Financial event extraction method and device, electronic equipment and storage medium

[0055] In order to better understand the present application scheme, the technical solutions in the present application embodiment will be apparent from the drawings in the present application, and the described embodiments are clearly described herein. It is an embodiment of the present application, not all of the embodiments. Based on the embodiments in the present application, one of ordinary skill in the art does not have all other embodiments obtained without creative labor, and should belong to the scope of the present application.

[0056] The terms "including" and "have" and "have" and "have" and any variations thereof are intended to cover the contained in the cover. For example, a series of steps or units comprising a series of steps or units are not limited to the listed steps or units, but alternatively include the steps or units not listed, or optionally also include For these processes, other steps or units inherent in the method, product, or equipment. Moreover, the terms "first", "second" and "third" are used to distinguish different objects, rather than being used to describe a particular order.

[0057] In the present application, "Embodiment" means that the specific features, structures, or characteristics described in connection with the embodiments may be included in at least one embodiment of the present application. This phrase is not necessarily a separate or alternative embodiment of the same embodiment in the various locations in the specification, nor is an independent or alternative embodiment of other embodiments. Those skilled in the art are, and the embodiments described herein can be combined with other embodiments.

[0058] The present application provides a financial event extraction method, which can be based on figure 1 See the application environment shown, see figure 1 In this application environment, terminal devices and servers, terminal devices, and server communication are connected, and their connection can be serial port connection, wireless network connection, Bluetooth connection, and network direct connection. Among them, the terminal device can include various devices having input capabilities and communication capabilities, which can be a tablet, handheld computer, laptop, etc., the server can include various devices with program code running capabilities and communication capabilities, which can be independent. The physical server can also be a server cluster or a distributed system, but also providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, and Cloud servers for infrastructure cloud computing services such as big data and artificial intelligence platforms.

[0059] In particular, the terminal device receives the program instruction input by the user through the communication component and to be extracted, and transmitting the program instruction from the communication interface to the server, and performs the extraction operation of the financial event by the server. The server runs the program instruction based on the supplied program, and the server runs the program instruction based on the provided program instruction, and performs a selection of the tangle to the extracted circle, and is quantified by the descent. The model predicts the final vector to obtain a predefined label corresponding to the words in the words to be drawn to be drawn, thereby extracting the corresponding financial events. Due to the introduction of financial knowledge maps during the entire extraction, it is conducive to the extraction of financial events, and the accuracy of financial event extraction is enhanced.

[0060] based on figure 1 The description of the financial event extraction method provided by the present application embodiment will be described in conjunction with other drawings.

[0061] See figure 2 , figure 2 A flow diagram of a financial event extraction method provided in the present application embodiment, which is applied to electronic devices, such as figure 2 As shown, including steps 201-204:

[0062] 201: Treatment of tangle treatment with the Chinese-speaking tools introduced into the financial knowledge map.

[0063] In the specific embodiment of the present application, the method further comprises:

[0064] Add the name of the node of the financial knowledge map to the financial nourse;

[0065] The financial noun table is added to the word table of the preset Chinese chronic tool to obtain a Chinese chronic tool introduced into the financial knowledge map.

[0066] Exemplary, the preset Chinese cordial tool can be a jieba word tool, a financial knowledge map is a vertical application of knowledge maps in the financial sector, which characterizes important entities in the financial sector, such as industries, standards, and countries. Institutions, important practitioners, etc., the edges in the map represent the relationship attributes between various entities. Briefly, the financial knowledge map is a large-scale financial information collection, adding a financial noun table in the existing Chinese word tool, which is conducive to accurately identifying important entities in the financial sector in the pylorque, facilitating Ensure the integrity of the financial sector's proprietary nouns to enhance the quality of the subsequent word vector.

[0067] 202: A first word vector to be extracted in the descertrans-processed words and the second word vector to extract the words in the drawing of the words based on the finance knowledge map.

[0068] In the specific embodiment of the present application, the first word vector refers to the vectorization of each word in the descended corpus after the word treatment, and the method can be obtained, such as the word embedding model, the word embedding model can be BERT (Bidirectional Encoder Reperesentation "model, a converter-based two-way encoding characterization model, Word2Vec (Word To Vector, a tool for converting words into vector forms) model, and so on.

