A method, device, computer equipment and storage medium for generating literature review

Abstract

The present invention relates to the technical field of literature review, and discloses a method, device, computer equipment and storage medium for generating literature review, the method comprising: obtaining a search subject and a search level; based on the search subject and the search level, retrieving a literature relationship tree from a literature database; obtaining at least one target influencing factor, a weight and a first quantity of each target influencing factor; based on at least one target influencing factor and the weight of each target influencing factor, extracting a first quantity of target documents from the literature relationship tree; and generating a literature review based on the first quantity of target documents. The present invention obtains a literature relationship tree by performing a primary screening based on the search subject and the search level, and further screens from the literature relationship tree by determining the target influencing factor and the corresponding weight, thereby generating a literature review, reducing the time for manual screening and reading of documents, obtaining a literature review that better meets the research needs of scientific researchers, and improving the professionalism of the literature review.

Description

Technical Field

[0001] The present invention relates to the technical field of literature review, and in particular to a method, device, computer equipment and storage medium for generating a literature review. Background Art

[0002] Literature review is a systematic review and summary of existing literature in a research field or topic, aiming to identify research trends, knowledge gaps, controversial points and future research directions. It can save researchers time in obtaining research information, help them quickly get a general idea, and establish a research framework. However, due to the large number of literature and the rich fields, researchers need to spend a lot of time and energy on writing literature reviews. Therefore, how to generate literature reviews is a problem that needs to be solved.

[0003] At present, the existing technology usually obtains the topic input by the researchers, screens out the literature related to the input topic from the existing literature in the literature database, and then generates a literature review based on the screened literature. However, screening only based on the input topic makes the screened literature very different from the literature required by the researchers, so the generated literature review cannot meet the research needs of the researchers and has low professionalism. Summary of the invention

[0004] In view of this, the present invention provides a method, apparatus, computer device and storage medium for generating a literature review to solve the problem that literature screening is performed only according to an input topic, resulting in the generated literature review being unable to meet the research needs of researchers and having low professionalism.

[0005] In a first aspect, the present invention provides a method for generating a literature review, the method comprising:

[0006] Get the search topic and search level;

[0007] Based on the search topic and search level, a document relationship tree is retrieved from the document database;

[0008] Obtaining at least one target influencing factor, a weight of each target influencing factor, and a first quantity, wherein a change in the target influencing factor causes a change in the subject of the document, and the weight of any target influencing factor indicates the degree of influence of the target influencing factor on the subject of the document;

[0009] Extracting a first number of target documents from the document relationship tree based on at least one target impact factor and a weight of each target impact factor;

[0010] Generate a literature review based on the first number of target documents.

[0011] The method for generating a literature review provided by the present invention obtains a literature relationship tree by performing an initial screening from a literature database based on a search topic and a search level, and can intuitively display the logical connection and hierarchical structure between the documents. Furthermore, by determining target influencing factors and the weights corresponding to the target influencing factors, further screening is performed from the literature relationship tree to extract target documents that meet the requirements of scientific researchers, and then a literature review is generated, which significantly reduces the time for manual screening and reading of documents, and can adjust the target influencing factors and the corresponding weights according to the different research needs of scientific researchers, so as to obtain a literature review that better meets the research needs of scientific researchers, thereby improving the professionalism of the literature review.

[0012] In an optional implementation, based on the search subject and the search level, a document relationship tree is retrieved from a document database, including:

[0013] When performing a search at the first search level, all documents in the document database are preprocessed to obtain all preprocessed documents;

[0014] For any preprocessed document, subject analysis and influencing factor analysis are performed to obtain at least one document subject, at least one subject influencing factor and the weight of each subject influencing factor corresponding to the preprocessed document. The change of the subject influencing factor causes the document subject to change. The weight of any subject influencing factor indicates the degree of influence of the subject influencing factor on the document subject.

[0015] Traversing all preprocessed documents, if there is a case where at least one document subject corresponding to any preprocessed document includes the search subject, the preprocessed document, at least one document subject corresponding to the preprocessed document, at least one subject influencing factor and the weight of each subject influencing factor are added to the document relationship tree;

[0016] Repeat the above traversal process until all preprocessed documents are traversed once, completing the retrieval of the first retrieval level, and obtaining the document relationship tree after the retrieval of the first retrieval level;

[0017] For all preprocessed documents retrieved in the first retrieval level, take any document topic other than the retrieval topic corresponding to each preprocessed document as a new retrieval topic, return to the above traversal process, and perform retrieval at the next retrieval level until the retrieval of all retrieval levels is completed to obtain a document relationship tree.

[0018] The literature review generation method provided by the present invention can more accurately capture the literature related to the search topic through a multi-layer progressive retrieval strategy, thereby improving the accuracy of the retrieval. In addition, in each layer of retrieval, not only the literature directly related to the search topic is focused on, but also an extended search is performed through the topics of the retrieved documents, thereby increasing the depth and breadth of the retrieval. The retrieved literature relationship tree not only presents the relationship between the documents, but also intuitively displays the development context of the topics at different levels, thereby accelerating the literature screening process and improving the efficiency of generating reviews.

