidate document sentence can be input into the encoder (query-aware sentence-encoder) corresponding to each candidate document sentence, and the encoder corresponding to each candidate document sentence can be used to extract features of each candidate document sentence to obtain the sentence features of each candidate document sentence, and then the sentence features of each candidate document sentence can be linearly transformed to obtain the query matrix, key-value matrix and value matrix of each candidate document sentence. Furthermore, the query matrix, key-value matrix and value matrix of each candidate document sentence, as well as the key-value matrix and value matrix of the search keyword can be input into the document-level encoder, and then the document-level encoder uses an attention mechanism to determine the correlation between each candidate document sentence and the search keyword based on the query matrix, key-value matrix and value matrix of each candidate document sentence, as well as the key-value matrix and value matrix of the search keyword.

[0073] Among them, the encoder corresponding to the search keyword can be composed of multiple layers of transformer encoders, and the encoder corresponding to any candidate document sentence can be composed of multiple layers of transformer encoders. For example, the encoder corresponding to the search keyword can be composed of 12 layers of transformer encoders, and the encoder corresponding to any candidate document sentence can be composed of 12 layers of transformer encoders, and the calculation of each layer of transformer is composed of three parts: Q, K, and V. The embodiment of the present application extracts the values ​​of K and V from the encoder corresponding to the search keyword, and performs cross attention calculation with Q obtained by the encoder corresponding to the candidate document sentence, which reduces the length of the input text, thereby greatly improving the calculation efficiency.

[0074] In one implementation, for any candidate document sentence among multiple candidate document sentences, the key-value matrix of any candidate document sentence can be concatenated with the key-value matrix of the search keyword to obtain a fused key-value matrix of any candidate document sentence, and the value matrix of any candidate document sentence can be concatenated with the value matrix of the search keyword to obtain a fused value matrix of any candidate document sentence. The fused key-value matrix and the fused value matrix of any candidate document sentence are used to perform attention learning on the query matrix of any candidate document sentence to obtain the correlation between any candidate document sentence and the search information.

[0075] by Figure 4 Taking the structural diagram of the method for determining the degree of association between the candidate document sentences and the search keywords as an example, the key-value matrix of a candidate document sentence can be spliced ​​with the key-value matrix of the search keyword to obtain the fused key-value matrix of the candidate document sentence, and the value matrix of the candidate document sentence can be spliced ​​with the value matrix of the search keyword to obtain the fused value matrix of the candidate document sentence. The fused key-value matrix and the fused value matrix of the candidate document sentence are used to perform attention learning on the query matrix of the candidate document sentence to obtain the degree of association between the candidate document sentence and the search information. The embodiment of the present application splices the values ​​of K and V in the query encoder with the values ​​of K and V in the query perception sentence-encoder, thereby realizing the interaction between the search keyword and the candidate document sentence.

[0076] Optionally, an embodiment of the present application may use a multi-head attention module to utilize the fused key-value matrix and the fused value matrix of the candidate document sentence to perform attention learning on the query matrix of the candidate document sentence to obtain the correlation between the candidate document sentence and the search information.

[0077] Among them, the multi-head attention module is formed by combining multiple Self-Attentions. Multi-HeadAttention contains multiple Self-Attention layers. First, the input X is passed to h different Self-Attentions respectively, and h output matrices Z are calculated. Taking h=8 as an example, 8 output matrices Z will be obtained at this time. Multi-HeadAttention concatenates these 8 output matrices Z together (Concat), and then passes them to a Linear layer to obtain the final output Z of Multi-Head Attention. There are multiple Self-Attentions in Multi-Head Attention in Transformer, which can capture the correlation coefficient attention score between words in multiple dimensions.

[0078] Optionally, in the embodiment of the present application, the key-value matrix of a candidate document sentence is concatenated with the key-value matrix of the search keyword to obtain the fused key-value matrix of the candidate document sentence, which can be understood as: adding the key-value matrix of a candidate document sentence to the key-value matrix of the search keyword to obtain the fused key-value matrix of the candidate document sentence. The value matrix of the candidate document sentence is concatenated with the value matrix of the search keyword to obtain the fused value matrix of the candidate document sentence, which can be understood as: adding the value matrix of the candidate document sentence to the value matrix of the search keyword to obtain the fused value matrix of the candidate document sentence.

