Energy policy monitoring and analysis method based on natural language processing

Abstract

The present invention relates to the technical field of energy policy monitoring, and in particular to an energy policy monitoring and analysis method based on natural language processing. The method obtains policy vocabulary and industrial node vocabulary, divides them to obtain vocabulary categories; obtains the node influence degree according to the corresponding edges of policy vocabulary and industrial node vocabulary in the vocabulary category; obtains the adjustment vocabulary category, takes the industrial node vocabulary in the adjustment vocabulary category as the adjustment vocabulary, obtains the adjustment influence degree based on the preset closed loop and node influence degree of the adjustment vocabulary, and selects the focus adjustment vocabulary based on the focus adjustment vocabulary, and obtains the concentrated adjustment degree based on the focus adjustment vocabulary; obtains the energy policy change evaluation text based on the short sentence, adjustment influence degree and concentrated adjustment degree of the policy vocabulary in the adjustment vocabulary category. The present invention accurately obtains the energy policy change evaluation text by comprehensively considering the impact of the energy policy revision part on the energy industry, so that enterprises can adjust the energy industry in a timely and accurate manner.

Description

Technical Field

[0001] The present invention relates to the technical field of energy policy monitoring, and in particular to an energy policy monitoring and analysis method based on natural language processing. Background Art

[0002] Energy policy monitoring and analysis refers to mining energy policy texts to analyze the mutual impact of energy policy revisions on different areas of the energy industry. Since technicians at different levels have strong subjective intentions in evaluating energy policy texts, it is difficult to synthesize the evaluation conclusions of energy policies. In the energy industry, various industrial nodes are interconnected to form a complete industrial chain. When the released energy policy adjusts some of the industrial nodes, it may have a hidden impact on the entire industrial chain. In order to accurately and comprehensively analyze the impact of energy policy adjustments on the energy industry chain, and then accurately monitor energy policies.

[0003] In the existing methods, the energy policy text is directly analyzed and mined to obtain the energy policy change evaluation text. The enterprise uses the energy policy change evaluation text as an auxiliary, combined with other aspects such as policy interpretation and analysis by relevant professionals, to conduct a comprehensive analysis of energy policy information to determine the overall impact of energy policy on the energy industry. However, in actual situations, the impact of the revised energy policy on some industrial nodes in the industrial chain is hidden, making the obtained energy policy change evaluation text inaccurate, which in turn makes it impossible for enterprises to accurately analyze the role and impact of energy policy revisions on the energy industry. Summary of the invention

[0004] In order to solve the technical problem that the impact of the revised part of the energy policy on some industrial nodes in the industrial chain is hidden, resulting in inaccurate energy policy change evaluation text, the purpose of the present invention is to provide an energy policy monitoring and analysis method based on natural language processing, and the technical solution adopted is as follows:

[0005] An embodiment of the present invention provides an energy policy monitoring and analysis method based on natural language processing, the method comprising the following steps:

[0006] Obtain policy vocabulary from energy policy texts and industry node vocabulary from energy industry structure diagrams;

[0007] Based on the vocabulary vectors corresponding to policy vocabulary and industrial node vocabulary, the policy vocabulary and industrial node vocabulary are uniformly divided to obtain multiple vocabulary categories;

[0008] According to the distribution of policy words in each vocabulary category and the distribution of industrial node words on the edges of the energy industry structure diagram, the node influence degree of each vocabulary category is obtained;

[0009] According to the policy vocabulary of the revised part in the energy policy text, the adjustment vocabulary category is screened out in the vocabulary category; the industrial node vocabulary in the adjustment vocabulary category is used as the adjustment vocabulary, and the adjustment influence degree of each adjustment vocabulary is obtained according to the number of industrial node vocabulary in the preset closed loop in the energy industry structure diagram and the node influence degree of each adjustment vocabulary category;

[0010] Based on the degree of adjustment impact, the words that are focused on adjustment are selected, and based on the degree of adjustment impact and quantity of the words that are focused on adjustment, the degree of concentrated adjustment of the energy policy text is obtained;

[0011] The short sentences in the energy policy text containing policy words in the adjustment vocabulary category, the adjustment influence degree and the concentrated adjustment degree of the focus adjustment words are input into the GPT model, and the energy policy change evaluation text is obtained through the GPT model.