[0019] In an optional implementation, all subject influencing factors in the document relationship tree include at least one target influencing factor;

[0020] Based on at least one target influencing factor and a weight of each target influencing factor, extracting a first number of target documents from the document relationship tree includes:

[0021] Based on at least one subject influencing factor corresponding to each preprocessed document in the document relationship tree, extracting at least one first document including all target influencing factors from the document relationship tree;

[0022] Determine whether the quantity of at least one first document is less than the first quantity;

[0023] When the number of at least one first document is not less than a first number, a first number of target documents are extracted from the at least one first document based on the weight of each subject influencing factor and the weight of each target influencing factor corresponding to the at least one first document.

[0024] The literature review generation method provided by the present invention first extracts the first documents containing all target influencing factors from the document relationship tree, and then compares the number of the first documents with the first number. When the first documents are sufficient, a secondary screening is performed from highly relevant first documents based on the target influencing factors and the weights of the target influencing factors, thereby ensuring that the screened documents are closely aligned with the research needs, thereby improving the professionalism of the literature review.

[0025] In an optional implementation, when the number of at least one first document is not less than a first number, extracting a first number of target documents from the at least one first document based on the weight of each subject influencing factor and the weight of each target influencing factor corresponding to the at least one first document includes:

[0026] For any first document and any target impact factor, determining a subject impact factor consistent with the target impact factor from at least one subject impact factor corresponding to the first document;

[0027] Determine the weight difference between the topic influence factor and the target influence factor;

[0028] The sum of the weight differences between each target impact factor and the subject impact factor consistent with each target impact factor is taken as the weight corresponding to the first document;

[0029] According to the weight corresponding to at least one first document, the documents are sorted from small to large, and the first number of first documents with the highest sorting positions are determined as target documents.

[0030] The literature review generation method provided by the present invention identifies, for each target influencing factor, a factor that matches the target influencing factor from multiple subject influencing factors contained in any first document, and calculates the weight difference between the two to reflect the overall degree of fit between the first document and the research needs. By quantifying and sorting the degree of fit, it is convenient to perform statistical analysis and comparison, so that the screened documents are more in line with the research needs of scientific researchers, thereby improving the professionalism of the literature review.

[0031] In an optional implementation, after determining whether the number of at least one first document is less than the first number, the method further includes:

[0032] When the number of at least one first document is less than the first number, obtaining at least one candidate search topic, the candidate search topic being a potential document topic of any first document and / or a similar topic of any target influencing factor;

[0033] taking the difference between the quantity of at least one first document and the first quantity as the second quantity;

[0034] For any candidate search topic, after removing at least one first document from the document relationship tree, extracting a second number of second documents including the candidate search topic;

[0035] At least one first document and a second number of second documents are used as a first number of target documents.

[0036] The literature review generation method provided by the present invention obtains alternative search topics when the number of first documents initially screened out is less than the expected first number, and screens again based on the alternative search topics, so that the screening obtains target documents that meet the first number, thereby increasing the flexibility of screening documents.

[0037] In an optional implementation, generating a literature review based on the first number of target documents includes:

[0038] For any target document, a document weight is assigned to the target document based on the weight corresponding to the subject influencing factor of the target document, the weight corresponding to at least one target influencing factor and at least one alternative search topic;

[0039] Based on the document weights corresponding to the first number of target documents, a text summarization algorithm is used to generate a plurality of document summaries corresponding to the first number of target documents;

[0040] A natural language generation algorithm is used to generate a literature review based on multiple literature abstracts.

[0041] The literature review generation method provided by the present invention extracts the literature abstract based on the literature weight of each target literature, and generates the literature review based on the extracted multiple literature abstracts using a natural language generation algorithm, so that the structure of the literature review is clear and the logic is coherent, thereby ensuring the accuracy and professionalism of the literature review.

[0042] In a second aspect, the present invention provides a device for generating a literature review, the device comprising:

[0043] A first acquisition module is used to acquire a search topic and a search level;

[0044] A retrieval module is used to retrieve a document relationship tree from a document database based on a retrieval subject and a retrieval level;

[0045] A second acquisition module is used to acquire at least one target influencing factor, a weight of each target influencing factor and a first quantity, wherein a change in the target influencing factor causes a change in the subject of the document, and the weight of any target influencing factor indicates the degree of influence of the target influencing factor on the subject of the document;

[0046] A first extraction module, configured to extract a first number of target documents from the document relationship tree based on at least one target influencing factor and a weight of each target influencing factor;

[0047] A generation module is used to generate a literature review based on a first number of target documents.

[0048] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the literature review generating method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0049] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the literature review generating method of the first aspect or any corresponding embodiment thereof.

[0050] In a fifth aspect, the present invention provides a computer program product, comprising computer instructions for causing a computer to execute the literature review generating method of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0052] Figure 1 is a flow chart of a method for generating a literature review according to an embodiment of the present invention;

[0053] Figure 2 is a flowchart of another method for generating a literature review according to an embodiment of the present invention;

[0054] Figure 3 is a schematic diagram of a document relationship tree according to an embodiment of the present invention;

[0055] Figure 4 is a flowchart of another literature review method according to an embodiment of the present invention;

[0056] Figure 5 is a structural block diagram of a device for generating a literature review according to an embodiment of the present invention;

[0057] Figure 6 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0059] Since the number of documents in the literature database is very large and the fields are rich, scientific researchers need to consume a lot of time and energy in writing literature reviews. When generating a literature review, the prior art usually obtains the subject input by the scientific researcher, screens out the documents related to the input subject from the literature database, and then generates a literature review. However, only screening is performed according to the input subject, so that the generated literature review cannot meet the research needs of the scientific researcher and has low professionalism. The literature review generation method provided by the embodiment of the present invention, by performing a primary screening from the literature database based on the retrieval subject and the retrieval level to obtain a document relationship tree, and by determining the target influencing factors and the corresponding weights of the target influencing factors, further screening is performed from the document relationship tree, and then a literature review is generated, which can obtain a literature review that is more in line with the research needs of the scientific researcher, and improves the professionalism of the literature review.