[0079] In one implementation, the number of multiple candidate document sentences can be K, where K is a positive integer. Then, K threads can be called to perform attention learning on the sentence features of corresponding candidate document sentences in the multiple candidate document sentences and the search features of the search keywords, respectively, to obtain the correlation between the corresponding candidate document sentences and the search keywords, where one thread corresponds to one candidate document sentence.

[0080] In an embodiment of the present application, the correlation between each candidate document sentence and the search keyword can be obtained in parallel, that is, the sentence features of the corresponding candidate document sentences in multiple candidate document sentences and the search features of the search keyword can be studied in parallel to obtain the correlation between the corresponding candidate document sentences and the search keyword, which can improve the efficiency of obtaining the correlation, thereby effectively improving the efficiency of determining the summary of the search results.

[0081] S203, based on the correlation between each candidate document sentence and the search keyword, determine a target candidate document sentence from the plurality of candidate document sentences, and use the target candidate document sentence as a summary of the search result.

[0082] The correlation between the target candidate document sentence and the search keyword is greater than the correlation between other candidate document sentences in the plurality of candidate document sentences and the search keyword.

[0083] In one implementation, after obtaining the relevance between each candidate document sentence and the search keyword, a candidate document sentence with a relevance greater than a preset relevance threshold can be determined from multiple candidate document sentences, and the determined candidate document sentence is the target candidate document sentence, that is, the target candidate document sentence refers to the candidate document sentence with a relevance greater than the preset relevance threshold, and the target candidate document sentence can be used as a summary of the search results. The preset relevance threshold is a preset percentage value, such as 85% or 90%.

[0084] In one implementation, after obtaining the relevance between each candidate document sentence and the search keyword, multiple candidate document sentences can be sorted based on the relevance, and a target number of candidate document sentences are determined from the sorted multiple candidate document sentences. The determined candidate document sentences are target candidate document sentences, that is, the target candidate document sentences refer to the target number of candidate document sentences with greater relevance, and the target candidate document sentences can be used as the summary of the search results. The target number is a preset value, such as 2 or 3.

[0085] S204: Output a summary of the search results.

[0086] After determining the summary of the search results, the summary of the search results can be displayed on the search interface. Figure 1 For example, after the user enters a search request in the search box, the user can respond to the search request and find at least one search result that matches the search keyword XX, including search result 1, search result 2, and search result 3, and then determine the summary of each search result through the summary determination method of the search result provided in the embodiment of the present application, and then display the summary of each search result on the search interface, that is, the summary of search result 1, the summary of search result 2, and the summary of search result 3. Optionally, one or more of the following can also be displayed on the search interface: the document title of the target document included in each search result, the cover image of the target document included in each search result, the publishing author of the target document included in each search result, the publishing date, interactive data, etc. The interactive data may include one or more of the following: the number of likes, the number of shares, the number of comments, the number of favorites, etc.

[0087] In an embodiment of the present application, after obtaining a search request and the search results found in response to the search request, attention learning is performed on the sentence features of each candidate document sentence in the target document included in the search results and the search features of the search keyword to determine the correlation between each candidate document sentence and the search keyword. Based on the correlation between each candidate document sentence and the search keyword, a target candidate document sentence is determined from multiple candidate document sentences, and the target candidate document sentence is used as a summary of the search results. The summary of the search results is then output, which can reduce the length of the text to be processed and effectively improve the efficiency of determining the summary of the search results.

[0088] An embodiment of the present application also provides another method for determining a summary of search results, specifically, obtaining a search request and search results found in response to the search request, the search request includes a search keyword, the search results include a target document, and the target document includes multiple document sentences; obtaining a feature vector of each document sentence included in the target document, and a feature vector of the search keyword; based on the feature vector of each document sentence and the feature vector of the search keyword, obtaining a matching degree between each document sentence and the search keyword; based on the matching degree between each document sentence and the search keyword, searching for a preset number of document sentences from the document sentences included in the target document; determining a summary of the search results based on the preset number of document sentences and the search keyword found; and outputting a summary of the search results.