[0012] Furthermore, the method for obtaining the node influence degree is:

[0013] For any vocabulary category, obtaining the number of edges of each industrial node vocabulary in the vocabulary category in the energy industry structure graph as the first number of the corresponding industrial node vocabulary;

[0014] The sum of all the first quantities is taken as the number of related edges within the vocabulary category;

[0015] The node influence degree of the vocabulary category is obtained according to the normalized result of the number of policy words in the vocabulary category and the normalized result of the number of related edges.

[0016] Furthermore, the method for obtaining the adjusted vocabulary category is:

[0017] The vocabulary categories of the policy vocabulary in the revised part of the energy policy text shall be regarded as the adjustment vocabulary categories.

[0018] Furthermore, the method for obtaining the adjustment impact degree is:

[0019] For any adjustment vocabulary, the product of the number of industrial node vocabulary in the preset closed loop where the adjustment vocabulary is located in the energy industry structure diagram and the node influence degree value of the adjustment vocabulary category where the adjustment vocabulary is located is obtained as the adjustment influence degree value of the adjustment vocabulary.

[0020] Furthermore, the method for acquiring the focus adjustment vocabulary is:

[0021] Arrange the adjustment words in order from low to high according to the adjustment influence degree to obtain the adjustment word sequence;

[0022] Obtaining the difference in adjustment influence between any two adjacent adjustment words in the adjustment word sequence, all of which are taken as the first difference;

[0023] The adjustment vocabulary with the highest adjustment influence among the two adjustment vocabulary corresponding to the largest first difference is used as the target adjustment vocabulary;

[0024] All adjustment words between the target adjustment word and the last adjustment word in the adjustment word sequence from left to right are regarded as emphasis adjustment words; wherein the emphasis adjustment words include the target adjustment word and the last adjustment word.

[0025] Furthermore, the method for obtaining the concentration adjustment degree is:

[0026] The ratio of the number of adjusted words to the number of focused adjusted words is used as the reference value in the first set;

[0027] The result of normalizing the largest first difference is used as the reference value in the second set;

[0028] According to the first centralized reference value and the second centralized reference value, the centralized adjustment degree of the energy policy text is obtained; wherein, the first centralized reference value and the second centralized reference value are both positively correlated with the centralized adjustment degree.

[0029] Furthermore, the method for obtaining the short sentence is:

[0030] The energy policy text is traversed from front to back, and each time a Chinese punctuation mark or period is detected, it is divided, and each divided text is regarded as a short sentence.

[0031] Furthermore, the method for obtaining the energy policy change evaluation text is:

[0032] The adjustment influence value of the focused adjustment vocabulary is used as the local weight in the GPT model;

[0033] The centralized adjustment degree value is used as the global weight in the GPT model;

[0034] The short sentences, local weights and global weights of the policy words in the adjustment vocabulary category in the energy policy text are input into the GPT model, and the energy policy change evaluation text is output through the GPT model.

[0035] Furthermore, the method for obtaining the vocabulary category is:

[0036] All policy words and industry node words are unified into a word set, and the word vector of each word in the word set is obtained by using the word2vec algorithm;

[0037] All vocabulary vectors are converted into vector space, and the corresponding vocabulary vectors in the vector space are clustered by an iterative self-organizing clustering algorithm to obtain multiple vocabulary vector clusters;

[0038] For any vocabulary vector cluster, the words corresponding to all the vocabulary vectors in the vocabulary vector cluster are constructed as a vocabulary category.

[0039] Furthermore, the method for obtaining the policy vocabulary and industry node vocabulary is:

[0040] Set the jieba segmentation mode to the precise mode, segment the energy policy text, and obtain the individual segmentations of the energy policy text as policy vocabulary;

[0041] Each industry node name in the energy industry structure diagram is directly used as the industry node vocabulary.

