Document information summarizing method, system, electronic equipment and medium
阅读说明：本技术 一种文档信息汇总方法、系统、电子设备及介质 (Document information summarizing method, system, electronic equipment and medium ) 是由 杨康 徐凯波 孙泽懿 王硕 于 2021-06-22 设计创作，主要内容包括：本申请公开了一种文档信息汇总方法、系统、电子设备及介质,文档信息汇总方法包括：隐藏向量值计算步骤：对源文档进行编码特征计算,获取源文档隐藏向量值后,根据所述源文档隐藏向量值计算获得目标文档隐藏向量值；输出向量计算步骤：根据所述源文档隐藏向量值与所述目标文档隐藏向量值计算获得输出向量值；文档信息汇总结果获取步骤:根据所述输出向量值计算获得所述文档信息汇总结果。本发明通过机器学习和图算法结合,利用机器学习完成汇总生成,利用图算法提取多文档之间的关系特征,从而实现基于机器学习的多文档信息生成,达到利用图算法弥补机器学习多文档信息生成不足的缺点。(The application discloses a method, a system, electronic equipment and a medium for summarizing document information, wherein the method for summarizing document information comprises the following steps: a hidden vector value calculating step: coding feature calculation is carried out on a source document, after a source document hiding vector value is obtained, a target document hiding vector value is obtained through calculation according to the source document hiding vector value; an output vector calculation step: calculating according to the source document hiding vector value and the target document hiding vector value to obtain an output vector value; and a step of obtaining a document information summarizing result, which is to calculate and obtain the document information summarizing result according to the output vector value. The invention combines machine learning and graph algorithm, completes summary generation by utilizing the machine learning, and extracts the relation characteristics among multiple documents by utilizing the graph algorithm, thereby realizing the generation of the multiple document information based on the machine learning and overcoming the defect of insufficient generation of the multiple document information by utilizing the graph algorithm.)
1. A document information summarizing method is characterized by comprising the following steps:
a hidden vector value calculating step: coding feature calculation is carried out on a source document, after a source document hiding vector value is obtained, a target document hiding vector value is obtained through calculation according to the source document hiding vector value;
an output vector calculation step: calculating according to the source document hiding vector value and the target document hiding vector value to obtain an output vector value;
and a step of obtaining a document information summarizing result, which is to calculate and obtain the document information summarizing result according to the output vector value.
2. The document information summarizing method according to claim 1, wherein said hidden vector value calculating step includes:
a source document hidden vector value calculation step: after coding feature calculation is carried out on a plurality of source documents through a neural network coder, a source document hidden vector value is obtained;
calculating a hidden vector value of the target document: and according to the source document hiding vector value, carrying out feature calculation on the target document through a graph algorithm, and then obtaining the target document hiding vector value.
3. The document information summarizing method according to claim 1, wherein the output vector calculating step includes obtaining the output vector value by an attention module calculation according to the source document hiding vector value and the target document hiding vector value.
4. The method according to claim 1, wherein the step of obtaining the summary document information result includes obtaining the summary document information result by a decoder calculation according to the output vector value.
5. A document information summarizing system applied to the document information summarizing method according to any one of claims 1 to 4, the document information summarizing system comprising:
a hidden vector value calculation unit: coding feature calculation is carried out on a source document, after a source document hiding vector value is obtained, a target document hiding vector value is obtained through calculation according to the source document hiding vector value;
an output vector calculation unit: calculating according to the source document hiding vector value and the target document hiding vector value to obtain an output vector value;
and the document information summarizing result acquisition unit calculates and acquires the document information summarizing result according to the output vector value.
6. The document information summarizing system according to claim 5, wherein said hidden vector value calculating step unit:
a source document hidden vector value calculation module: after coding feature calculation is carried out on a plurality of source documents through a neural network coder, a source document hidden vector value is obtained;
a target document hidden vector value calculation module: and according to the source document hiding vector value, carrying out feature calculation on the target document through a graph algorithm, and then obtaining the target document hiding vector value.