[0060] According to an embodiment of the present invention, an embodiment of a method for generating a literature review is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0061] In this embodiment, a method for generating a literature review is provided, which can be used in electronic devices. Figure 1 is a flowchart of a method for generating a literature review according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0062] Step S101, obtaining the search subject and search level. Specifically, since there are a large number of documents in a wide range of fields in the document database, it is very difficult for researchers to search for documents and write literature reviews. Therefore, researchers can input the search subject and search level of the documents to be searched through the input device of the electronic device, such as a keyboard, so that the processor of the electronic device searches according to the input requirements of the researchers, thereby improving the search efficiency and further improving the efficiency of generating the literature review.

[0063] Step S102, based on the search subject and the search level, retrieve a document relationship tree from the document database. Specifically, the search subject can be one or more, and the search level can be one or more layers. By searching based on the search subject, the retrieved documents are consistent with the search subject, and by setting the search level, one or more searches can be performed according to the needs of scientific researchers. The obtained document relationship tree can intuitively display the logical connection and hierarchical structure between the documents, so that scientific researchers can understand the current research status through the document relationship tree.

[0064] Step S103, obtaining at least one target influencing factor, the weight of each target influencing factor and the first quantity, the change of the target influencing factor causes the subject of the document to change, and the weight of any target influencing factor represents the degree of influence of the target influencing factor on the subject of the document. Specifically, the target influencing factor will have an impact on the subject of the document, such as experimental conditions, sample characteristics, method selection, external environment, etc. When the target influencing factor changes, the subject of the document will also change. The weight of the target influencing factor is the degree of influence on the subject of the document. Assuming that the influence of target influencing factor A "experimental conditions" on the subject of the document is greater than the influence of target influencing factor B "method selection" on the subject of the document, the weight corresponding to target influencing factor A should be higher than the weight corresponding to target influencing factor B. Since the document relationship tree only provides search results and cannot directly provide documents for generating a literature review, the target influencing factors and their corresponding weights input by the scientific researcher can be obtained, and the document can be screened again from the document relationship tree, so that the screened documents are more in line with the research needs of the scientific researcher, thereby improving the professionalism of the literature review. Optionally, the target influencing factor and its weight can be adjusted according to the research needs of the scientific researcher.

[0065] Step S104, based on at least one target influencing factor and the weight of each target influencing factor, extract a first number of target documents from the document relationship tree. Specifically, a literature review is usually obtained by comprehensive analysis of multiple documents, so after screening based on the target influencing factors and their weights input by the researchers, the first number of target documents are extracted, so that the obtained target documents are consistent with the research needs of the researchers.

[0066] Step S105, generating a literature review based on the first number of target documents. Specifically, since the target documents obtained through the above two screening processes can meet the research needs of researchers, the professionalism of the literature review is improved compared to the literature review generated by the literature retrieved only by the search topic.

[0067] The literature review generation method provided in the embodiment of the present invention can intuitively display the logical connection and hierarchical structure between the documents by performing initial screening from the literature database based on the retrieval subject and the retrieval level, and further screen from the literature relationship tree by determining the target influencing factors and the weights corresponding to the target influencing factors, and extract the target documents that meet the requirements of the researchers, and then generate a literature review, which significantly reduces the time for manual screening and reading of documents, and can adjust the target influencing factors and the corresponding weights according to the different research needs of the researchers, so as to obtain a literature review that better meets the research needs of the researchers, thereby improving the professionalism of the literature review.

[0068] In this embodiment, a method for generating a literature review is provided, which can be used in the above-mentioned electronic device. Figure 2 is a flowchart of another method for generating a literature review according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0069] Step S201, obtain the search subject and search level. Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0070] Step S202: based on the search subject and the search level, retrieve and obtain a document relationship tree from the document database.

[0071] Specifically, the above step S202 includes:

[0072] Step S2021, when performing a search at the first search level, preprocess all documents in the document database to obtain all preprocessed documents. Specifically, preprocess all documents in the document database, such as removing redundant spaces, non-word characters, case conversion, etc., so that the preprocessed documents are convenient for subsequent analysis.

[0073] Step S2022, for any preprocessed document, subject analysis and influencing factor analysis are performed to obtain at least one document subject, at least one subject influencing factor and the weight of each subject influencing factor corresponding to the preprocessed document. The change of the subject influencing factor causes the document subject to change, and the weight of any subject influencing factor indicates the degree of influence of the subject influencing factor on the document subject. Specifically, for any preprocessed document, the LDA (Latent Dirichlet Allocation) model is used for subject analysis to obtain at least one document subject corresponding to the preprocessed document, such as gastric cancer and early gastric cancer. Then, the TextBlob sentiment analysis and keyword extraction methods are used to perform influencing factor analysis to obtain at least one subject influencing factor corresponding to the preprocessed document and the weight corresponding to each subject influencing factor, such as stomach medicine A and stomach medicine A corresponding to 80%. The change of the subject influencing factor can cause the document subject to change, and the weight of the subject influencing factor indicates the degree of influence on the document subject. Optionally, the subject analysis and influencing factor analysis performed on the document are both prior arts and can be implemented in other ways, and the embodiments of the present invention will not be repeated here. By performing subject analysis and influencing factor analysis on the preprocessed document, it is beneficial to perform document retrieval.