[0042] The present invention has the following beneficial effects:

[0043] Based on the vocabulary vectors corresponding to policy vocabulary and industrial node vocabulary, the policy vocabulary and industrial node vocabulary are uniformly divided to obtain multiple vocabulary categories, and the policy vocabulary and industrial node vocabulary are matched and integrated to determine the industrial node vocabulary to be adjusted by the policy vocabulary, which is conducive to the subsequent accurate analysis of the impact of energy policies on industrial nodes; in order to analyze the degree of influence of policy vocabulary in each vocabulary category on industrial node vocabulary, the node influence degree of each vocabulary category is obtained according to the distribution of policy vocabulary in each vocabulary category and the distribution of industrial node vocabulary on the edges of the energy industry structure diagram, which accurately reflects the impact of energy policies on the industrial nodes corresponding to the industrial node vocabulary in each vocabulary category; in order to efficiently analyze the impact of energy policies on the energy industry, according to the policy vocabulary of the revised part in the energy policy text, the adjustment vocabulary category is screened out in the vocabulary category to accurately determine the industrial nodes that need to be adjusted in the energy industry. Therefore, the industrial node vocabulary in the adjustment vocabulary category is used as the adjustment vocabulary to accurately determine the industrial nodes that need to be adjusted; in order to accurately obtain The impact of the revised energy policy on the energy industry, and then according to the number of industrial node words in the preset closed loop of each adjustment word in the energy industry structure diagram and the node influence degree of the adjustment word category in which each adjustment word is located, the adjustment influence degree of each adjustment word is obtained, accurately reflecting the impact of each adjustment word when it is adjusted; in order to analyze the concentration of the revised energy policy, so as to more accurately analyze the impact of the revised energy policy on the energy industry in the future, and then based on the adjustment influence degree, the focus adjustment words are screened out, and based on the adjustment influence degree and number of the focus adjustment words, the concentration adjustment degree of the energy policy text is obtained, accurately reflecting the concentration degree of the energy policy revision on the energy industry; and then according to the short sentences in the energy policy text where the policy words in the adjustment word category are located, the adjustment influence degree and the concentration adjustment degree of the focus adjustment words, accurately and comprehensively obtain the energy policy change evaluation text, provide enterprises with a comprehensive impact of the revised energy policy, which is conducive to enterprises to make timely and accurate adjustments to the corresponding links in the energy industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0045] Figure 1 A schematic flow chart of an energy policy monitoring and analysis method based on natural language processing provided by one embodiment of the present invention;

[0046] Figure 2 An energy industry structure diagram provided by an embodiment of the present invention;

[0047] Figure 3 A flow chart of a method for obtaining a centralized adjustment degree provided by an embodiment of the present invention;

[0048] Figure 4 A structural diagram of an energy policy monitoring and analysis system based on natural language processing provided by one embodiment of the present invention;

[0049] Figure 5 A schematic diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0050] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the energy policy monitoring and analysis method based on natural language processing proposed by the present invention, its specific implementation method, structure, characteristics and effects are described in detail as follows in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0051] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0052] The specific scheme of the energy policy monitoring and analysis method based on natural language processing provided by the present invention is described in detail below with reference to the accompanying drawings.

[0053] Embodiment 1:

[0054] The specific scenario of this embodiment is: this embodiment only analyzes the text content of the energy policy that can be evaluated. At the same time, the energy policy in this embodiment is the revised energy policy. By analyzing the energy policy text, the impact of the revised energy policy on the energy industry is evaluated.

[0055] This paper proposes an energy policy monitoring and analysis method based on natural language processing, please refer to Figure 1 , which shows a schematic flow chart of an energy policy monitoring and analysis method based on natural language processing provided by an embodiment of the present invention, the method comprising the following steps:

[0056] Step S1: Obtain policy vocabulary from energy policy text and industry node vocabulary from energy industry structure diagram.