7. The document information summarizing system of claim 6, wherein the output vector value is computed and obtained by an attention module according to the source document hiding vector value and the target document hiding vector value.
8. The document information summarizing system according to claim 7, wherein the document information summarizing result is obtained by a decoder calculation according to the output vector value.
9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the document information summarizing method according to any one of claims 1 to 4 when executing the computer program.
10. An electronic device readable storage medium having stored thereon computer program instructions which, when executed by the processor, implement the document information summarizing method according to any one of claims 1 to 4.
Technical Field
The present application relates to the field of deep learning technologies, and in particular, to a method, a system, an electronic device, and a medium for summarizing document information.
Background
At present, in an information explosion age, various information reports are needed in many industries, and the information sources may be originated from different articles, news and icons, which needs a certain summarizing capability, so that finally obtained information is diversified and accurate. The prior art needs high cost for summarizing contents from multiple documents, either manually or by machine. Manual collection requires a large amount of labor cost, and wastes time and labor; machine summarization is mainly biased to single-document summary generation and has high quality, but many information have certain relevance among cross-documents, certain relevance is lost only from single-document summary generation, and a multi-document summarization technology using a machine has certain performance and methodology loss, so that a plurality of defects existing in machine learning multi-document information generation cannot be overcome by the prior art.
Disclosure of Invention
The embodiment of the application provides a document information summarizing method, a document information summarizing system, electronic equipment and a document information summarizing medium, and at least solves the problems of lack of relevance of document information, high cost and the like in the process of manually summarizing multi-document information by a machine.
The invention provides a document information summarizing method, which comprises the following steps:
a hidden vector value calculating step: coding feature calculation is carried out on a source document, after a source document hiding vector value is obtained, a target document hiding vector value is obtained through calculation according to the source document hiding vector value;
an output vector calculation step: calculating according to the source document hiding vector value and the target document hiding vector value to obtain an output vector value;
and a step of obtaining a document information summarizing result, which is to calculate and obtain the document information summarizing result according to the output vector value.
In the above document information summarizing method, the hidden vector value calculating step includes:
a source document hidden vector value calculation step: after coding feature calculation is carried out on a plurality of source documents through a neural network coder, a source document hidden vector value is obtained;
calculating a hidden vector value of the target document: and according to the source document hiding vector value, carrying out feature calculation on the target document through a graph algorithm, and then obtaining the target document hiding vector value.
In the above document information summarizing method, the step of calculating the output vector includes calculating and obtaining the output vector value through an attention module according to the source document hiding vector value and the target document hiding vector value.
In the above document information summarizing method, the obtaining step of the document information summarizing result includes obtaining the document information summarizing result by a decoder according to the output vector value.
The present invention further provides a document information summarizing system, which is suitable for the document information summarizing method described above, and the document information summarizing system includes:
a hidden vector value calculation unit: coding feature calculation is carried out on a source document, after a source document hiding vector value is obtained, a target document hiding vector value is obtained through calculation according to the source document hiding vector value;
an output vector calculation unit: calculating according to the source document hiding vector value and the target document hiding vector value to obtain an output vector value;
and the document information summarizing result acquisition unit calculates and acquires the document information summarizing result according to the output vector value.
In the above document information summarizing system, the hidden vector value calculating step unit:
a source document hidden vector value calculation module: after coding feature calculation is carried out on a plurality of source documents through a neural network coder, a source document hidden vector value is obtained;
a target document hidden vector value calculation module: and according to the source document hiding vector value, carrying out feature calculation on the target document through a graph algorithm, and then obtaining the target document hiding vector value.
In the document information summarizing system, the output vector value is calculated and obtained through an attention module according to the source document hiding vector value and the target document hiding vector value.
In the document information summarizing system, the document information summarizing result is obtained through calculation of a decoder according to the output vector value.