[0074] In some optional implementations, after subject analysis of the preprocessed document, a causal relationship analysis can also be performed using a Transformer model based on the document subject and search subject of the preprocessed document to determine whether the document subject of the document is the cause or effect of the search subject. If the document subject of the document is the cause of the search subject, the weight corresponding to the subject influence factor of the document represents the degree of influence on the document subject. If the document subject of the document is the effect of the search subject, the degree of influence of the subject influence factor of the document on the document subject is divided into favorable and unfavorable aspects, and the weight corresponding to the subject influence factor of the document represents the degree of favorable influence or unfavorable influence on the document subject.

[0075] In some optional implementations, the subject influencing factors and their weights obtained after performing an influencing factor analysis on any preprocessed document can be represented by the following codes.

[0076] Struct Impact Analysis Report

[0077] {

[0078] Article_info; / / Basic information of the preprocessed document

[0079] articleImpactList; / / All topic impact factors corresponding to the preprocessed literature

[0080] }

[0081] Stuct ImpactInfo

[0082] {

[0083] factorName; / / The name of the factor affecting the topic

[0084] factorDetails; / / Details of the factors affecting the topic

[0085] double impactFactor; / / The weight of the topic impact factor

[0086] }

[0087] Step S2023, traverse all preprocessed documents, if there is any preprocessed document corresponding to at least one document theme including the search theme, add the preprocessed document, at least one document theme corresponding to the preprocessed document, at least one theme influencing factor and the weight of each theme influencing factor to the document relationship tree. Specifically, assuming that the search theme is "gastric cancer", all preprocessed documents in the document database are traversed, if the document theme of any preprocessed document traversed is "gastric cancer, early gastric cancer", then the preprocessed document, at least one document theme corresponding to the preprocessed document, at least one theme influencing factor and the weight of each theme influencing factor are added to the document relationship tree, and an edge is constructed between the preprocessed document and the search theme. Continue to judge the next preprocessed document, if the document theme of the preprocessed document is "brain cancer" and does not include the search theme "gastric cancer", then do not process the preprocessed document, and continue to traverse the next preprocessed document.

[0088] Step S2024, repeat the above traversal process until all pre-processed documents are traversed once, the retrieval of the first retrieval level is completed, and the document relationship tree after the retrieval of the first retrieval level is obtained. Specifically, when all pre-processed documents in the document database are traversed once, the retrieval of the first retrieval level is completed, and the document relationship tree after the retrieval is obtained.

[0089] Step S2025, for all preprocessed documents retrieved in the first level of search hierarchy, take any document subject other than the search subject corresponding to each preprocessed document as a new search subject, return to the above traversal process, and perform a search in the next level of search hierarchy until all search levels are searched to obtain a document relationship tree. Specifically, in the document relationship tree after a search, there are multiple preprocessed documents retrieved once. Assume that the document subject of any preprocessed document is "gastric cancer, early gastric cancer". Since "gastric cancer" is the search subject, "early gastric cancer" can be used as a new search subject, and repeat the above steps S2023 to S2024 to perform a secondary search. By performing a secondary search on each preprocessed document obtained in the first search, a document relationship tree after the secondary search is obtained. Based on the search subject of each level of search hierarchy and the initially set search level, repeat the above steps S2023 to S2025 until all search levels are searched to obtain a document relationship tree. Optionally, Figure 3 is a schematic diagram of a document relationship tree according to an embodiment of the present invention, assuming that the search level is 2. Figure 3 It can be seen that 4 pre-processed documents were retrieved in the first search level: Document 1, Document 2, Document 3 and Document 4. Then the 4 pre-processed documents were searched again and the following results were obtained: Figure 3The document relationship tree shown in the figure. Among them, document 1 is searched at the second level of the search layer, and document 11 and document 12 are obtained; document 2 is searched at the second level of the search layer, and document 21 is obtained; document 3 is searched at the second level of the search layer, and document 31 is obtained; document 4 is searched at the second level of the search layer, and document 41, document 42 and document 43 are obtained. Through the multi-layer progressive search strategy, the documents related to the search topic can be captured more accurately, which improves the accuracy of the search, and in each level of the search, not only the documents directly related to the search topic are focused on, but also the search is extended through the topics of the retrieved documents, which increases the depth and breadth of the search, so that the document relationship tree obtained by the search not only presents the relationship between the documents, but also intuitively shows the development context of the topics at different levels, while accelerating the document screening process and improving the efficiency of generating reviews.

[0090] Step S203, obtaining at least one target influencing factor, the weight of each target influencing factor and the first quantity, the change of the target influencing factor causes the document theme to change, and the weight of any target influencing factor indicates the degree of influence of the target influencing factor on the document theme. Figure 1 Step S103 of the illustrated embodiment will not be described in detail here.

[0091] Step S204: extract a first number of target documents from the document relationship tree based on at least one target impact factor and the weight of each target impact factor. Figure 1 Step S104 of the illustrated embodiment will not be described in detail here.

[0092] Step S205: Generate a literature review based on the first number of target documents. Figure 1 Step S105 of the illustrated embodiment will not be described in detail here.

