Abstract generation method, device and equipment

Abstract

The present invention discloses a summary generation method, device and equipment, which relates to the field of data processing technology, wherein the method comprises: obtaining an original text; extracting content from the original text to obtain a first summary and a knowledge graph of the original text; extracting key entities and key relationships from the knowledge graph according to the first summary to obtain key content; translating the key content into a target language to obtain translation content; and generating a target summary according to the translation content. In the above manner, the present invention realizes the automatic generation of text summaries across languages.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method, device and equipment for generating a summary. Background Art

[0002] The internationalization strategy of an enterprise is the development strategy of its products and services outside the local market. As the strength of enterprises continues to grow and the domestic market gradually becomes saturated, some entrepreneurs have begun to turn their attention to the global overseas market.

[0003] One of the most important aspects of a company's internationalization strategy is that product introductions can be supported in multiple languages. Existing product introductions are usually literal translations, but some special nouns or specific words may be translated incorrectly after literal translation. In addition, some product introductions are very cumbersome in the initial version and have no abstracts. In order to achieve the company's internationalization strategy, how to automatically generate text abstracts across languages ​​is an urgent problem to be solved by people in this field. Summary of the invention

[0004] In order to solve the above problems, a summary generation method, apparatus and device according to an embodiment of the present invention are proposed.

[0005] According to one aspect of an embodiment of the present invention, a method for generating a summary is provided, comprising:

[0006] Get the original text;

[0007] Extracting content from the original text to obtain a first summary and a knowledge graph of the original text;

[0008] According to the first summary, key entities and key relationships are extracted in the knowledge graph to obtain key content;

[0009] Translating the key content into a target language to obtain translated content;

[0010] A target summary is generated according to the translated content.

[0011] Optionally, performing content extraction on the original text to obtain a first summary and a knowledge graph of the original text includes:

[0012] Extracting text information of the original text to obtain a first summary of the original text;

[0013] The entity content and relationship content of the original text are extracted to obtain a knowledge graph of the original text.

[0014] Optionally, extracting text information of the original text to obtain a first summary of the original text includes:

[0015] Analyzing text information in the original text to obtain an analysis result, wherein the text information includes at least one sentence;

[0016] Taking sentences in a preset format in the analysis result as extraction units, obtaining at least one extraction unit;

[0017] Predicting the at least one extraction unit to obtain a prediction result;

[0018] A preset number of extraction units with the highest prediction scores in the prediction results are concatenated to obtain a first summary of the original text.

[0019] Optionally, taking sentences in a preset format in the analysis result as extraction units to obtain at least one extraction unit includes:

[0020] If the analysis result is that the sentence containing the clause does not have a clause node, the whole sentence containing the clause is used as the extraction unit;

[0021] If the clause node of the first sentence whose analysis result is a clause includes the clause node of the second sentence whose analysis result is a clause, the first sentence is used as the extraction unit.

[0022] Optionally, extracting entity content and relationship content of the original text to obtain a knowledge graph of the original text includes:

[0023] Extracting entity content and relationship content of the original text to obtain an extraction result;

[0024] Classifying the entity content and the relationship content in the extraction result to obtain at least one classification result, wherein each classification result is in the form of a triple of a head entity, a relationship, and a tail entity;

[0025] The at least one classification result is concatenated to obtain a knowledge graph of the original text.

[0026] Optionally, after obtaining the extraction results, it also includes:

[0027] The entity contents with the same information in the extraction results are unified into the same entity content, wherein the entity content is one of the same entity contents.

[0028] Optionally, according to the first summary, key entities and key relationships are extracted from the knowledge graph to obtain key content, including:

[0029] In the knowledge graph, extract entity content and relationship content appearing in the first summary to obtain key entities and key relationships;

[0030] The key entities and the key relationships are concatenated to obtain the key content.

[0031] According to another aspect of an embodiment of the present invention, a summary generation device is provided, the device comprising:

[0032] The acquisition module is used to obtain the original text;

[0033] A processing module is used to extract content from the original text to obtain a first summary and a knowledge graph of the original text; extract key entities and key relationships from the knowledge graph according to the first summary to obtain key content; and translate the key content into a target language to obtain translated content;

[0034] A generation module is used to generate a target summary according to the translation content.