[0089] Optionally, there are multiple ways to determine the summary of the search results based on a preset number of document sentences and search keywords. For example, see Figure 2 The summary determination method of the search results shown determines the summary of the search results. Specifically, the sentence features of each document sentence in a preset number of document sentences and the search features of the search keyword are subjected to attention learning to determine the correlation between each document sentence and the search keyword. Based on the correlation between each document sentence and the search keyword, a target document sentence is determined from the preset number of document sentences, and the target document sentence is used as the summary of the search results. For another example, a traditional summary determination method of search results can be used to determine the summary of the search results. Exemplarily, the search keyword and each document sentence in a preset number of document sentences are spliced ​​to obtain a spliced ​​document sentence; the search keyword is feature extracted by an encoder corresponding to the search keyword to obtain the search feature of the search keyword; the spliced ​​document sentence is feature extracted by an encoder corresponding to each document sentence to obtain the sentence features of each document sentence; the sentence features of any document sentence and the search features of the search keyword are fused to determine whether the document sentence is used as the summary of the search results.

[0090] by Figure 5Taking the schematic diagram of the architecture of the summary determination system of the search results shown as an example, after obtaining the search keyword (query), the search keyword can be input into the encoder (query encoder) corresponding to the search keyword, and the encoder corresponding to the search keyword can be used to extract features of the search keyword to obtain a feature vector of the search keyword. In addition, each document sentence included in the target document can be input into the encoder (query-aware sentence-encoder) corresponding to each document sentence, and the encoder corresponding to each document sentence can be used to extract features of each document sentence to obtain a feature vector of each document sentence. Further, the document-level encoder can be called to obtain the matching degree between each document sentence and the search keyword based on the feature vector of each document sentence and the feature vector of the search keyword, and then based on the matching degree between each document sentence and the search keyword, a preset number of document sentences are searched from the document sentences included in the target document; based on the preset number of document sentences and the search keyword found, the summary of the search result is determined.

[0091] Compared with the traditional method of determining the summary of the search results based on the various document sentences and search keywords included in the target document, the embodiment of the present application first searches for a preset number of document sentences related to the search keywords from the multiple document sentences included in the target document, and then determines the summary of the search results based on the preset number of document sentences and search keywords found, which can effectively improve the efficiency of determining the summary of the search results.

[0092] Based on the above description, see Figure 6 , Figure 6 is a flow chart of another method for determining a summary of a search result provided in an embodiment of the present application. The method for determining a summary of a search result can be executed by a search engine, a content publishing platform, a browser or a computer device; Figure 6 The summary determination scheme of the search results shown includes but is not limited to steps S601 to S608, wherein:

[0093] S601, obtaining a search request and search results found in response to the search request, wherein the search request includes a search keyword, the search results include a target document, and the target document includes a plurality of document sentences.

[0094] For example, a user may input a search request through a search engine, so after detecting the search request, the search request may be responded to by finding at least one search result matching the search keyword in the search request. Any search result may include a target document, and the target document may include multiple document sentences.

[0095] S602, acquiring feature vectors of each document sentence included in the target document and feature vectors of search keywords.

[0096] by Figure 5 For example, after obtaining the search keyword (query), the search keyword can be input into the encoder corresponding to the search keyword (query encoder), and the encoder corresponding to the search keyword can be used to extract features of the search keyword to obtain a feature vector of the search keyword. In addition, each document sentence included in the target document can be input into the encoder corresponding to each document sentence (query-aware sentence-encoder), and the encoder corresponding to each document sentence can be used to extract features of each document sentence to obtain a feature vector of each document sentence.

[0097] In one implementation, at least one document published to a content publishing platform may be obtained, and feature extraction may be performed on each document sentence included in each document in the at least one document to obtain a feature vector of each document sentence included in each document, and the feature vector of each document sentence included in each document may be stored in a memory. Based on this, a method of obtaining the feature vector of each document sentence included in a target document may include: obtaining the feature vector of each document sentence included in a target document from a memory.