[0093] The literature review generation method provided in the embodiment of the present invention can intuitively display the logical connection and hierarchical structure between the documents by performing initial screening from the literature database based on the retrieval subject and the retrieval level, and further screen from the literature relationship tree by determining the target influencing factors and the weights corresponding to the target influencing factors, and extract the target documents that meet the requirements of the researchers, and then generate a literature review, which significantly reduces the time for manual screening and reading of documents, and can adjust the target influencing factors and the corresponding weights according to the different research needs of the researchers, so as to obtain a literature review that better meets the research needs of the researchers, thereby improving the professionalism of the literature review.

[0094] In this embodiment, a method for generating a literature review is provided, which can be used in the above-mentioned electronic device. Figure 4 is a flowchart of another literature review method according to an embodiment of the present invention, such as Figure 4As shown, the process includes the following steps:

[0095] Step S401, obtain the search subject and search level. Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0096] Step S402: Based on the search topic and search level, retrieve the document relationship tree from the document database. Figure 2 Step S202 of the illustrated embodiment will not be described in detail here.

[0097] Step S403, obtaining at least one target influencing factor, the weight of each target influencing factor and the first quantity, the change of the target influencing factor causes the document topic to change, and the weight of any target influencing factor indicates the degree of influence of the target influencing factor on the document topic. Figure 1 Step S103 of the illustrated embodiment will not be described in detail here.

[0098] Step S404: extracting a first number of target documents from the document relationship tree based on at least one target influencing factor and the weight of each target influencing factor, wherein all subject influencing factors in the document relationship tree include at least one target influencing factor.

[0099] Specifically, the above step S404 includes:

[0100] Step S4041, based on at least one subject influencing factor corresponding to each preprocessed document in the document relationship tree, extract at least one first document including all target influencing factors from the document relationship tree. Specifically, assuming that the target influencing factor is "stomach medicine A, stomach medicine B", determine the first document including "stomach medicine A, stomach medicine B" from the document relationship tree.

[0101] Step S4042, determining whether the number of at least one first document is less than a first number. Specifically, the first number is the number of documents input by the researcher for generating a literature review. To ensure that the research needs of the researcher are met, it is necessary to determine whether the number of first documents including "stomach medicine A, stomach medicine B" meets the first number.

[0102] Step S4043, when the number of at least one first document is not less than the first number, based on the weight of each subject influencing factor and the weight of each target influencing factor corresponding to the at least one first document, extract a first number of target documents from the at least one first document.

[0103] In some optional implementations, the above step S4043 includes:

[0104] Step a1, for any first document and any target influencing factor, determine the subject influencing factor that is consistent with the target influencing factor from at least one subject influencing factor corresponding to the first document. Specifically, when the number of first documents is not less than the first number, that is, at this time, the scientific researcher's demand for the number of documents can be met, and documents that are more in line with the scientific researcher's research needs can be screened. Since both the subject influencing factor and the target influencing factor can affect the subject of the document, it is necessary to determine whether the subject influencing factor and the target influencing factor of the first document are consistent. If they are consistent, it means that the first document meets the research needs of the scientific researcher.

[0105] Step a2, determine the weight difference between the topic influencing factor and the target influencing factor. Specifically, the weight of the influencing factor reflects the degree of influence of the influencing factor, that is, quantifying the influencing factor. Therefore, by determining the weight difference between two influencing factors, the consistency between the two influencing factors can be clarified. The smaller the weight difference, the higher the consistency between the two influencing factors.

[0106] Step a3, taking the sum of the weight differences between each target influencing factor and the subject influencing factor consistent with each target influencing factor as the weight corresponding to the first document. Specifically, if the target influencing factor is "stomach medicine A, stomach medicine B", the weight difference between the target influencing factor "stomach medicine A" and the subject influencing factor "stomach medicine A" and the weight difference between the target influencing factor "stomach medicine B" and the subject influencing factor "stomach medicine B" are determined respectively, and the sum of the two weight differences is taken as the weight corresponding to the first document, which can reflect the degree of fit between the first document and the research needs of scientific researchers. The smaller the weight corresponding to the first document, the higher the degree of fit.

[0107] Step a4, sorting from small to large according to the weight corresponding to at least one first document, and determining the first number of first documents with the top ranking position as the target documents. Specifically, assuming that the number of first documents is 10 and the first number is 5, sort the 10 first documents from small to large according to the weight, and determine the first documents with the top 5 ranking positions as the target documents. By first extracting the first documents containing all the target influencing factors from the document relationship tree, and then comparing the number of first documents with the first number, when there are sufficient first documents, a secondary screening is performed from highly relevant first documents based on the target influencing factors and the weights of the target influencing factors to ensure that the screened documents are closely aligned with the research needs, thereby improving the professionalism of the literature review.