[0035] According to another aspect of an embodiment of the present invention, there is provided a computing device, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0036] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned summary generation method.

[0037] According to another aspect of the embodiments of the present invention, a computer storage medium is provided, wherein at least one executable instruction is stored in the storage medium, and the executable instruction enables a processor to perform operations corresponding to the above-mentioned summary generation method.

[0038] According to the solution provided by the above embodiment of the present invention, the original text is obtained; content is extracted from the original text to obtain a first summary and a knowledge graph of the original text; key entities and key relationships are extracted from the knowledge graph based on the first summary to obtain key content; the key content is translated into a target language to obtain translated content; and a target summary is generated based on the translated content, thereby realizing automatic generation of text summaries across languages.

[0039] The above description is only an overview of the technical solution of the embodiment of the present invention. In order to more clearly understand the technical means of the embodiment of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the embodiment of the present invention more obvious and easy to understand, the specific implementation method of the embodiment of the present invention is specifically listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the embodiments of the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0041] Figure 1 A flowchart of a method for generating a summary provided by an embodiment of the present invention is shown;

[0042] Figure 2 A schematic diagram of a cross-language summarization solution based on knowledge enhancement provided by an embodiment of the present invention is shown;

[0043] Figure 3 A schematic diagram showing the structure of a summary generation device provided by an embodiment of the present invention is shown;

[0044] Figure 4 A schematic diagram of the structure of a computing device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0045] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to enable the scope of the present invention to be fully communicated to those skilled in the art.

[0046] Figure 1 FIG. 4 is a flowchart of a method for generating a summary according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0047] Step 11, obtaining the original text;

[0048] Step 12, extracting content from the original text to obtain a first summary and a knowledge graph of the original text;

[0049] Step 13, extracting key entities and key relationships in the knowledge graph according to the first summary to obtain key content;

[0050] Step 14, translating the key content into a target language to obtain translated content;

[0051] Step 15: Generate a target summary based on the translated content.

[0052] In this embodiment, the original text is obtained; content is extracted from the original text to obtain a first summary and a knowledge graph of the original text; key entities and key relationships are extracted from the knowledge graph based on the first summary to obtain key content; the key content is translated into a target language to obtain translated content; and a target summary is generated based on the translated content, thereby achieving automatic generation of text summaries across languages.

[0053] In an optional embodiment of the present invention, step 12 may include:

[0054] Step 121, extracting text information of the original text to obtain a first summary of the original text;

[0055] Step 122, extracting the entity content and relationship content of the original text to obtain a knowledge graph of the original text.

[0056] In this embodiment, the core idea of ​​extracting information and knowledge is to extract semantic information and knowledge information from the text, and to assist the extraction of core knowledge information through semantic information, wherein semantic information refers to information expressed in daily language, which includes not only information provided by language, such as weather forecasts, propositions or descriptive statements, predictions, information provided by scientific theories, but also information provided by thermometers, scales, mathematical formulas, traffic signs, disease symptoms, test data, and pictures and videos.

[0057] In another optional embodiment of the present invention, step 121 may include:

[0058] Step 1211, analyzing the text information in the original text to obtain an analysis result, wherein the text information includes at least one sentence;

[0059] Step 1212, taking the sentences in the preset form in the analysis result as extraction units, and obtaining at least one extraction unit; if the sentence whose analysis result is a clause does not have a clause node, taking the whole sentence containing the clause as the extraction unit; if the clause node of the first sentence whose analysis result is a clause includes the clause node of the second sentence whose analysis result is a clause, taking the first sentence as the extraction unit.

[0060] Step 1213, predicting the at least one extraction unit to obtain a prediction result;

[0061] Step 1214: concatenate a preset number of extraction units with the highest prediction scores in the prediction results to obtain a first summary of the original text.

[0062] In this embodiment, extraction units of pre-set types are extracted from the original text through an extraction method, and the first summary is obtained by splicing and combining the extraction units. The types of extraction units are generally divided into sentences, phrases or words, but are not limited to the above. The first summary obtained by this method can achieve a better summary effect, and can separate the key information and non-key information in the evidence through fine-grained extraction units, avoiding the redundant existence of sentences and the extraction of unnecessary information. Moreover, since the first summary is not the final target summary required, the problem of stiff connection of the extracted sentences can be tolerated.