[0098] In the embodiment of the present application, since there is no need to concatenate the document sentences and the search keywords, the feature vectors of each document sentence included in the target document can be calculated and cached in advance, thereby saving online computing time.

[0099] S603, based on the feature vector of each document sentence and the feature vector of the search keyword, obtain the matching degree between each document sentence and the search keyword.

[0100] S604: searching for a preset number of document sentences from the document sentences included in the target document based on the matching degree between each document sentence and the search keyword.

[0101] The matching degree between the found document sentence and the search keyword is greater than the matching degree between other document sentences included in the target document and the search keyword.

[0102] S605: The found document sentences are used as a plurality of candidate document sentences.

[0103] S606: Perform attention learning on the sentence features of each candidate document sentence in the plurality of candidate document sentences and the search features of the search keyword to determine the correlation between each candidate document sentence and the search keyword.

[0104] S607: Determine a target candidate document sentence from the plurality of candidate document sentences based on the relevance between each candidate document sentence and the search keyword, and use the target candidate document sentence as a summary of the search result.

[0105] The correlation between the target candidate document sentence and the search keyword is greater than the correlation between other candidate document sentences in the plurality of candidate document sentences and the search keyword.

[0106] S608: Output a summary of the search results.

[0107] It can be understood that steps S606 to S608 can refer to the description of steps S202 to S204 in the above embodiment, and will not be repeated in the embodiment of the present application.

[0108] In an embodiment of the present application, a preset number of candidate document sentences related to the search keyword are extracted from multiple document sentences included in the target document through a coarse-grained extraction stage; then, through a fine-grained extraction stage, sentence features of each candidate document sentence in the multiple candidate document sentences and search features of the search keyword are subjected to attention learning to determine a summary of the search results, which can effectively improve the efficiency of determining the summary of the search results while ensuring the accuracy of the summary of the search results.

[0109] An embodiment of the present application further provides a computer storage medium, in which program instructions are stored. When the program instructions are executed, they are used to implement the corresponding methods described in the above embodiments.

[0110] See also Figure 7 , Figure 7 It is a structural diagram of a device for determining a summary of a search result according to an embodiment of the present application.

[0111] In one implementation of the device for determining a summary of a search result in an embodiment of the present application, the device for determining a summary of a search result includes the following structure.

[0112] An acquisition unit 701 is used to acquire a search request and a search result found in response to the search request; wherein the search request includes a search keyword, the search result includes a target document, and the target document includes a plurality of candidate document sentences;

[0113] A relevance determination unit 702, configured to perform attention learning on the sentence features of each candidate document sentence in the plurality of candidate document sentences and the search features of the search keyword to determine the relevance between each candidate document sentence and the search keyword;

[0114] A document sentence determination unit 703 is used to determine a target candidate document sentence from the multiple candidate document sentences based on the associations between the respective candidate document sentences and the search keyword, and use the target candidate document sentence as a summary of the search result; wherein the association between the target candidate document sentence and the search keyword is greater than the association between other candidate document sentences in the multiple candidate document sentences and the search keyword;

[0115] The output unit 704 is configured to output a summary of the search results.

[0116] In one embodiment, the relevance determination unit 702 performs attention learning on the sentence features of each candidate document sentence in the plurality of candidate document sentences and the search features of the search keyword to determine the relevance between each candidate document sentence and the search keyword, including:

[0117] Performing linear transformation on sentence features of each candidate document sentence in the plurality of candidate document sentences to obtain a query matrix, a key-value matrix, and a value matrix of each candidate document sentence;

[0118] Performing a linear transformation on the search features of the search keyword to obtain a key-value matrix and a value matrix of the search keyword;

[0119] The attention mechanism is adopted to determine the association between each candidate document sentence and the search keyword based on the query matrix, key-value matrix and value matrix of each candidate document sentence, and the key-value matrix and value matrix of the search keyword.