[0069] Exemplary, the above-described financial knowledge map obtains a second word vector of the word to extract the words after the finishes process, including:

[0070] With the target node in the financial knowledge map, a node directly connected to the target node and the side of the connection of the two nodes;

[0071] The triplet is performed to quantify the quantization process to obtain the fourth word vector of the entity in the desired corpus to extract the particle;

[0072] The word mapping the node of the finishes of the finishes of the desired pattern is mapped to zero vector;

[0073] The fourth word vector and the zero vector are determined to be the second word vector.

[0074] For example, it is assumed that the tangle to be drawn is "Zhang XX releases a strong down-cut signal" in the Congress testimony, "Zhang XX" is the entity in the tangurated corpus. If there is a node that is directly connected to it in the financial knowledge map "Manager", connecting the edges between the two nodes indicating the relationship between the two entities, such as "A Company", by "Zhang XX, A Company, Manager" can build a three-way group, among them, "Zhang XX "indicates the head entity," A company "expressed the relationship," manager "means tail entities, such as image 3 As shown, the TRANSE algorithm can be used to quantify the quantization process, resulting in vectorization of the entity and relationship, thereby obtaining the fourth word vector of the entity in the patch to be drawn after the word processing, such as "XX" The word vector, and the words that are not appearing in the financial knowledge map in the finishes of the word processing, the word vector and zero vector, all the fourth word vector and zero vector. In this embodiment, the knowledge in the financial knowledge map is lacking in the form of financial concepts or entity alias to be extracted, such as "Zhang XX" is the hidden information of the Company A, and the corresponding event can also be taken. Effective enhancement of the recall of the event extraction model, and then enhance the accuracy of the event extraction.

[0075] 203: Specifies the first word vector and the second word vector to obtain a third word vector of the word to be drawn in the desired corpus.

[0076] In the specific embodiment of the present application, please continue to see image 3 The first word vector is V_EMBED, the second word vector is V_Graph, and the third word vector of each word in the tangle after the word processing is denoted as V. t = [V_EMBED, V_GRAPH], the third word vector V t As the input of the sequence labeling, or will be based on the third word vector V t The new vector obtained is input as a sequence labeling model.

[0077] 204: The predefined tag corresponding to the word to extract in the drawing of the words based on the third word vector is predicted, and the predetermined tag corresponding to the words is performed according to the predefined tag corresponding to the word.

[0078] In the specific embodiment of the present application, since the content of the financial event can be generally divided into two categories: a class as a policy of government or professional institutional policy, the other is a description of the variation of the specific industries or major asset value trend. User truly interested information often only involves a clear event body and event predicate, so it is necessary to extract this information and exclude the contents of the decorative class in the text.

[0079] This application has been further refined under the labeling system labeling the existing conventional sequence, and the predefined label system is:

[0080] SB: Subject Beginning, indicating the beginning of the event body description.

[0081] Si: Subject Intermediate, indicating the main contents of the event entity description.

[0082] SE: Subject End, indicating the end of the event main body description.

[0083] PB: predicate beginning, indicating the beginning of the event predicate.

[0084] PI: PRedicate Intermediate, indicating the main content of the event predicate.

[0085] PE: PREDICATE END, indicating the end of the event predicate description.

[0086] O: Others, indicating that the characters are not related.

[0087]Based on the predefined tag system, the sequence label model is used to classify the third word vector, and the sequence annotation model can be the LSTM (Long Short-Term Memory, long short-term memory network) model, or BI-LSTM ( Bidirectional LSTM NetWorks, two-way long short-term memory network) model, or Bi-LSTM-CRF (Bidirectional LSTM NetWorks-Conditional Random Field, two-way long short-term memory network - condition random) model, and so on. Compared to the regular input of the sequence labeling the model, the second word vector V_Graph spliced ​​after the first word vector V_EMBED, the dimension of the input vector increases, but the process of the sequence labeling model is consistent with the original, the last one The label of the layer output is a predefined label, such as image 3 In: Y1, Y2, ... YN-1, YN, where n represents the number of words, specifically as follows:

[0088] Zhang XX releases a strong interest rate cut signal in the congressional testimony

[0089] SB Si SE O O O O Pb Pi Pi Pi Pi Pi PE

[0090] Since this application is extracted only the event body and event predicate, the above example is based on the predicted predefined label, the final extracted financial event is "Zhang XX release strong down interest signal".