[0063] Specifically, the method for extracting the first summary includes: the first step of taking the clause as the extraction unit:

[0064] In the first sub-step, the syntactic component analysis of the sentence is performed using an open source university treebank method, and each clause node S and SBAR is regarded as a clause unit, where the clause node S refers to a simple declarative clause, that is, a sentence that is not introduced by a subordinate conjunction and does not show subject-verb inversion, and the SBAR refers to a clause introduced by a subordinate conjunction.

[0065] In the second sub-step, after removing the root node, if a clause node is included in a clause node at a higher level, the clause node at the highest level is selected as the extracted clause unit. At the same time, if there is no clause node after a sentence is parsed, the entire sentence is directly selected as the extraction unit, which can retain more complete semantic information.

[0066] The second step is to extract sub-sentences based on the BERT sub-sentence extraction model (Sub-Sentential Extraction, SSE):

[0067] In the third sub-step, the open source SSE model is used to perform average pooling on the tokens within the clause range to obtain the representation of the clause unit. After collecting the representations of all clause units, a layer of Transformer layer is used to mix the information between the upper and lower sentences, and a layer of fully connected layer is used to perform binary classification to predict whether the current clause unit needs to be extracted. The Transformer layer refers to a model that uses the attention mechanism to increase the model training speed.

[0068] The fourth sub-step is to learn whether each clause needs to be extracted during training, and select the top N clauses with the highest scores for concatenation and output during prediction.

[0069] In yet another optional embodiment of the present invention, step 122 may include:

[0070] Step 1221, extracting entity content and relationship content of the original text to obtain an extraction result;

[0071] Step 1222, classifying the entity content and the relationship content in the extraction result to obtain at least one classification result, wherein each classification result is in the form of a triple of a head entity, a relationship, and a tail entity;

[0072] Step 1223, concatenate the at least one classification result to obtain a knowledge graph of the original text.

[0073] In this embodiment, taking the field of network security as an example, obtaining the knowledge graph of the original text includes the following three steps: The third step is entity and relationship extraction, that is, extracting STIX entities (STIX TM Domain Objects (SDO) and STIX relationship (STIX TM STIX (Structured Threat Information Expression) is a language and serialization format for exchanging network threat intelligence. Its application scenarios include collaborative threat analysis, automated threat intelligence exchange, automated threat detection and response, etc., but are not limited to the above:

[0074] The fifth sub-step is corpus construction: Build a corpus based on the rich STIX standard threat intelligence data within the enterprise and threat knowledge base data such as hacker organizations, malware, tools and attack methods.

[0075] The sixth sub-step is entity extraction: entity extraction is performed through the Transformer and Conditional Random Field (CRF) model.

[0076] The seventh sub-step is relationship classification: based on the results of entity extraction and the predefined security ontology, the relationship between the head entity and the tail entity is classified; for example, the relationship between the head entity of the attack mode type and the tail entity of the malware, according to the definition of the security ontology, there are only two types of relationships: transmission and utilization; therefore, for the determined head and tail entity pairs, the relationship between them is converted into a multi-classification problem based on the security ontology, which is processed by the Transformer and normalized exponential function (Softmax) multi-classification model.

[0077] The fourth step is to align entities and relationships: the specific explanation is as follows: 12211.

[0078] Step 5: Knowledge graph splicing: splice the <head entity, relationship, tail entity> triples to obtain the knowledge graph of the text.

[0079] In another optional embodiment of the present invention, after step 1221, the following steps may be further included:

[0080] Step 12211, unifying entity contents with the same information in the extraction result into the same entity content, wherein the entity content is one of the same entity contents.

[0081] In this embodiment, taking the field of network security as an example, an entity link dictionary is constructed for entities and relationship information such as attack organizations through the security ontology model and threat knowledge base data in the corpus. Aliases and other information similar to attack organizations are replaced based on the dictionary, for example: the entity name is "APT-C-09", and the aliases are "Dropping Elephant", "Chinastrats", "Hangover", "CANDLEFISH", "Patchwork", "Viceroy Tiger", "Monsoon", "Sarit", "Maha Grass", and "White Elephant". Among them, if an alias appears, it is necessary to replace the entity name with the entity name to achieve the alignment of the entity and the relationship, that is, replace the alias with the entity name.