The invention also provides an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, and is characterized in that the processor implements any one of the document information summarizing methods when executing the computer program.
The invention also provides an electronic device readable storage medium, wherein the electronic device readable storage medium stores computer program instructions, and when the computer program instructions are executed by the processor, the method for summarizing the document information is realized, and the natural language processing capability is improved.
Compared with the related art, the document information summarizing method, the document information summarizing system, the electronic equipment and the medium provided by the invention can replace manual work to summarize document information contents from multiple documents, and make up the defect of generating the multiple document information by a traditional machine learning method through a graph algorithm.
The details of one or more embodiments of the application are set forth in the accompanying drawings and the description below to provide a more thorough understanding of the application.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:
FIG. 1 is a flowchart of a document information summarization method according to an embodiment of the present application;
FIG. 2 is a diagram of a multi-document information generation framework according to an embodiment of the present application;
FIG. 3 is a schematic structural diagram of a document information summarizing system according to the present invention;
fig. 4 is a frame diagram of an electronic device according to an embodiment of the present application.
Wherein the reference numerals are:
a hidden vector value calculation unit: 51;
an output vector calculation unit: 52;
a document information summary result acquisition unit: 53;
a source document hidden vector value calculation module: 511;
a target document hidden vector value calculation module: 512;
80 parts of a bus;
a processor: 81;
a memory: 82;
a communication interface: 83.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments provided in the present application without any inventive step are within the scope of protection of the present application.
It is obvious that the drawings in the following description are only examples or embodiments of the present application, and that it is also possible for a person skilled in the art to apply the present application to other similar contexts on the basis of these drawings without inventive effort. Moreover, it should be appreciated that such a development effort might be complex and tedious, but would nevertheless be a routine undertaking of design, fabrication, and manufacture for those of ordinary skill having the benefit of this disclosure, and thus should not be construed as a limitation of this disclosure.
Reference in the specification to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the specification. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those of ordinary skill in the art will explicitly and implicitly appreciate that the embodiments described herein may be combined with other embodiments without conflict.
Unless defined otherwise, technical or scientific terms referred to herein shall have the ordinary meaning as understood by those of ordinary skill in the art to which this application belongs. Reference to "a," "an," "the," and similar words throughout this application are not to be construed as limiting in number, and may refer to the singular or the plural. The present application is directed to the use of the terms "including," "comprising," "having," and any variations thereof, which are intended to cover non-exclusive inclusions; for example, a process, method, system, article, or apparatus that comprises a list of steps or modules (elements) is not limited to the listed steps or elements, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus. Reference to "connected," "coupled," and the like in this application is not intended to be limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The term "plurality" as referred to herein means two or more. "and/or" describes an association relationship of associated objects, meaning that three relationships may exist, for example, "A and/or B" may mean: a exists alone, A and B exist simultaneously, and B exists alone. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship. Reference herein to the terms "first," "second," "third," and the like, are merely to distinguish similar objects and do not denote a particular ordering for the objects.