[0091] Exemplary, in the word menu of the preset Chinese chronic tool, the method is also included in the Chinese word tool introduced into the financial knowledge map, the method further includes:

[0092] Taking samples, the sample text is scavemented by the Chinese chronic tool introduced into the financial knowledge map;

[0093] A fifth word vector of the word in the sample text after the word processing is obtained and the sixth word vector of the words in the sample text after the finance is obtained by the financial knowledge map;

[0094] The fifth word vector and the sixth word vector are spliced ​​to obtain a seventh word vector of the word text after the word treatment;

[0095] The seventh word vector input sequence labeling model is trained to obtain a predefined label corresponding to the word words in the sample text after word treatment;

[0096] The target loss is determined based on the predefined tag corresponding to the word corresponding to the word in the sample text after the word treatment;

[0097] Adjust the parameters of the sequence labeling model to minimize the target loss and obtain a well-trained sequence label model.

[0098] In the specific embodiment of the present application, the fifth word vector refers to the word vector to quantify the words in the sample text after word treatment, and the sixth word vector refers to the use of the TRANSE algorithm based on the financial knowledge map. The word vector obtained by quantization, the processing of the training stage can be referred to in the above steps 201-204, and details are not described herein again. The target loss can be the largest likelihood loss, and its formula is expressed as follows:

[0099]

[0100] Wherein, L (θ) indicates the value of the target loss, n represents the number of samples, k represents the kar of the n-sample text, T represents the length of the sample, P θ Probability distribution of predefined tags indicating the sequence label model output, y j Pre-defined label representing the word on the position J, X 1:j The splicing word of the words in all positions before position J and position J, θ represent the parameters of the sequence labeling model.

[0101] In this embodiment, the introduction of the financial knowledge map makes the training of the sequence labeling model no longer requires a large amount of labeling data, breaking the bottleneck completely trained by large data driving models, realizing the training model of data and knowledge dual drive, which is conducive to Reduce training overhead.

[0102] Exemplary, the above-mentioned third word vector predicts a predefined tag corresponding to the word to extract the words in the desired corpus, including:

[0103] The weight of each word vector in the third word vector is calculated using a preset focal mechanism;

[0104] The attention vector of the word to be extracted in the words after each word is calculated by the weight of each word.

[0105] A predefined label corresponding to the words to be drawn in the descending tangle is predicted based on the attention vector.

[0106] In the specific embodiment of the present application, after obtaining a third word vector, the preset focal mechanism is used to capture the subject information to extract the values ​​to the importance of learning words, to allocate different weights in the words to be drawn. , Compared to the traditional attention mechanism depends on the hidden vector in SEQ2SEQ (Sequence-To-Sequence, a Natural Language) Decoder, the preset focal mechanism is an improved attention mechanism. The improvement point is through the subject marker S t Mark the third word vector to get the weight of each word vector Among them, t represents the Time.

[0107] in,

[0108]

[0109] in, W (act) , B (act) All the parameters learned during the training stage minimized by the loss function, S t Represents the subject marker S t Marked hidden vector, It is an operational symbol for connecting two vectors, that is, the preset focal mechanism is dependent on the subject marker S. t The marked hidden vector. will and Multiplication Get the payment vectors corresponding to the first third word, the extraction of financial events is used to make the sequence labeling model more attention to the subject information of the tangua to be extracted, and further extract it to useful information.

[0110] Exemplary, the above-described predetermined predefined tag, which is predicts the word to extract the words in the desired corpus according to the attention vectors, including:

[0111] The attention vector is encoded to obtain the vector to be classified;

[0112] The sequence of sequences to be classified by the classified vector input to the classification is predicted to obtain a predetermined prediction of the word corresponding to the words to extract the words in the desired language.

[0113] In the specific embodiment of the present application, the attention vector output for the preset focal mechanism is again used, and the subject marker S is employed. t Code it to get the vector S ' t , To be classified vector S ' t Enter a well-trained sequence label model for classification forecast, eventually output a predefined label to make the topic information to extract the topic information easier to be captured by the sequence labeling model.