[0082] In another optional embodiment of the present invention, step 13 may include:

[0083] Step 131, in the knowledge graph, extracting entity content and relationship content appearing in the first summary to obtain key entities and key relationships;

[0084] Step 132, concatenate the key entity and the key relationship to obtain the key content.

[0085] In this embodiment, the most important thing in the target summary is the core context and factual knowledge of the article, so the entity information of the summary is extracted based on the entity extraction model in the knowledge extraction method. That is, the knowledge information of the auxiliary core is extracted through the first summary to obtain the core knowledge graph covered by the first summary, which can restore the factual knowledge in the original text to the greatest extent. The method is to perform multi-constrained minimum cost maximum connected subgraph calculation on the text knowledge graph, which is specifically divided into the following three steps:

[0086] Step 6: Initialize the weight of each edge in the knowledge graph to 1.

[0087] In the seventh step, the entity information of the first summary is used as the constraint set of the graph nodes, requiring that the minimum cost maximum connected subgraph calculated from the knowledge graph covers as many entity nodes as possible.

[0088] The eighth step is to extract the minimum cost maximum connected subgraph in the text knowledge graph based on the Prime algorithm (Prime) minimum spanning tree algorithm. The maximal connected subgraph is also called the maximal connected subgraph, which is a subgraph in mathematics and graph theory, that is, a subgraph that connects all the nodes of the graph with the least edges, so the maximal connected subgraph is not unique.

[0089] In another optional embodiment of the present invention, step 14 may include:

[0090] In the ninth step, the entities and relationships in the multi-constrained minimum cost maximum connectivity subgraph calculated in the eighth step are translated into the target language through a translation interface.

[0091] In yet another optional embodiment of the present invention, step 15 may include:

[0092] The tenth step is to generate text in the target language through a latent variable autoencoder based on the translated entity and relationship information.

[0093] Figure 2 A schematic diagram of a cross-language summary solution based on knowledge enhancement provided by an embodiment of the present invention is shown. Figure 2 As shown in the figure, the overall solution consists of four parts: information and knowledge extraction module, multi-constrained minimum cost maximum connected subgraph calculation module, translation module and text generation module. This solution is aimed at the vertical field of threat intelligence and aims to provide technical support for the internationalization of threat intelligence applications.

[0094] In the above-mentioned embodiment of the present invention, the overall solution is based on the core knowledge graph of factual knowledge, which can take over downstream translation and text generation tasks for different target languages, has high robustness and flexibility, is conducive to industrial deployment, and creatively combines the entity information of extractive summaries. Through factual knowledge extraction, a core knowledge graph of text information is constructed based on core factual knowledge and is highly interpretable.

[0095] Figure 3 FIG. 3 is a schematic diagram showing the structure of a summary generating device 30 provided in an embodiment of the present invention. Figure 3 As shown, the device comprises:

[0096] An acquisition module 31 is used to acquire the original text;

[0097] The processing module 32 is used to extract the content of the original text to obtain a first summary and a knowledge graph of the original text; extract key entities and key relationships in the knowledge graph according to the first summary to obtain key content; and translate the key content into a target language to obtain translated content;

[0098] The generating module 33 is used to generate a target summary according to the translated content.

[0099] Optionally, the processing module 32 is further configured to extract text information of the original text to obtain a first summary of the original text;

[0100] The entity content and relationship content of the original text are extracted to obtain a knowledge graph of the original text.

[0101] Optionally, the processing module 32 is further used to analyze text information in the original text to obtain an analysis result, wherein the text information includes at least one sentence;

[0102] Taking sentences in a preset format in the analysis result as extraction units, obtaining at least one extraction unit;

[0103] Predicting the at least one extraction unit to obtain a prediction result;

[0104] A preset number of extraction units with the highest prediction scores in the prediction results are concatenated to obtain a first summary of the original text.