[0057] Specifically, the newly released energy policy text is obtained from the public policy release webpage, the jieba segmentation mode is set to the precise mode, the energy policy text is segmented, and each segmentation of the energy policy text is obtained as the policy vocabulary. Among them, the jieba segmentation is a public technology and will not be described in detail. The known energy industry chain is public knowledge, such as Figure 2 Shown is the energy industry structure diagram, which accurately reflects the relationship between different industry nodes.

[0058] The essence of energy policy revision is to adjust the industrial nodes in the energy industry. By adjusting the industrial nodes in the industrial chain, the energy industry can develop healthily and stably. Therefore, the energy policy text must contain words that describe the industrial nodes, so as to achieve an accurate and directional description of the adjusted industrial nodes. In order to subsequently determine the industrial nodes that are directly adjusted by the energy policy text, this embodiment directly uses each industrial node name in the energy industry structure diagram as an industrial node vocabulary, that is, one industrial node corresponds to one industrial node vocabulary. At this point, all policy vocabulary and industrial node vocabulary are obtained.

[0059] Step S2: Based on the vocabulary vectors corresponding to the policy vocabulary and the industry node vocabulary, the policy vocabulary and the industry node vocabulary are uniformly divided to obtain multiple vocabulary categories.

[0060] Specifically, in order to analyze the impact of energy policies on industrial nodes, and then obtain the semantic similarity between policy vocabulary and industrial node vocabulary in the energy policy text, determine the industrial nodes that need to be adjusted that are intuitively described in the energy policy. Therefore, this embodiment first obtains the vocabulary vectors of policy vocabulary and industrial node vocabulary, and then based on the vocabulary vectors corresponding to the policy vocabulary and industrial node vocabulary, uniformly divides the policy vocabulary and industrial node vocabulary to obtain multiple vocabulary categories, wherein a vocabulary category may only contain policy vocabulary or industrial node vocabulary, or may contain both policy vocabulary and industrial node vocabulary. It should be noted that the policy vocabulary in a certain vocabulary category can be defaulted to the vocabulary that directly describes how to adjust the industrial node vocabulary in the vocabulary category, therefore, the degree of correlation between the policy vocabulary and industrial node vocabulary in the same vocabulary category is the highest.

[0061] Preferably, in some possible implementations of the present embodiment, the method for obtaining vocabulary categories is as follows: first, all policy vocabulary and industry node vocabulary are unified into a vocabulary set, which is conducive to accurately analyzing the semantic similarity between policy vocabulary and industry node vocabulary. Then, the vocabulary vector of each vocabulary in the vocabulary set is obtained through the word2vec algorithm, and all vocabulary vectors are converted into the vector space to more accurately analyze the degree of correlation between vocabulary vectors, and then the corresponding vocabulary vectors in the vector space are clustered through the iterative self-organizing clustering algorithm to obtain multiple vocabulary vector clusters. Among them, the word2vec algorithm and the iterative self-organizing clustering algorithm are both well-known technologies and will not be described in detail. Finally, for any vocabulary vector cluster cluster, the vocabulary corresponding to all vocabulary vectors in the vocabulary vector cluster cluster is constructed as a vocabulary category.

[0062] At this point, all vocabulary categories are obtained, that is, policy vocabulary and industry node vocabulary are matched and integrated, which facilitates the subsequent analysis of specific industry nodes affected by the revision of energy policies.

[0063] Step S3: According to the distribution of policy words in each vocabulary category and the distribution of industrial node words on the edges of the energy industry structure diagram, the node influence degree of each vocabulary category is obtained.