[0120] In one embodiment, the relevance determination unit 702 adopts an attention mechanism to determine the relevance between each candidate document sentence and the search keyword based on the query matrix, key-value matrix and value matrix of each candidate document sentence, and the key-value matrix and value matrix of the search keyword, including:

[0121] For any candidate document sentence among the multiple candidate document sentences, concatenate the key-value matrix of the candidate document sentence with the key-value matrix of the search keyword to obtain a fused key-value matrix of the candidate document sentence;

[0122] Concatenate the value matrix of any candidate document sentence with the value matrix of the search keyword to obtain a fused value matrix of any candidate document sentence;

[0123] Attention learning is performed on the query matrix of any candidate document sentence using the fused key-value matrix and the fused value matrix of any candidate document sentence to obtain the correlation between any candidate document sentence and the search information.

[0124] In one embodiment, the search result summary determination device further includes:

[0125] The feature extraction unit 705 is used to input each candidate document sentence into the encoder corresponding to each candidate document sentence, perform feature extraction on each candidate document sentence through the encoder corresponding to each candidate document sentence, and obtain sentence features of each candidate document sentence; input the search keyword into the encoder corresponding to the search keyword, perform feature extraction on the search keyword through the encoder corresponding to the search keyword, and obtain the search feature of the search keyword.

[0126] In one embodiment, the number of the plurality of candidate document sentences is K, where K is a positive integer;

[0127] The relevance determination unit 702 performs attention learning on the sentence features of each candidate document sentence in the plurality of candidate document sentences and the search features of the search keyword to determine the relevance between each candidate document sentence and the search keyword, including:

[0128] Call K threads to perform attention learning on sentence features of corresponding candidate document sentences among the multiple candidate document sentences and search features of the search keyword, so as to obtain the correlation between the corresponding candidate document sentences and the search keyword; wherein one thread corresponds to one candidate document sentence.

[0129] In one embodiment, the acquisition unit 701 is further configured to acquire a feature vector of each document sentence included in the target document and a feature vector of the search keyword; based on the feature vector of each document sentence and the feature vector of the search keyword, acquire a matching degree between each document sentence and the search keyword;

[0130] The search result summary determination device further includes:

[0131] The document sentence searching unit 706 is used to search for a preset number of document sentences from the document sentences included in the target document based on the matching degree between the respective document sentences and the search keyword; wherein the matching degree between the found document sentences and the search keyword is greater than the matching degree between other document sentences included in the target document and the search keyword; and the found document sentences are used as the multiple candidate document sentences.

[0132] In one embodiment, the acquisition unit 701 is further configured to acquire at least one document published to the content publishing platform;

[0133] The search result summary determination device further includes:

[0134] A feature extraction unit 705 is used to extract features from each document sentence included in each document in the at least one document to obtain a feature vector of each document sentence included in each document;

[0135] A storage unit 707, used to store the feature vectors of each document sentence included in each document into a memory;

[0136] The acquisition unit 701 acquires the feature vector of each document sentence included in the target document, including:

[0137] The feature vector of each document sentence included in the target document is obtained from the memory.

[0138] In an embodiment of the present application, after the acquisition unit 701 acquires the search request and the search results found in response to the search request, the association determination unit 702 performs attention learning on the sentence features of each candidate document sentence in the target document included in the search results and the search features of the search keyword to determine the association between each candidate document sentence and the search keyword. The document sentence determination unit 703 determines a target candidate document sentence from multiple candidate document sentences based on the association between each candidate document sentence and the search keyword, and uses the target candidate document sentence as a summary of the search results. Then, the output unit 704 outputs the summary of the search results, which can reduce the length of the text to be processed and effectively improve the efficiency of determining the summary of the search results.

[0139] See also Figure 8 , Figure 8 80 is a schematic diagram of the structure of a computer device provided in an embodiment of the present application. The computer device in the embodiment of the present application includes a power supply module and other structures, and includes a processor 801, a storage device 802, and a communication interface 803. The processor 801, the storage device 802, and the communication interface 803 can exchange data, and the processor 801 implements the corresponding search result summary determination method.

[0140] The storage device 802 may include a volatile memory, such as a random-access memory (RAM); the storage device 802 may also include a non-volatile memory, such as a flash memory, a solid-state drive (SSD), etc.; the storage device 802 may also include a combination of the above-mentioned types of memory.