An image algorithm refers to an algorithm used to process an image. The method comprises the steps of image denoising, image transformation, image analysis, image compression, image enhancement, image blurring processing and the like. Machine Learning (ML) is a multi-domain cross discipline, and relates to a plurality of disciplines such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and the like. The special research on how a computer simulates or realizes the learning behavior of human beings so as to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve the performance of the computer. It is the core of artificial intelligence, and is a fundamental way for computer to possess intelligence, and its application is extensive in every field of artificial intelligence, and it mainly uses induction, synthesis, rather than deduction. The environment provides some information to the learning part of the system, the learning part uses the information to modify the knowledge base to improve the efficiency of the system execution part to complete the task, the execution part completes the task according to the knowledge base, and simultaneously feeds back the obtained information to the learning part. In a specific application, the environment, the knowledge base and the execution part determine specific work content, and the problem to be solved by the learning part is completely determined by the part 3. We describe the impact of these 3 sections on the design learning system separately below. The most important factor affecting the design of a learning system is the information provided by the environment to the system. Or more specifically the quality of the information. The knowledge base stores general principles that direct the execution of part of the actions, but the environment provides a learning system with a wide variety of information. If the quality of the information is high, the difference from the general principle is small, and the learning part is easy to process. If the learning system is provided with the disordered specific information for guiding the execution of specific actions, the learning system needs to delete unnecessary details after obtaining enough data, summarize and popularize to form a general principle of guiding actions, and put the general principle into a knowledge base, so that the task of learning part is relatively heavy and the design is relatively difficult. Because the information obtained by a learning system is often incomplete, the reasoning performed by the learning system is not completely reliable, and the rules summarized by the learning system may or may not be correct. This is verified by the execution effect. The correct rule can improve the efficiency of the system and should be reserved; incorrect rules should be modified or deleted from the database. The knowledge base is the second factor that affects the design of the learning system. Knowledge can be represented in a variety of forms, such as feature vectors, first-order logic statements, production rules, semantic networks and frameworks, and so forth. These representations have their own features, and the following 4 aspects are taken into consideration when selecting the representation: the expression ability is strong; the reasoning is easy; the knowledge base is easy to modify; the knowledge representation is easily scalable. One problem that may ultimately be addressed with knowledge bases is that learning systems cannot acquire knowledge from the air without any knowledge at all, and each learning system requires information provided by certain knowledge understanding environments, analyzes comparisons, makes assumptions, checks and modifies these assumptions. Thus, more precisely, the learning system is an extension and improvement of existing knowledge. The execution part is the core of the whole learning system, because the action of the execution part is the action of the learning part aiming for improvement. There are 3 problems associated with the execution part: complexity, feedback, and transparency.
With the rapid development of the graph technology, various graph algorithms and knowledge maps are applied, so that the development of the traditional machine learning field is more advanced. Therefore, the downstream application performance of the atlas is better and better, and the essence of the atlas algorithm is to construct the incidence relation between different subjects and objects. Therefore, when the information is summarized in a cross-document mode, the characteristic of the graph algorithm can be effectively utilized to improve the performance of multi-document information summarization.
According to the invention, by combining the graph algorithm and the machine learning method, the labor cost is reduced, and the insufficient performance of the machine learning multi-document information generation is improved, so that the purpose of effectively summarizing the multi-document information is achieved.
The present invention will be described with reference to specific examples.
Example one
The embodiment provides a document information summarizing method. Referring to fig. 1 to 2, fig. 1 is a flowchart of a document information summarizing method according to an embodiment of the present application; fig. 2 is a multi-document information generation framework diagram according to an embodiment of the present application, and as shown in fig. 1 to 2, the document information summarizing method includes the following steps:
hidden vector value calculation step S1: coding feature calculation is carried out on a source document, after a source document hiding vector value is obtained, a target document hiding vector value is obtained through calculation according to the source document hiding vector value;
output vector calculation step S2: calculating according to the source document hiding vector value and the target document hiding vector value to obtain an output vector value;
and a step S3 of obtaining a document information summary result, wherein the document information summary result is obtained through calculation according to the output vector value.
In an embodiment, the concealment vector value calculation step S1 includes:
source document hidden vector value calculating step S11: after coding feature calculation is carried out on a plurality of source documents through a neural network coder, a source document hidden vector value is obtained;
target document hidden vector value calculating step S12: and according to the source document hiding vector value, carrying out feature calculation on the target document through a graph algorithm, and then obtaining the target document hiding vector value.