[0064] Specifically, in actual situations, energy policies are diverse, that is, one energy policy can directly adjust multiple industrial nodes. In order to obtain the impact of energy policies on industrial nodes, this embodiment analyzes each vocabulary category. The more policy words in a vocabulary category, the more accurate the adjustment intention of the industrial node corresponding to the industrial node words in the vocabulary category, and the greater the possibility of adjustment of the industrial node corresponding to the industrial node words in the vocabulary category. Figure 2 In the energy industry structure graph, there are edges, for example, Figure 2 In the energy industry structure diagram, new energy vehicle production has 3 edges. When the industrial node vocabulary in this vocabulary category has more edges, it means that when the industrial node corresponding to the industrial node vocabulary in this vocabulary category is adjusted, the degree of influence on the overall industrial chain of the energy industry is greater. Therefore, this embodiment obtains the node influence degree of each vocabulary category according to the distribution of policy vocabulary in each vocabulary category and the distribution of edges of industrial node vocabulary in the energy industry structure diagram. The greater the node influence degree, the greater the impact of adjusting the industrial node corresponding to the industrial node vocabulary in the corresponding vocabulary category on the overall industrial chain.

[0065] Preferably, in some possible implementations of this embodiment, the method for obtaining the node influence degree is: for any vocabulary category, the number of edges of each industrial node vocabulary in the vocabulary category in the energy industry structure diagram is obtained as the first number of the corresponding industrial node vocabulary; the larger the first number, the more important and influential the industrial node corresponding to the corresponding industrial node vocabulary is in the industrial chain. In order to analyze the overall influence of the industrial nodes corresponding to the industrial node vocabulary in the vocabulary category, the sum of all the first numbers is taken as the number of related edges in the vocabulary category; the larger the number of related edges, the more important the industrial node corresponding to the industrial node vocabulary in the vocabulary category is in the overall energy industry structure diagram, and when the industrial node corresponding to the industrial node vocabulary in the vocabulary category is adjusted, the greater the impact on other industrial nodes. In order to more accurately analyze the influence of the industrial node corresponding to the industrial node vocabulary in the vocabulary category in the industrial chain, the number of policy vocabulary in the vocabulary category is further obtained. When the number of policy vocabulary in the vocabulary category is larger, it means that the industrial node corresponding to the industrial node vocabulary in the vocabulary category is more likely to be adjusted, and then the industrial node corresponding to the industrial node vocabulary in the vocabulary category has a greater impact. Therefore, the node influence degree of the vocabulary category is obtained according to the normalized result of the number of policy words in the vocabulary category and the normalized result of the number of related edges.

[0066] In order to accurately represent the degree of influence of the industrial nodes corresponding to the industrial node vocabulary in each vocabulary category, this embodiment specifically quantifies the node influence degree as a node influence degree value. The larger the node influence degree value, the greater the impact of the industrial node adjustment corresponding to the industrial node vocabulary in the corresponding vocabulary category. The calculation formula of the node influence degree value is: ; In the formula, is the node influence value of the kth vocabulary category; is the number of policy words in the kth word category; is the number of all words in the kth vocabulary category; is the number of industry node words in the kth word category; is the first number of the sth industry node vocabulary in the kth vocabulary category; is the number of related edges in the kth vocabulary category. right Normalize it by M Normalization is performed. In other embodiments, normalization methods such as sigmoid function, function conversion, maximum and minimum normalization, etc. can be used to normalize and Normalization is performed, which is not limited here. Meanwhile, in other embodiments, and The addition result of , which is not limited here. So far, the node influence value of each vocabulary category is obtained.

[0067] Step S4: According to the policy vocabulary of the revised part in the energy policy text, the adjustment vocabulary category is screened out in the vocabulary category; the industrial node vocabulary in the adjustment vocabulary category is used as the adjustment vocabulary, and the adjustment influence degree of each adjustment vocabulary is obtained according to the number of industrial node vocabulary in the preset closed loop in the energy industry structure diagram and the node influence degree of each adjustment vocabulary category.

[0068] Specifically, in order to obtain the industrial nodes that are adjusted according to the energy policy, it is conducive to more efficient analysis of the impact of the energy policy on the energy industry. Therefore, this embodiment first compares the energy policy text with the energy policy text before the revision, and obtains the part of the energy policy text that is inconsistent with the energy policy text before the revision, that is, the revised part of the energy policy text. The direct impact of the energy policy on the energy industry must be reflected in the revised part of the energy policy text. Therefore, according to the policy vocabulary of the revised part of the energy policy text, the adjustment vocabulary category is screened out in the vocabulary category.