[0108] Step S4044, when the number of at least one first document is less than the first number, at least one alternative search topic is obtained, and the alternative search topic is a potential document topic of any first document and / or a similar topic of any target influencing factor. Specifically, assuming that the number of first documents is 3 and the first number is 5, the number of first documents is less than the first number at this time, which cannot meet the scientific researcher's demand for the number of documents, and the document screening needs to be performed again at this time. There are two ways to obtain alternative search topics: the first way is that when performing a topic analysis on any first document in the above step S2022, not only the document topic of the first document can be obtained, but also the potential document topic of the first document can be obtained. Among them, the potential document topic can have an impact on the document topic, but its impact is less than the impact corresponding to the document topic. For example, the document topic of the first document is "immune response: immune response", and the LDA model can also obtain the potential document topics "cytokines: cytokines", "T-cells: T cells" and "inflammation: inflammation". The second way is to use the synonyms or synonyms of the target influencing factors input by the scientific researcher as similar topics. For example, if the target influencing factor "neoplasm: tumor" is input by researchers, similar topics "tumor: tumor" and "cancer: cancer" can be obtained. The topics obtained by combining any one of the above methods or both methods are used as alternative search topics, so that literature screening can be performed again based on the alternative search topics.

[0109] Step S4045, the difference between the quantity of at least one first document and the first quantity is used as the second quantity. Specifically, since the difference between the quantity of the first document and the first quantity is 2, the quantity requirement can be met by selecting two more documents.

[0110] Step S4046, for any candidate search topic, after removing at least one first document from the document relationship tree, extract a second number of second documents including the candidate search topic. Specifically, since there may be more than one candidate search topic, the second document including any candidate search topic can be screened out from the document relationship tree after removing the first document. Optionally, the order of the candidate search topics can be set voluntarily during the screening, and the embodiment of the present invention does not limit this.

[0111] Step S4047, at least one first document and a second number of second documents are used as the first number of target documents. Specifically, when the number of the first documents initially screened out is less than the expected first number, alternative search topics are obtained, and screening is performed again based on the alternative search topics, so that the target documents that meet the first number are screened out, thereby increasing the flexibility of screening documents.

[0112] Step S405: Generate a literature review based on the first number of target documents.

[0113] Specifically, the above step S405 includes:

[0114] Step S4051, for any target document, a document weight is assigned to the target document based on the weight corresponding to the subject influencing factor of the target document, the weight corresponding to at least one target influencing factor, and at least one alternative search subject. Specifically, for any target document, if the target document belongs to the first document, that is, it is determined based on the weight difference between the target influencing factor and the subject influencing factor, then a document weight can be assigned to the target document based on the weight difference. The smaller the weight difference, the greater the document weight, and the more it meets the research needs of scientific researchers. If the target document belongs to the second document, that is, it is determined based on the alternative search subject, then a document weight can be assigned to the target document by itself. The greater the document weight, the more it meets the research needs of scientific researchers.

[0115] Step S4052: Based on the document weights corresponding to the first number of target documents, a text summarization algorithm is used to generate multiple document summaries corresponding to the first number of target documents. Specifically, based on the document weight corresponding to each target document, a text summarization algorithm is used to generate multiple document summaries, ensuring that the target document with a larger document weight generates a more detailed document summaries.

[0116] Step S4053, using a natural language generation algorithm, a literature review is generated based on multiple document abstracts. Specifically, a natural language generation algorithm is used to generate a coherent and logical literature review based on multiple document abstracts. Optionally, the above-mentioned generation of a literature review is a prior art and will not be repeated here. By extracting the document abstract based on the document weight of each target document, and using a natural language generation algorithm to generate a literature review based on the extracted multiple document abstracts, the structure of the literature review is clear and logically coherent, ensuring the accuracy and professionalism of the literature review.

[0117] The literature review generation method provided in the embodiment of the present invention can intuitively display the logical connection and hierarchical structure between the documents by performing initial screening from the literature database based on the retrieval subject and the retrieval level, and further screen from the literature relationship tree by determining the target influencing factors and the weights corresponding to the target influencing factors, and extract the target documents that meet the requirements of the researchers, and then generate a literature review, which significantly reduces the time for manual screening and reading of documents, and can adjust the target influencing factors and the corresponding weights according to the different research needs of the researchers, so as to obtain a literature review that better meets the research needs of the researchers, thereby improving the professionalism of the literature review.

[0118] In the present embodiment, a document review generating device is also provided, and the device is used to implement the above-mentioned embodiment and preferred implementation mode, and the description that has been made will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware is also possible and conceived.

[0119] This embodiment provides a device for generating a literature review. Figure 5 As shown, including:

[0120] The first acquisition module 501 is used to acquire the search topic and the search level.

[0121] The search module 502 is used to retrieve a document relationship tree from a document database based on a search topic and a search level.

[0122] The second acquisition module 503 is used to obtain at least one target influencing factor, the weight of each target influencing factor and the first quantity. The change of the target influencing factor causes the document theme to change. The weight of any target influencing factor represents the degree of influence of the target influencing factor on the document theme.

[0123] The first extraction module 504 is used to extract a first number of target documents from the document relationship tree based on at least one target influencing factor and the weight of each target influencing factor.

[0124] The generating module 505 is used to generate a literature review based on the first number of target documents.

[0125] In some optional implementations, the retrieval module 502 includes:

[0126] The preprocessing unit is used to preprocess all documents in the document database when performing a search at the first search level to obtain all preprocessed documents.

[0127] The analysis unit is used to perform subject analysis and influencing factor analysis on any preprocessed document to obtain at least one document subject, at least one subject influencing factor and the weight of each subject influencing factor corresponding to the preprocessed document. The change of the subject influencing factor causes the document subject to change. The weight of any subject influencing factor represents the degree of influence of the subject influencing factor on the document subject.