[0141] The processor 801 may be a central processing unit (CPU). The processor 801 may also be a combination of a CPU and a GPU. In the server, multiple CPUs and GPUs may be included as needed to determine the summary of the corresponding search results. In one embodiment, the storage device 802 is used to store program instructions. The processor 801 may call program instructions to implement various methods involved in the embodiments of the present application.

[0142] In a first possible implementation, a processor 801 of a computer device calls a program instruction stored in a storage device 802 to obtain a search request and search results found in response to the search request; wherein the search request includes a search keyword, the search results include a target document, and the target document includes a plurality of candidate document sentences; attention learning is performed on sentence features of each candidate document sentence in the plurality of candidate document sentences and search features of the search keyword to determine the degree of association between each candidate document sentence and the search keyword; based on the degree of association between each candidate document sentence and the search keyword, a target candidate document sentence is determined from the plurality of candidate document sentences, and the target candidate document sentence is used as a summary of the search results; wherein the degree of association between the target candidate document sentence and the search keyword is greater than the degree of association between other candidate document sentences in the plurality of candidate document sentences and the search keyword; and a summary of the search results is output.

[0143] In one embodiment, the processor 801 performs the following operations when performing attention learning on the sentence features of each candidate document sentence in the multiple candidate document sentences and the search features of the search keyword to determine the degree of association between each candidate document sentence and the search keyword:

[0144] Performing linear transformation on sentence features of each candidate document sentence in the plurality of candidate document sentences to obtain a query matrix, a key-value matrix, and a value matrix of each candidate document sentence;

[0145] Performing a linear transformation on the search features of the search keyword to obtain a key-value matrix and a value matrix of the search keyword;

[0146] The attention mechanism is adopted to determine the association between each candidate document sentence and the search keyword based on the query matrix, key-value matrix and value matrix of each candidate document sentence, and the key-value matrix and value matrix of the search keyword.

[0147] In one embodiment, the processor 801 is used to perform the following operations when determining the association between each candidate document sentence and the search keyword based on the query matrix, key-value matrix and value matrix of each candidate document sentence and the key-value matrix and value matrix of the search keyword using the attention mechanism:

[0148] For any candidate document sentence among the multiple candidate document sentences, concatenate the key-value matrix of the candidate document sentence with the key-value matrix of the search keyword to obtain a fused key-value matrix of the candidate document sentence;

[0149] Concatenate the value matrix of any candidate document sentence with the value matrix of the search keyword to obtain a fused value matrix of any candidate document sentence;

[0150] Attention learning is performed on the query matrix of any candidate document sentence using the fused key-value matrix and the fused value matrix of any candidate document sentence to obtain the correlation between any candidate document sentence and the search information.

[0151] In one embodiment, the processor 801 is further configured to perform the following operations:

[0152] Inputting each candidate document sentence into an encoder corresponding to each candidate document sentence, and extracting features of each candidate document sentence through the encoder corresponding to each candidate document sentence to obtain sentence features of each candidate document sentence;

[0153] The search keyword is input into an encoder corresponding to the search keyword, and features of the search keyword are extracted by the encoder corresponding to the search keyword to obtain search features of the search keyword.

[0154] In one embodiment, the number of the plurality of candidate document sentences is K, where K is a positive integer;

[0155] When the processor 801 performs attention learning on the sentence features of each candidate document sentence in the plurality of candidate document sentences and the search features of the search keyword to determine the degree of association between each candidate document sentence and the search keyword, the processor 801 is configured to perform the following operations:

[0156] Call K threads to perform attention learning on sentence features of corresponding candidate document sentences among the multiple candidate document sentences and search features of the search keyword, so as to obtain the correlation between the corresponding candidate document sentences and the search keyword; wherein one thread corresponds to one candidate document sentence.

[0157] In one embodiment, the processor 801 is further configured to perform the following operations:

[0158] Acquire feature vectors of each document sentence included in the target document and feature vectors of the search keyword;

[0159] Based on the feature vectors of the respective document sentences and the feature vectors of the search keywords, obtaining the matching degree between the respective document sentences and the search keywords;

[0160] Based on the matching degree between each document sentence and the search keyword, a preset number of document sentences are searched from the document sentences included in the target document; wherein the matching degree between the found document sentences and the search keyword is greater than the matching degree between other document sentences included in the target document and the search keyword; and the found document sentences are used as the multiple candidate document sentences.