[0105] Optionally, the processing module 32 is further configured to use the entire sentence containing the clause as an extraction unit if the analysis result is that the sentence containing the clause does not have a clause node;

[0106] If the clause node of the first sentence whose analysis result is a clause includes the clause node of the second sentence whose analysis result is a clause, the first sentence is used as the extraction unit.

[0107] Optionally, the processing module 32 is further used to extract entity content and relationship content of the original text to obtain an extraction result;

[0108] Classifying the entity content and the relationship content in the extraction result to obtain at least one classification result, wherein each classification result is in the form of a triple of a head entity, a relationship, and a tail entity;

[0109] The at least one classification result is concatenated to obtain a knowledge graph of the original text.

[0110] Optionally, the processing module 32 is further configured to unify entity contents with the same information in the extraction result into the same entity content, wherein the entity content is one of the same entity contents.

[0111] Optionally, the processing module 32 is further used to extract entity content and relationship content appearing in the first summary in the knowledge graph to obtain key entities and key relationships;

[0112] The key entities and the key relationships are concatenated to obtain the key content.

[0113] It should be understood that the above Figure 1 to Figure 2 The description of the illustrated method embodiment is only an explanation of the technical solution of the present invention in the form of optional examples, and does not constitute a limitation on the summary generation method of the present invention. In other implementation modes, the execution steps and order of the summary generation method of the present invention may be different from the above-mentioned embodiment, and the embodiment of the present invention is not limited to this.

[0114] It should be noted that this embodiment is a device embodiment corresponding to the above method embodiment, and all implementation methods in the above method embodiment are applicable to the embodiment of this device and can achieve the same technical effect.

[0115] An embodiment of the present invention provides a non-volatile computer storage medium, wherein the computer storage medium stores at least one executable instruction, and the computer executable instruction can execute the digest generation method in any of the above method embodiments.

[0116] Figure 4 A schematic diagram of the structure of a computing device provided by an embodiment of the present invention is shown, and the specific embodiment of the present invention does not limit the specific implementation of the computing device.

[0117] like Figure 4 As shown, the computing device may include: a processor, a communications interface, a memory, and a communication bus.

[0118] The processor, the communication interface, and the memory communicate with each other via a communication bus. The communication interface is used to communicate with other devices such as a client or other server network elements. The processor is used to execute a program, specifically, to execute the relevant steps in the above-mentioned summary generation method embodiment for a computing device.

[0119] Specifically, the program may include program codes including computer operation instructions.

[0120] The processor may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. The one or more processors included in the computing device may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0121] The memory is used to store programs. The memory may include a high-speed RAM memory, and may also include a non-volatile memory, such as at least one disk memory.

[0122] The program can be specifically used to enable the processor to execute the summary generation method in any of the above method embodiments. The specific implementation of each step in the program can refer to the corresponding descriptions in the corresponding steps and units in the above summary generation method embodiments, which will not be repeated here. A person skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the above-described device and module can refer to the corresponding process description in the above method embodiments, which will not be repeated here.

[0123] The algorithm or display provided herein is not inherently related to any particular computer, virtual system or other device. Various general purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious that the structure required for constructing this type of system. In addition, the embodiment of the present invention is not directed to any specific programming language either. It should be understood that various programming languages ​​can be utilized to realize the content of the embodiment of the present invention described herein, and the description of the above specific language is to disclose the best mode of the embodiment of the present invention.

[0124] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0125] Similarly, it should be understood that in order to streamline the embodiments of the present invention and to aid in understanding one or more of the various aspects of the invention, in the above description of exemplary embodiments of the present invention, the various features of the embodiments of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof.

[0126] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and furthermore may be divided into a plurality of submodules or subunits or subcomponents.

[0127] Furthermore, those skilled in the art will appreciate that although some embodiments herein include certain features included in other embodiments but not other features, the combination of features from different embodiments is meant to be within the scope of the present invention and to form different embodiments.