[0069] Preferably, in some possible implementations of this embodiment, the method for obtaining the adjustment vocabulary category is as follows: this embodiment obtains all policy vocabulary corresponding to the revised part in the energy policy text as target policy vocabulary, and the industrial nodes corresponding to the industrial node vocabulary in the vocabulary category where the target policy vocabulary is located are all industrial nodes that need to be adjusted. In order to accurately analyze the impact of the energy policy on each industrial node in the subsequent period, the vocabulary category where the target policy vocabulary is located is used as the adjustment vocabulary category, and at the same time, the industrial node vocabulary in the adjustment vocabulary category is used as the adjustment vocabulary.

[0070] In other implementations of the embodiments of the present invention, some target policy words may be selected and the word categories in which the some target policy words are located may be used as adjustment word categories. The method for obtaining the adjustment word categories is not limited herein.

[0071] In actual situations, the revision of energy policy is mainly aimed at adjusting some industrial nodes, but the connections between industrial nodes will transmit the impact of the adjustment, that is, the industrial nodes are connected through Figure 2The arrows shown point to conduct influence transmission. In order to analyze the maximum impact of each adjustment vocabulary in the energy industry structure diagram, this embodiment first screens out the minimum closed loop of each adjustment vocabulary in the energy industry structure diagram through the Dijkstra algorithm. It should be noted that the minimum closed loop is not the self-loop of the adjustment vocabulary. In this embodiment, the preset closed loop is set as the minimum closed loop. The implementer can set the preset closed loop according to the actual situation, which is not limited here. Among them, the Dijkstra algorithm is a well-known technology and will not be repeated. The minimum closed loop where each adjustment vocabulary is located in the energy industry structure diagram represents the substitutability of the corresponding adjustment vocabulary in the industrial chain. When the more industrial node vocabulary exists in the minimum closed loop where a certain adjustment vocabulary is located, the more important the industrial node corresponding to the adjustment vocabulary is in the industrial chain, and it is necessary to pass through more industrial nodes before and after to reach the industrial node corresponding to the adjustment vocabulary. Therefore, the adjustment influence of the corresponding industrial node of the adjustment vocabulary is greater. At the same time, the greater the node influence of the adjustment vocabulary category in which the known adjustment vocabulary is located, the greater the adjustment influence of the corresponding adjustment vocabulary. Therefore, this embodiment obtains the adjustment influence degree of each adjustment word according to the number of industrial node words in the preset closed loop in the energy industry structure diagram and the node influence degree of the adjustment word category in which each adjustment word is located. Among them, the greater the adjustment influence degree, the greater the impact of the industrial node corresponding to the corresponding adjustment word after adjustment.

[0072] In order to accurately express the adjustment impact of each adjustment vocabulary, this embodiment specifically quantifies the adjustment impact as an adjustment impact value. The larger the adjustment impact value, the more obvious the impact of the industrial node corresponding to the adjustment vocabulary on the industrial chain after being adjusted. The calculation formula of the adjustment impact value is: ; In the formula, is the adjustment influence value of the pth adjustment word; is the number of industrial node words in the minimum closed loop of the pth adjusted word in the energy industry structure diagram; is the node influence value of the adjusted vocabulary category in which the pth adjusted vocabulary belongs. and The addition result of , which is not limited here.

[0073] At this point, the adjustment impact value of each adjustment word is obtained.

[0074] Step S5: based on the adjustment impact degree, select the focus adjustment words, and based on the adjustment impact degree and quantity of the focus adjustment words, obtain the concentration adjustment degree of the energy policy text.