In the concrete implementation, in the first step, coding feature calculation is performed on a plurality of source documents to obtain a source document hidden vector value, and the formula is as follows:
Hi＝Encoder(Xi)
wherein Xi represents the ith source document, i.e., X1, X2, X3 in fig. 2, wherein X1, X2, X3, X4 are source files; encoder is used for extracting features from a source document and can be various neural network structures such as transform, BilSTM and the like; hi denotes the source document hidden vector extracted by the neural network encoder, i.e. H1, H2, H3 shown in fig. 2;
secondly, coding feature calculation is carried out on the target document to obtain a hidden vector value of the target document, and the formula is as follows:
Gi＝Graph(Xi+1)
wherein Xi +1 represents target output text extracted from multiple documents, namely X4 in the diagram of FIG. 2, and X4 is target output, namely X4 is to be summarized from three documents, namely X1, X2 and X3; the Graph representation Graph algorithm is used for extracting the feature vector of the target output X4 and can be various algorithm models, such as a GCN Graph convolution network and the like; gi represents the hidden feature vector of the target text extracted by the graph algorithm.
In an embodiment, the output vector calculating step S2 includes obtaining the output vector value through attention module calculation according to the source document hiding vector value and the target document hiding vector value.
In a specific implementation, in the third step, the obtained Hi and Gi are subjected to attention calculation, and the formula is expressed as follows:
Eout＝Attn(Concat(Hi),Gi+1)
wherein Hi represents three hidden vectors of H1, H2 and H3 calculated in the first step, and Concat represents a splicing operation, namely, the three vectors are transversely spliced; gi +1 represents the feature vector of the target text extracted by the graph algorithm in the second step; attn represents an attention calculation process for calculating feature importance degree values of different source documents and target documents, and the main purpose is to calculate the contribution degree of three vectors of H1, H2 and H3 to an output target feature vector; eout is the output vector of Attention calculated in the third step.
In an embodiment, the step S3 of obtaining the summary result of the document information includes obtaining the summary result of the document information through a decoder calculation according to the output vector value.
In the concrete implementation, the fourth step, the Decoder is used to decode and calculate the hidden state vector Eout calculated by the memorability attack, and the output result is obtained, the formula is expressed as follows:
Output＝Decoder(Eout)
wherein, Decoder is used to decode and calculate the value calculated by Attn module, which can be multiple neural network models, such as BilSTM, transform model; output is the result of the resulting cross-document text summary.
Example two
Referring to fig. 3, fig. 3 is a schematic structural diagram of a document information summarizing system according to the present invention. As shown in fig. 3, the document information summarization method of the present invention is applicable to the above document information summarization method, and the document information summarization system includes:
the hidden vector value calculation unit 51: coding feature calculation is carried out on a source document, after a source document hiding vector value is obtained, a target document hiding vector value is obtained through calculation according to the source document hiding vector value;
output vector calculation unit 52: calculating according to the source document hiding vector value and the target document hiding vector value to obtain an output vector value;
and the document information summary result acquisition unit 53 calculates and acquires the document information summary result according to the output vector value.
In an embodiment, the concealment vector value calculation step unit 51:
source document hiding vector value calculation module 511: after coding feature calculation is carried out on a plurality of source documents through a neural network coder, a source document hidden vector value is obtained;
target document hidden vector value calculation module 512: and according to the source document hiding vector value, carrying out feature calculation on the target document through a graph algorithm, and then obtaining the target document hiding vector value.
In an embodiment, the output vector value is obtained through calculation of an attention module according to the source document hiding vector value and the target document hiding vector value.
In an embodiment, the document information summary result is obtained through a decoder calculation according to the output vector value.
EXAMPLE III
Referring to fig. 4, this embodiment discloses a specific implementation of an electronic device. The electronic device may include a processor 81 and a memory 82 storing computer program instructions.
Specifically, the processor 81 may include a Central Processing Unit (CPU), or A Specific Integrated Circuit (ASIC), or may be configured to implement one or more Integrated circuits of the embodiments of the present Application.