[0114] It can be seen that the present application example is tethected by the extracted corpus by using the Chinese chronic tool to introduce the financial knowledge map, and the first word vector of the word to be drawn after the word treatment is obtained and the financing method is obtained after obtaining the finishes based on financial knowledge map. The second word vector of the word in the word is drawn, the first word vector and the second word vector are spliced, and the third word vector of the words to be drawn in the tangle is obtained, based on the third word vector, the word is predicted The predefined label corresponding to the words in the tangle to be extracted, performing the extraction of financial events based on the predefined tag corresponding to the word. Since the information introduced in the financial knowledge map is introduced during the financial event extraction, the Chinese-speaking instrument introduced into the financial knowledge map is conducive to ensuring the integrity of important professional vocabulary when performing the designs of the descent, and ensures the first word vector. The quality of the second word vector is higher, then, the predetermined prediction of the predefined label is made with a higher quality first word vector and the second word vector. Accuracy of financial events. In addition, the predefined label is refined to the original label system, which only focuses on event main body and event predicate. Under the premise of retaining the semantics of the event, it is greatly improved the financial event description. Quantitative, which is conducive to reducing the cost of further processing of the extracted financial events.

[0115] See Figure 4 , Figure 4 A flow chart of another financial event extraction method provided in the present application embodiment is equally based on figure 1 The application environment shown, such as Figure 4 As shown, including steps 401-408:

[0116] 401: Treatment of tanguo treatment with Chinese chronic tools in introducing financial knowledge maps;

[0117] 402: The first word vector of the word to be extracted in the tangle is obtained;

[0118] 403: With the target node in the financial knowledge map, the node directly connected to the target node and the side of the two nodes are constructed, and the target node is a representation of the financing. The node of the entity to which the entity to extract the corpus;

[0119] 404: Treume the three-component group to quantify, to obtain the fourth word vector of the entity in which the entity to be drawn in the tangle is obtained;

[0120] 405: Map the word that is not partially extracted in the descending the tie to extract into zero vector;

[0121] 406: Determine the fourth word vector and the zero vector to a second word vector;

[0122] 407: Specifies the first word vector and the second word vector to obtain a third word vector of the words to be drawn in the desired corpus;

[0123] 408: A predefined label corresponding to the word to extract in the drawing of the words after the third word vector is predicted, and the extraction of the financial event is performed according to a predefined tag corresponding to the word.

[0124] Among them, the specific embodiments of steps 401-408 are figure 2 The description thereof has been described in the examples and can achieve the same or similar beneficial effects, in order to avoid repetition, will not be described again.

[0125] For the description of the above financial event extraction method, please see Figure 5 , Figure 5 A structural diagram of a financial event extraction device provided in the present application embodiment, such as Figure 5 As shown, the apparatus includes:

[0126] The word module 501 is used to treat the tandem tool to be drawn by the introduction of a financial knowledge map.

[0127]The encoding module 502 is used to obtain a first word vector of the word to extract the words after the word processing, and the second word vector of the word to extract the words after the finance knowledge spectrum is obtained by the financial knowledge map;

[0128] The splicing module 503 is configured to splicing the first word vector and the second word vector to obtain a third word vector of the word to extract the words after the word treatment;

[0129] Processing module 504 for predicting a predefined label corresponding to the word to extract the words in the drawing of the word after the third word vector, according to the predefined tag corresponding to the predefined tag.

[0130] In a possible embodiment, the coding module 502 is specifically used in: the second word vector of the word to extract the words after the financing of the finishes is obtained based on the financial knowledge map.

[0131] With the target node in the financial knowledge map, a node directly connected to the target node and the side of the two nodes are constructed, and the target node is a representation of the finishes of the financial knowledge map. The node of the entity to be extracted in the tang;

[0132] The triplet is performed to quantify the quantization process to obtain the fourth word vector of the entity in the desired corpus to extract the particle;

[0133] The word mapping the node of the finishes of the finishes of the desired pattern is mapped to zero vector;

[0134] The fourth word vector and the zero vector are determined to be the second word vector.

[0135] In one possible embodiment, the word module 501 is also used in:

[0136] Add the name of the node of the financial knowledge map to the financial nourse;

[0137] The financial noun table is added to the word table of the preset Chinese chronic tool to obtain a Chinese chronic tool introduced into the financial knowledge map.

[0138] In a possible embodiment, the processing module 504 is specifically used for: Treatment Module 504 for pre-defined tags in the predetermined tag, predicting the words of the words after the word treatment, predicting the words.

[0139] The weight of each word vector in the third word vector is calculated using a preset focal mechanism;

[0140] The attention vector of the word to be extracted in the words after each word is calculated by the weight of each word.

[0141] A predefined label corresponding to the words to be drawn in the descending tangle is predicted based on the attention vector.

[0142] In a possible embodiment, the processing module 504 is specifically used in: Treatment module 504 for predetermined predetermined tags in accordance with the attention vectors predicting the words of the words.