[0128] The various component embodiments of the present invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or digital signal processor (DSP) may be used in practice to implement some or all of the functions of some or all of the components according to the embodiments of the present invention. The embodiments of the present invention may also be implemented as a device or apparatus program (e.g., a computer program and a computer program product) for executing part or all of the methods described herein. Such a program implementing an embodiment of the present invention may be stored on a computer-readable medium, or may have the form of one or more signals. Such a signal may be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0129] It should be noted that the above embodiments illustrate the embodiments of the present invention rather than limiting the present invention, and the word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The embodiments of the present invention can be implemented by means of hardware including a number of different elements and by means of a suitably programmed computer. The use of the words first, second, and third, etc. does not indicate any order. These words can be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be understood as limiting the order of execution.

Claims

1. A method for generating a summary, characterized in that: The method comprises: Get the original text; Extracting content from the original text to obtain a first summary and a knowledge graph of the original text; According to the first summary, key entities and key relationships are extracted in the knowledge graph to obtain key content; Translating the key content into a target language to obtain translated content; Based on the translation content, a target summary is generated, Extracting text information of the original text to obtain a first summary of the original text includes: Analyzing text information in the original text to obtain an analysis result, wherein the text information includes at least one sentence; Taking sentences in a preset format in the analysis result as extraction units, obtaining at least one extraction unit; Predicting the at least one extraction unit to obtain a prediction result; splicing a preset number of extraction units with the highest prediction scores in the prediction results to obtain a first summary of the original text, The sentences in the preset form in the analysis result are used as extraction units to obtain at least one extraction unit, including: If the analysis result is that the sentence containing the clause does not have a clause node, the whole sentence containing the clause is used as the extraction unit; If the clause node of the first sentence whose analysis result is a clause includes the clause node of the second sentence whose analysis result is a clause, the first sentence is used as the extraction unit.

2. The method for generating a summary according to claim 1, characterized in that: Extracting content from the original text to obtain a knowledge graph of the original text includes: The entity content and relationship content of the original text are extracted to obtain a knowledge graph of the original text.

3. The method for generating a summary according to claim 2, characterized in that: Extracting entity content and relationship content of the original text to obtain a knowledge graph of the original text includes: Extracting entity content and relationship content of the original text to obtain an extraction result; Classifying the entity content and the relationship content in the extraction result to obtain at least one classification result, wherein each classification result is in the form of a triple of a head entity, a relationship, and a tail entity; The at least one classification result is concatenated to obtain a knowledge graph of the original text.

4. The method for generating a summary according to claim 3, characterized in that: After getting the extraction results, it also includes: The entity contents with the same information in the extraction results are unified into the same entity content, wherein the entity content is one of the same entity contents.

5. The method for generating a summary according to claim 1, wherein: According to the first summary, key entities and key relationships are extracted from the knowledge graph to obtain key content, including: In the knowledge graph, extract entity content and relationship content appearing in the first summary to obtain key entities and key relationships; The key entities and the key relationships are concatenated to obtain the key content.

6. A summary generation device, characterized in that: The device comprises: The acquisition module is used to obtain the original text; A processing module is used to extract content from the original text to obtain a first summary and a knowledge graph of the original text; extract key entities and key relationships from the knowledge graph according to the first summary to obtain key content; and translate the key content into a target language to obtain translated content; A generation module, used to generate a target summary according to the translation content, Extracting text information of the original text to obtain a first summary of the original text includes: Analyzing text information in the original text to obtain an analysis result, wherein the text information includes at least one sentence; Taking sentences in a preset format in the analysis result as extraction units, obtaining at least one extraction unit; Predicting the at least one extraction unit to obtain a prediction result; splicing a preset number of extraction units with the highest prediction scores in the prediction results to obtain a first summary of the original text, The sentences in the preset form in the analysis result are used as extraction units to obtain at least one extraction unit, including: If the analysis result is that the sentence containing the clause does not have a clause node, the whole sentence containing the clause is used as the extraction unit; If the clause node of the first sentence whose analysis result is a clause includes the clause node of the second sentence whose analysis result is a clause, the first sentence is used as the extraction unit.

7. A computing device comprising: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and when the at least one executable instruction is executed, the processor executes the summary generation method according to any one of claims 1 to 5.

8. A computer storage medium, wherein at least one executable instruction is stored in the storage medium, and when the executable instruction is executed, a computing device executes the summary generation method according to any one of claims 1 to 5.

Images