Memory 82 may include, among other things, mass storage for data or instructions. By way of example, and not limitation, memory 82 may include a Hard Disk Drive (Hard Disk Drive, abbreviated to HDD), a floppy Disk Drive, a Solid State Drive (SSD), flash memory, an optical Disk, a magneto-optical Disk, tape, or a Universal Serial Bus (USB) Drive or a combination of two or more of these. Memory 82 may include removable or non-removable (or fixed) media, where appropriate. The memory 82 may be internal or external to the anomaly data monitoring device, where appropriate. In a particular embodiment, the memory 82 is a Non-Volatile (Non-Volatile) memory. In particular embodiments, Memory 82 includes Read-Only Memory (ROM) and Random Access Memory (RAM). The ROM may be mask-programmed ROM, Programmable ROM (PROM), Erasable PROM (FPROM), Electrically Erasable PROM (EFPROM), Electrically rewritable ROM (EAROM), or FLASH Memory (FLASH), or a combination of two or more of these, where appropriate. The RAM may be a Static Random-Access Memory (SRAM) or a Dynamic Random-Access Memory (DRAM), where the DRAM may be a Fast Page Mode Dynamic Random-Access Memory (FPMDRAM), an Extended data output Dynamic Random-Access Memory (EDODRAM), a Synchronous Dynamic Random-Access Memory (SDRAM), and the like.
The memory 82 may be used to store or cache various data files for processing and/or communication use, as well as possible computer program instructions executed by the processor 81.
The processor 81 realizes any of the document information summarizing methods in the above-described embodiments by reading and executing computer program instructions stored in the memory 82.
In some of these embodiments, the electronic device may also include a communication interface 83 and a bus 80. As shown in fig. 4, the processor 81, the memory 82, and the communication interface 83 are connected via the bus 80 to complete communication therebetween.
The communication interface 83 is used for implementing communication between modules, devices, units and/or equipment in the embodiment of the present application. The communication port 83 may also be implemented with other components such as: and data communication is carried out among external equipment, image/abnormal data monitoring equipment, a database, external storage, an image/abnormal data monitoring workstation and the like.
The bus 80 includes hardware, software, or both to couple the components of the electronic device to one another. Bus 80 includes, but is not limited to, at least one of the following: data Bus (Data Bus), Address Bus (Address Bus), Control Bus (Control Bus), Expansion Bus (Expansion Bus), and Local Bus (Local Bus). By way of example, and not limitation, Bus 80 may include an Accelerated Graphics Port (AGP) or other Graphics Bus, an Enhanced Industry Standard Architecture (EISA) Bus, a Front-Side Bus (FSB), a Hyper Transport (HT) Interconnect, an ISA (ISA) Bus, an InfiniBand (InfiniBand) Interconnect, a Low Pin Count (LPC) Bus, a memory Bus, a microchannel Architecture (MCA) Bus, a PCI (Peripheral Component Interconnect) Bus, a PCI-Express (PCI-X) Bus, a Serial Advanced Technology Attachment (SATA) Bus, a Video Electronics Bus (audio Electronics Association), abbreviated VLB) bus or other suitable bus or a combination of two or more of these. Bus 80 may include one or more buses, where appropriate. Although specific buses are described and shown in the embodiments of the application, any suitable buses or interconnects are contemplated by the application.
The electronic device may be connected to an anomaly data monitoring system to implement the methods described in connection with fig. 1-2.
The technical features of the embodiments described above may be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the embodiments described above are not described, but should be considered as being within the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
In summary, the invention combines machine learning and graph algorithm, completes summary generation by machine learning, and extracts relationship features among multiple documents by graph algorithm, thereby realizing multiple document information generation based on machine learning, and achieving the purpose of making up for the defect of insufficient generation of the multiple document information by machine learning by graph algorithm. Therefore, the problems of lack of relevance of the document information, high cost and the like in the process of manually summarizing the multi-document information by a machine are solved.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent application shall be subject to the protection scope of the appended claims.