[0143] The attention vector is encoded to obtain the vector to be classified;

[0144] The sequence of sequences to be classified by the classified vector input to the classification is predicted to obtain a predetermined prediction of the word corresponding to the words to extract the words in the desired language.

[0145] In a possible embodiment, if Image 6 As shown, the apparatus also includes a training module 505 for:

[0146] Taking samples, the sample text is scavemented by the Chinese chronic tool introduced into the financial knowledge map;

[0147] A fifth word vector of the word in the sample text after the word processing is obtained and the sixth word vector of the words in the sample text after the finance is obtained by the financial knowledge map;

[0148] The fifth word vector and the sixth word vector are spliced ​​to obtain a seventh word vector of the word text after the word treatment;

[0149] The seventh word vector input sequence labeling model is trained to obtain a predefined label corresponding to the word words in the sample text after word treatment;

[0150] The target loss is determined based on the predefined tag corresponding to the word corresponding to the word in the sample text after the word treatment;

[0151] Adjust the parameters of the sequence labeling model to minimize the target loss and obtain a well-trained sequence label model.

[0152] According to an embodiment of the present application, Figure 5 or Image 6 Each unit of the financial event extraction device may be composed of or all of them into one or several additional units, or some (some) units can also be removed into a plurality of units that are functionally smaller. Configuration, this can achieve the same operation without affecting the implementation of the technical effects of the embodiments of the present application. The above unit is based on logic functionality, in actual applications, the functionality of one unit can also be implemented by a plurality of units, or the functionality of the plurality of units is implemented by one unit. In other embodiments of the present application, other units may also include other units based on financial event extraction devices, which can also be implemented by other units, and can be implemented by multiple units.

[0153] According to another embodiment of the present application, the general-purpose computing device, such as a computer such as a computer such as a computer such as a memory element, such as a memory element, including a central processing unit (CPU), random access storage medium (RAM), a read-only storage medium (ROM). Run can perform figure 2 or Figure 4 Computer programs (including program code) of the respective steps described in the corresponding methods shown in the respective methods are constructed Figure 5 or Image 6 Financial event extraction device, and the financial event extraction method of the present application embodiment is implemented. The computer program may be described, for example, on the computer readable recording medium, and is mounted in the computing device through a computer readable recording medium, and operate therein.

[0154] Based on the above method embodiments and the embodiment of the apparatus embodiment, the present application also provides an electronic device. See Figure 7 The electronic device includes at least a processor 701, an input device 702, an output device 703, and a computer storage medium 704. The processor 701, input device 702, output device 703, and computer storage medium 704 can be connected via a bus or otherwise connected.

[0155] The computer storage medium 704 can be stored in a memory of the electronic device, the computer storage medium 704 for storing computer programs, the computer program including program instructions, the processor 701 for performing the computer storage medium 704 stored programs instruction. Processor 701 (or CPU (CPU (Central Processing Unit, Central Processor)) is the calculation core of the electronic device and the control core, which adapted to implement one or more instructions, specifically suitable for loading and executing one or more instructions The corresponding method flow or corresponding functions.

[0156] In one embodiment, the processor 701 of the electronic device provided herein can be used to perform a series of financial event extraction processing:

[0157] Treatment of tether to the extracted corpus by introducing financial knowledge maps;

[0158] The first word vector of the word to be drawn in the drawing of the word processing is obtained and the second word vector of the word to extract the word to extract the words after the finance knowledge map is obtained;

[0159] The first word vector and the second word vector are spliced ​​to obtain a third word vector of the word to be drawn in the tangurated corpus;

[0160] Based on the third word vector predicts a predefined tag corresponding to the word to extract the words in the drawing of the word process, the extraction of the financial event is performed according to a predefined tag corresponding to the word.

[0161] In still another embodiment, the processor 701 performs the second word vector of the word to extract the words based on the financial knowledge map to acquire the word processing, including:

[0162] With the target node in the financial knowledge map, a node directly connected to the target node and the side of the two nodes are constructed, and the target node is a representation of the finishes of the financial knowledge map. The node of the entity to be extracted in the tang;

[0163] The triplet is performed to quantify the quantization process to obtain the fourth word vector of the entity in the desired corpus to extract the particle;

[0164] The word mapping the node of the finishes of the finishes of the desired pattern is mapped to zero vector;

[0165] The fourth word vector and the zero vector are determined to be the second word vector.