[0128] A traversal unit is used to traverse all preprocessed documents. If there is at least one document topic corresponding to any preprocessed document that includes the search topic, the preprocessed document, at least one document topic corresponding to the preprocessed document, at least one topic influencing factor and the weight of each topic influencing factor are added to the document relationship tree.

[0129] The first retrieval unit is used to repeat the above traversal process until all pre-processed documents are traversed once, completing the retrieval of the first retrieval level and obtaining the document relationship tree after the retrieval of the first retrieval level.

[0130] The second retrieval unit is used to return to the above traversal process for all preprocessed documents retrieved in the first retrieval level, using any document subject other than the retrieval subject corresponding to each preprocessed document as a new retrieval subject, and perform retrieval at the next retrieval level until the retrieval of all retrieval levels is completed to obtain a document relationship tree.

[0131] In some optional implementations, all subject influencing factors in the document relationship tree include at least one target influencing factor;

[0132] The first extraction module 504 includes:

[0133] The first extraction unit is used to extract at least one first document including all target influencing factors from the document relationship tree based on at least one subject influencing factor corresponding to each preprocessed document in the document relationship tree.

[0134] The judging unit is used to judge whether the quantity of at least one first document is less than the first quantity.

[0135] The second extraction unit is used to extract a first number of target documents from the at least one first document based on the weight of each subject influencing factor and the weight of each target influencing factor corresponding to the at least one first document when the number of the at least one first document is not less than the first number.

[0136] In some optional embodiments, the second extraction unit includes:

[0137] The first determination subunit is used to determine, for any first document and any target influence factor, a subject influence factor consistent with the target influence factor from at least one subject influence factor corresponding to the first document.

[0138] The second determination subunit is used to determine the weight difference between the topic influencing factor and the target influencing factor.

[0139] The third determination subunit is used to use the sum of the weight differences between each target impact factor and the subject impact factor consistent with each target impact factor as the weight corresponding to the first document.

[0140] The fourth determining subunit is used to sort the first documents from small to large according to the weight corresponding to at least one first document, and determine a first number of first documents with forward ranking positions as target documents.

[0141] In some optional implementations, after the determination unit, the device further includes:

[0142] The third acquisition module is used to acquire at least one alternative search topic when the number of at least one first document is less than the first number. The alternative search topic is a potential document topic of any first document and / or a similar topic of any target influencing factor.

[0143] The first determining module is used to take the difference between the quantity of at least one first document and the first quantity as the second quantity.

[0144] The second extraction module is used to extract a second number of second documents including the alternative search topic after removing at least one first document from the document relationship tree for any alternative search topic.

[0145] The second determining module is used to take at least one first document and a second number of second documents as the first number of target documents.

[0146] In some optional implementations, the generating module 505 includes:

[0147] The weight assigning unit is used to assign a document weight to any target document based on the weight corresponding to the subject influencing factor of the target document, the weight corresponding to at least one target influencing factor and at least one alternative search subject.

[0148] The first generating unit is used to generate a plurality of document summaries corresponding to the first number of target documents by using a text summarization algorithm based on document weights corresponding to the first number of target documents.

[0149] The second generation unit is used to generate a literature review based on multiple literature abstracts by using a natural language generation algorithm.

[0150] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0151] The literature review generation device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0152] The embodiment of the present invention also provides a computer device having the above Figure 5 The literature review generation device shown.

[0153] See also Figure 6 , Figure 6is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 6 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 6 A processor 10 is taken as an example.

[0154] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0155] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0156] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0157] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0158] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 6 The example of connecting through bus is taken in the following.

[0159] The input device 30 can receive input digital or character information, and generate key signal input related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator bar, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display and a plasma display. In some optional embodiments, the display device may be a touch screen.

[0160] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0161] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0162] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the present invention.

Claims

1. A method for generating a literature review, characterized in that: The method comprises: Get the search topic and search level; Based on the search subject and the search level, a document relationship tree is retrieved from a document database; Obtain at least one target influencing factor, a weight of each target influencing factor, and a first quantity, wherein a change in the target influencing factor causes a change in the subject of the document, and the weight of any target influencing factor indicates the degree of influence of the target influencing factor on the subject of the document, wherein the target influencing factor includes experimental conditions, sample characteristics, method selection, and external environment; Based on the at least one target influencing factor and the weight of each target influencing factor, extracting a first number of target documents from the document relationship tree, wherein all subject influencing factors in the document relationship tree include the at least one target influencing factor; Generate a literature review based on the first number of target documents; Wherein, the step of retrieving a document relationship tree from a document database based on the search subject and the search level includes: When performing a search at the first search level, all documents in the document database are preprocessed to obtain all preprocessed documents; For any preprocessed document, subject analysis and influencing factor analysis are performed to obtain at least one document subject, at least one subject influencing factor and the weight of each subject influencing factor corresponding to the preprocessed document. The change of the subject influencing factor causes the document subject to change. The weight of any subject influencing factor indicates the degree of influence of the subject influencing factor on the document subject. Traversing all preprocessed documents, if there is at least one document subject corresponding to any preprocessed document that includes the search subject, adding the preprocessed document, at least one document subject corresponding to the preprocessed document, at least one subject influencing factor and the weight of each subject influencing factor to the document relationship tree; Repeat the above traversal process until all preprocessed documents are traversed once, completing the retrieval of the first retrieval level, and obtaining the document relationship tree after the retrieval of the first retrieval level; For all pre-processed documents retrieved in the first search level, any document subject other than the search subject corresponding to each pre-processed document is used as a new search subject, and the traversal process is returned to the above-mentioned process to perform a search at the next search level until all search levels are completed to obtain the document relationship tree; The step of extracting a first number of target documents from the document relationship tree based on the at least one target influencing factor and the weight of each target influencing factor comprises: Based on at least one subject influencing factor corresponding to each preprocessed document in the document relationship tree, extracting at least one first document including all target influencing factors from the document relationship tree; Determining whether the quantity of the at least one first document is less than a first quantity; When the number of the at least one first document is less than the first number, obtaining at least one candidate search topic, wherein the candidate search topic is a potential document topic of any first document and / or a similar topic of any target influencing factor; Taking the difference between the quantity of the at least one first document and the first quantity as the second quantity; For any candidate search topic, after removing the at least one first document from the document relationship tree, extracting a second number of second documents including the candidate search topic; taking the at least one first document and the second number of second documents as the first number of target documents; Alternatively, when the number of the at least one first document is not less than the first number, for any first document and any target influence factor, a subject influence factor consistent with the target influence factor is determined from at least one subject influence factor corresponding to the first document; Determining a weight difference between the subject influencing factor and the target influencing factor; The sum of the weight differences between each target impact factor and the subject impact factor consistent with each target impact factor is used as the weight corresponding to the first document; The at least one first document is sorted from small to large according to the weight corresponding to the at least one first document, and a first number of first documents with leading positions in the sorting are determined as target documents.