[0075] Specifically, the adjustment of energy policies to industrial nodes has different manifestations. When a small number of industrial nodes have a large adjustment impact, it means that the concentration of the revised part of the energy policy is more obvious. In order to analyze the concentration of the revised part of the energy policy text, this embodiment analyzes the adjustment impact of each adjustment vocabulary, and then selects the focus adjustment vocabulary with a greater impact adjusted by the energy policy. When the number of focus adjustment vocabulary is smaller, it means that the energy policy really focuses on adjusting fewer industrial nodes, which indirectly shows that the concentration of the revised part of the energy policy is more obvious. At the same time, when the adjustment impact of the focus adjustment vocabulary is greater, it means that the concentration of changes in the revised part of the energy policy is more obvious. Therefore, this embodiment obtains the degree of centralized adjustment of the energy policy text based on the adjustment impact and number of the focus adjustment vocabulary. The greater the degree of centralized adjustment, the more obvious the adjustment of the revised part of the energy policy to the industrial nodes.

[0076] Preferably, in some possible implementations of this embodiment, the method for obtaining the concentration adjustment degree can refer to Figure 3 , which shows a flow chart of a method for obtaining a concentration adjustment degree provided by an embodiment of the present invention, the method comprising the following steps:

[0077] Step S501: Acquire focus adjustment vocabulary.

[0078] By adjusting the degree of impact, the focused adjustment vocabulary is selected from the adjustment vocabulary, which facilitates the subsequent accurate and efficient analysis of the concentrated adjustment degree of the energy policy text.

[0079] In some possible implementations of this embodiment, the method for obtaining the focused adjustment vocabulary is: arranging the adjustment vocabulary in order from low to high according to the adjustment influence, that is, arranging the adjustment vocabulary in order from small to large according to the adjustment influence value, and obtaining the adjustment vocabulary sequence. Then, the absolute value of the difference between the adjustment influence values ​​of any two adjacent adjustment vocabulary in the adjustment vocabulary sequence is obtained, and all are used as the first difference; the larger the first difference, the more likely the adjustment vocabulary with the largest adjustment influence value in the two adjacent adjustment vocabulary is to be the focused adjustment vocabulary, which indirectly indicates that the difference in the adjustment influence value of the two adjacent adjustment vocabulary is more obvious. Therefore, in this embodiment, the adjustment vocabulary with the highest adjustment influence in the two adjustment vocabulary corresponding to the largest first difference is used as the target adjustment vocabulary. It should be noted that when there are at least two first differences, the first difference that appears for the first time is mainly used to analyze and obtain the target adjustment vocabulary. Then, all the adjustment vocabulary between the target adjustment vocabulary and the last adjustment vocabulary in the adjustment vocabulary sequence from left to right are used as the focused adjustment vocabulary; wherein, the focused adjustment vocabulary includes the target adjustment vocabulary and the last adjustment vocabulary in the adjustment vocabulary sequence from left to right. So far, the focused adjustment vocabulary in the energy industry structure diagram is obtained.

[0080] Step S502: Acquire the concentration adjustment degree.

[0081] When the number of focused adjustment words is smaller, it means that the energy policy revision part actually focuses on adjusting fewer industrial nodes, and the concentration of changes in the energy policy revision part is more obvious. When the first difference is larger, it directly reflects that the concentration of changes in the energy policy revision part is more obvious. Therefore, this embodiment obtains the degree of concentrated adjustment of the energy policy text according to the number of focused adjustment words and the first difference.

[0082] In some possible implementations of this embodiment, the method for obtaining the degree of centralized adjustment is: taking the ratio of the number of adjustment words to the number of emphasis adjustment words as the first centralized reference value; the smaller the number of emphasis adjustment words, the larger the first centralized reference value, and the greater the degree of centralized adjustment of the energy policy text. The result of normalizing the largest first difference is taken as the second centralized reference value; the larger the second centralized reference value, the greater the degree of centralized adjustment of the energy policy text. Therefore, according to the first centralized reference value and the second centralized reference value, the degree of centralized adjustment of the energy policy text is obtained; wherein, the first centralized reference value and the second centralized reference value are both positively correlated with the degree of centralized adjustment.