[0166] In still another embodiment, the processor 701 is also used for:

[0167] Add the name of the node of the financial knowledge map to the financial nourse;

[0168] The financial noun table is added to the word table of the preset Chinese chronic tool to obtain a Chinese chronic tool introduced into the financial knowledge map.

[0169] In still another embodiment, the processor 701 performs the predefined label corresponding to the word to extract the words based on the third word vector predict the words after the word processing, including:

[0170] The weight of each word vector in the third word vector is calculated using a preset focal mechanism;

[0171] The attention vector of the word to be extracted in the words after each word is calculated by the weight of each word.

[0172] A predefined label corresponding to the words to be drawn in the descending tangle is predicted based on the attention vector.

[0173] In still another embodiment, the processor 701 performs the predefined label corresponding to the word to extract the words according to the attention vectors predict the word processing, including:

[0174] The attention vector is encoded to obtain the vector to be classified;

[0175] The sequence of sequences to be classified by the classified vector input to the classification is predicted to obtain a predetermined prediction of the word corresponding to the words to extract the words in the desired language.

[0176] In still another embodiment, after adding the financial noun table to the word table of the predetermined Chinese chronic tool, the processor 701 is also used in:

[0177] Taking samples, the sample text is scavemented by the Chinese chronic tool introduced into the financial knowledge map;

[0178] A fifth word vector of the word in the sample text after the word processing is obtained and the sixth word vector of the words in the sample text after the finance is obtained by the financial knowledge map;

[0179]The fifth word vector and the sixth word vector are spliced ​​to obtain a seventh word vector of the word text after the word treatment;

[0180] The seventh word vector input sequence labeling model is trained to obtain a predefined label corresponding to the word words in the sample text after word treatment;

[0181] The target loss is determined based on the predefined tag corresponding to the word corresponding to the word in the sample text after the word treatment;

[0182] Adjust the parameters of the sequence labeling model to minimize the target loss and obtain a well-trained sequence label model.

[0183] Exemplary, electronic devices include, but not limited to, processor 701, input device 702, output device 703, and computer storage medium 704. It is also possible to include a memory, power supply, application client module, and the like. The input device 702 can be a keyboard, a touch screen, a radio frequency receiver, or the like, and the output device 703 can be a speaker, a display, a radio transmitter, and the like. Those skilled in the art will appreciate that the schematic diagram is merely an example of an electronic device, and does not constitute a defined to the electronic device, which may include more or less components, or a combination of parts, or different components.

[0184] In addition, since the processor 701 of the electronic device implements the step of the above financial event extraction method, the embodiment of the above financial event extraction method is applied to the electronic device, and can achieve the same or similar The beneficial effect.

[0185] The present application embodiment also provides a computer storage medium (Memory), the computer storage medium, is a memory device in an electronic device for storing programs and data. It will be appreciated that the computer storage medium here may include both a built-in storage medium in the terminal, of course, can also include an extended storage medium supported by the terminal. The computer storage medium provides storage space that stores the operating system of the terminal. Further, in this storage space, one or more instructions suitable for being loaded and executed by the processor 701 are also stored, which may be one or more computer programs (including program code). It should be noted that the computer storage medium here may be a high speed RAM memory, or may be a non-Volatile Memory, such as at least one disk memory; optionally, may be at least one located away from the aforementioned processing The computer storage medium of 701. In one embodiment, one or more instructions stored in the computer storage medium may be loaded and executed by the processor 701 to achieve the corresponding steps of the financial event extraction method.

[0186] Exemplary, computer programs for computer storage media include computer program code, which can be source code form, object code form, executable file, or some intermediate form. The computer readable medium can include any entity or device capable of carrying the computer program code, a recording medium, a U disk, a mobile hard disk, a disk, an optical disk, a computer memory, a read-only memory (ROM, Read-Only Memory) , Random access memory (RAM, RANDOM Access Memory), electrical carrier signal, telecommunications signal, and software distribution media.

[0187] In addition, since the computer program of the computer storage medium is executed by the processor, all embodiments of the above financial event extraction method are suitable for the computer storage medium, and can be achieved. The same or similar benefits.

[0188] The present application examples have been described in detail herein, and specific examples are used herein to explain the principles and embodiments of the present application, and the above embodiments are intended to help understand the method of this application and their core ideas; In the art, according to the thoughts of the present invention, there will be changes in the specific embodiments and applications, in summary, the contents of this specification should not be construed as limiting the present application.