2. The method according to claim 1, characterized in that Generating a literature review based on the first number of target documents includes: For any target document, assign a document weight to the target document based on the weight corresponding to the subject influencing factor of the target document, the weight corresponding to the at least one target influencing factor and the at least one alternative search subject; Based on the document weights corresponding to the first number of target documents, a text summarization algorithm is used to generate a plurality of document summaries corresponding to the first number of target documents; A natural language generation algorithm is used to generate the literature review based on the multiple literature abstracts.

3. A device for generating a literature review, characterized in that: The device comprises: A first acquisition module is used to acquire a search topic and a search level; A retrieval module, used to retrieve a document relationship tree from a document database based on the retrieval subject and the retrieval level; A second acquisition module is used to acquire at least one target influencing factor, a weight of each target influencing factor and a first quantity, wherein a change in the target influencing factor causes a change in the subject of the document, and the weight of any target influencing factor indicates the degree of influence of the target influencing factor on the subject of the document, wherein the target influencing factor includes experimental conditions, sample characteristics, method selection and external environment; A first extraction module, configured to extract a first number of target documents from the document relationship tree based on the at least one target influencing factor and a weight of each target influencing factor, wherein all subject influencing factors in the document relationship tree include the at least one target influencing factor; A generating module, used for generating a literature review based on the first number of target documents; Wherein, the retrieval module specifically includes: When performing a search at the first search level, all documents in the document database are preprocessed to obtain all preprocessed documents; For any preprocessed document, subject analysis and influencing factor analysis are performed to obtain at least one document subject, at least one subject influencing factor and the weight of each subject influencing factor corresponding to the preprocessed document. The change of the subject influencing factor causes the document subject to change. The weight of any subject influencing factor indicates the degree of influence of the subject influencing factor on the document subject. Traversing all preprocessed documents, if there is at least one document subject corresponding to any preprocessed document that includes the search subject, adding the preprocessed document, at least one document subject corresponding to the preprocessed document, at least one subject influencing factor and the weight of each subject influencing factor to the document relationship tree; Repeat the above traversal process until all preprocessed documents are traversed once, completing the retrieval of the first retrieval level, and obtaining the document relationship tree after the retrieval of the first retrieval level; For all pre-processed documents retrieved in the first search level, any document subject other than the search subject corresponding to each pre-processed document is used as a new search subject, and the traversal process is returned to the above-mentioned process to perform a search at the next search level until all search levels are completed to obtain the document relationship tree; The first extraction module specifically includes: Based on at least one subject influencing factor corresponding to each preprocessed document in the document relationship tree, extracting at least one first document including all target influencing factors from the document relationship tree; Determining whether the quantity of the at least one first document is less than a first quantity; When the number of the at least one first document is less than the first number, obtaining at least one candidate search topic, wherein the candidate search topic is a potential document topic of any first document and / or a similar topic of any target influencing factor; Taking the difference between the quantity of the at least one first document and the first quantity as the second quantity; For any candidate search topic, after removing the at least one first document from the document relationship tree, extracting a second number of second documents including the candidate search topic; taking the at least one first document and the second number of second documents as the first number of target documents; Alternatively, when the number of the at least one first document is not less than the first number, for any first document and any target influence factor, a subject influence factor consistent with the target influence factor is determined from at least one subject influence factor corresponding to the first document; Determining a weight difference between the subject influencing factor and the target influencing factor; The sum of the weight differences between each target impact factor and the subject impact factor consistent with each target impact factor is used as the weight corresponding to the first document; The at least one first document is sorted from small to large according to the weight corresponding to the at least one first document, and a first number of first documents with leading positions in the sorting are determined as target documents.

4. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the literature review generation method described in any one of claims 1 or 2 by executing the computer instructions.

5. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the literature review generation method described in any one of claims 1 or 2.

6. A computer program product, characterized in that It includes computer instructions, which are used to enable a computer to execute the literature review generation method described in any one of claims 1 or 2.

Images