[0161] In one embodiment, the processor 801 is further configured to perform the following operations:

[0162] Obtain at least one document published to the content publishing platform;

[0163] Performing feature extraction on each document sentence included in each document of the at least one document to obtain a feature vector of each document sentence included in each document;

[0164] Storing the feature vectors of each document sentence included in each document into a memory;

[0165] When acquiring the feature vector of each document sentence included in the target document, the processor 801 is used to perform the following operations:

[0166] The feature vector of each document sentence included in the target document is obtained from the memory.

[0167] In an embodiment of the present application, after obtaining a search request and the search results found in response to the search request, the processor 801 performs attention learning on the sentence features of each candidate document sentence in the target document included in the search results and the search features of the search keyword to determine the correlation between each candidate document sentence and the search keyword, and based on the correlation between each candidate document sentence and the search keyword, determines a target candidate document sentence from multiple candidate document sentences, and uses the target candidate document sentence as a summary of the search results, and then outputs a summary of the search results, which can reduce the length of the text to be processed and effectively improve the efficiency of determining the summary of the search results.

[0168] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the computer-readable storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM). The computer-readable storage medium can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the blockchain node, etc.

[0169] The above disclosure is only part of the embodiments of the present application, which certainly cannot be used to limit the scope of rights of the present application. Ordinary technicians in this field can understand that all or part of the processes of implementing the above embodiments and making equivalent changes according to the claims of this application are still within the scope of the present invention.

Claims

1. A method for determining a summary of a search result, characterized in that: include: Obtaining a search request and search results found in response to the search request; wherein the search request includes a search keyword, the search results include a target document, and the target document includes a plurality of candidate document sentences; Attention learning is performed on the sentence features of each candidate document sentence in the multiple candidate document sentences and the search features of the search keyword to determine the association between each candidate document sentence and the search keyword, which includes: inputting the search keyword into the encoder corresponding to the search keyword to obtain the key-value matrix and value matrix of the search keyword; inputting each candidate document sentence into the encoder corresponding to the corresponding candidate document sentence to obtain the query matrix, key-value matrix and value matrix of each candidate document sentence, and performing attention calculation based on the query matrix of each candidate document sentence, the fused key-value matrix obtained by concatenating the key-value matrix of each candidate document sentence and the key-value matrix of the search keyword, and the fused value matrix obtained by concatenating the value matrix of each candidate document sentence and the value matrix of the search keyword; Based on the associations between the candidate document sentences and the search keyword, a target candidate document sentence is determined from the multiple candidate document sentences, and the target candidate document sentence is used as a summary of the target document; wherein the association between the target candidate document sentence and the search keyword is greater than the association between other candidate document sentences in the multiple candidate document sentences and the search keyword; Outputs a summary of the target document.

2. The method according to claim 1, characterized in that The performing attention learning on the sentence features of each candidate document sentence in the plurality of candidate document sentences and the search features of the search keyword to determine the degree of association between each candidate document sentence and the search keyword includes: Performing linear transformation on sentence features of each candidate document sentence in the plurality of candidate document sentences to obtain a query matrix, a key-value matrix, and a value matrix of each candidate document sentence; Performing a linear transformation on the search features of the search keyword to obtain a key-value matrix and a value matrix of the search keyword; The attention mechanism is adopted to determine the association between each candidate document sentence and the search keyword based on the query matrix, key-value matrix and value matrix of each candidate document sentence, and the key-value matrix and value matrix of the search keyword.

3. The method according to claim 2, characterized in that The attention mechanism is used to determine the association between each candidate document sentence and the search keyword based on the query matrix, key-value matrix and value matrix of each candidate document sentence, and the key-value matrix and value matrix of the search keyword, including: For any candidate document sentence among the multiple candidate document sentences, concatenate the key-value matrix of the candidate document sentence with the key-value matrix of the search keyword to obtain a fused key-value matrix of the candidate document sentence; Concatenate the value matrix of any candidate document sentence with the value matrix of the search keyword to obtain a fused value matrix of any candidate document sentence; Attention learning is performed on the query matrix of any candidate document sentence using the fused key-value matrix and the fused value matrix of any candidate document sentence to obtain the correlation between any candidate document sentence and the search keyword.

4. The method according to claim 2 or 3, characterized in that: The method further comprises: Inputting each candidate document sentence into an encoder corresponding to each candidate document sentence, and extracting features of each candidate document sentence through the encoder corresponding to each candidate document sentence to obtain sentence features of each candidate document sentence; The search keyword is input into an encoder corresponding to the search keyword, and features of the search keyword are extracted by the encoder corresponding to the search keyword to obtain search features of the search keyword.

5. The method according to claim 2, characterized in that: The number of the plurality of candidate document sentences is K, where K is a positive integer; The performing attention learning on the sentence features of each candidate document sentence in the plurality of candidate document sentences and the search features of the search keyword to determine the degree of association between each candidate document sentence and the search keyword includes: Call K threads to perform attention learning on sentence features of corresponding candidate document sentences among the multiple candidate document sentences and search features of the search keyword, so as to obtain the correlation between the corresponding candidate document sentences and the search keyword; wherein one thread corresponds to one candidate document sentence.

6. The method according to claim 1, characterized in that The method further comprises: Acquire feature vectors of each document sentence included in the target document and feature vectors of the search keyword; Based on the feature vectors of the respective document sentences and the feature vectors of the search keywords, obtaining the matching degree between the respective document sentences and the search keywords; Based on the matching degree between each document sentence and the search keyword, searching for a preset number of document sentences from the document sentences included in the target document; wherein the matching degree between the found document sentences and the search keyword is greater than the matching degree between other document sentences included in the target document and the search keyword; The found document sentences are used as the multiple candidate document sentences.

7. The method according to claim 6, characterized in that The method further comprises: Obtain at least one document published to a content publishing platform; Performing feature extraction on each document sentence included in each document of the at least one document to obtain a feature vector of each document sentence included in each document; Storing the feature vectors of each document sentence included in each document into a memory; The step of obtaining a feature vector of each sentence in the target document includes: The feature vector of each document sentence included in the target document is obtained from the memory.

8. A device for determining a summary of a search result, characterized in that: The device comprises: An acquisition unit, configured to acquire a search request and a search result found in response to the search request; wherein the search request includes a search keyword, the search result includes a target document, and the target document includes a plurality of candidate document sentences; A relevance determination unit, used for performing attention learning on the sentence features of each candidate document sentence in the multiple candidate document sentences and the search features of the search keyword to determine the relevance between each candidate document sentence and the search keyword, including: inputting the search keyword into an encoder corresponding to the search keyword to obtain a key-value matrix and a value matrix of the search keyword; inputting each candidate document sentence into an encoder corresponding to the corresponding candidate document sentence to obtain a query matrix, a key-value matrix and a value matrix of each candidate document sentence, and performing attention calculation based on the query matrix of each candidate document sentence, a fused key-value matrix obtained by concatenating the key-value matrix of each candidate document sentence and the key-value matrix of the search keyword, and a fused value matrix obtained by concatenating the value matrix of each candidate document sentence and the value matrix of the search keyword; a document sentence determination unit, configured to determine a target candidate document sentence from the plurality of candidate document sentences based on the associations between the respective candidate document sentences and the search keyword, and use the target candidate document sentence as a summary of the target document; wherein the association between the target candidate document sentence and the search keyword is greater than the association between other candidate document sentences in the plurality of candidate document sentences and the search keyword; An output unit is used to output the summary of the target document.

9. A computer device, characterized in that: The computer device comprises a processor, a storage device and a communication interface, wherein the processor, the storage device and the communication interface are connected to each other, wherein: The storage device is used to store a computer program, wherein the computer program includes program instructions; The processor is used to call the program instructions to execute the method for determining a summary of a search result according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the method for determining a summary of a search result according to any one of claims 1 to 7 is executed.

11. A computer program product, characterized in that The computer program product comprises a computer program, and the computer program is suitable for being loaded by a processor and executing the method for determining a summary of a search result according to any one of claims 1 to 7.

Images