[0083] In order to accurately express the degree of centralized adjustment of the energy policy text, this embodiment specifically quantifies the degree of centralized adjustment as a centralized adjustment degree value. The larger the centralized adjustment degree value, the more centralized the adjustment of the energy policy revision on the industrial node. The calculation formula of the centralized adjustment degree value is: ; Where D is the centralized adjustment value of the energy policy text; To adjust the number of words; Adjust the number of words for emphasis; It is the reference value in the first set; is the first difference; is the second centralized reference value; norm is a normalization function. In other embodiments, the centralized adjustment degree value of the energy policy text can be obtained by adding the first centralized reference value and the second centralized reference value, which is not limited here.

[0084] Step S6: Input the short sentences in the energy policy text containing the policy words in the adjustment vocabulary category, the adjustment influence degree and the concentrated adjustment degree of the focus adjustment words into the GPT model, and obtain the energy policy change evaluation text through the GPT model.

[0085] Specifically, in order to accurately analyze the impact of energy policies on the energy industry, it is beneficial for enterprises to make reasonable adjustments to the energy industry in a timely manner. Therefore, this embodiment obtains the energy policy change evaluation text through the generative pre-trained deep learning (Generative Pre-trained Transformer, GPT) model. Based on the obtained energy policy change evaluation text, it is helpful for enterprises to conduct a preliminary analysis of the revision of energy policies, and then conduct a more accurate analysis of the impact of energy policies. Among them, the GPT model is a well-known technology and will not be described in detail.

[0086] The specific process of obtaining the energy policy change evaluation text through the GPT model is as follows:

[0087] (1) Obtain the short sentences in the energy policy text where the policy words in the adjustment vocabulary category are located, and then input these short sentences into the GPT model to determine the relevant energy information that needs to be analyzed in the GPT model.

[0088] Preferably, in some possible implementations of this embodiment, the method for obtaining short sentences is: traverse the energy policy text from front to back, and divide it once each time a Chinese punctuation mark is detected, and each segment of the text after division is regarded as a short sentence. That is, a short sentence is a sentence in the energy policy text. This embodiment only inputs the short sentences in the energy policy text where the policy vocabulary in the adjustment vocabulary category is located into the GPT model, which can more accurately and efficiently analyze the impact of the revised part of the energy policy on the energy industry.

[0089] (2) The adjustment impact values ​​of the key adjustment words are input into the GPT model as local weights, and the GPT model is used to determine the most important impact information of the revised energy policy on the corporate energy industry.

[0090] (3) The concentration adjustment degree value is input into the GPT model as a global weight to determine the overall impact of energy policies on the energy industry.

[0091] (4) Output the energy policy change evaluation text through the softmax layer in the GPT model.

[0092] At this point, the company has accurately obtained a comprehensive energy policy change evaluation text. The company will use the obtained energy policy change evaluation text as an auxiliary means to conduct a preliminary analysis of the revision of the energy policy, which will help the company to conduct a more accurate analysis of the energy policy in the future and make reasonable adjustments to the industrial nodes in the energy industry in a timely manner.

[0093] In summary, this embodiment obtains policy vocabulary and industrial node vocabulary, divides them to obtain vocabulary categories; obtains the node influence degree according to the corresponding edges of policy vocabulary and industrial node vocabulary in the vocabulary category; obtains the adjustment vocabulary category, takes the industrial node vocabulary in the adjustment vocabulary category as the adjustment vocabulary, obtains the adjustment influence degree based on the preset closed loop and node influence degree of the adjustment vocabulary, and selects the focus adjustment vocabulary based on the focus adjustment vocabulary, and obtains the concentrated adjustment degree based on the focus adjustment vocabulary; obtains the energy policy change evaluation text based on the short sentence, adjustment influence degree and concentrated adjustment degree of the policy vocabulary in the adjustment vocabulary category. The present invention accurately obtains the energy policy change evaluation text by comprehensively considering the impact of the energy policy revision part on the energy industry, so that enterprises can adjust the energy industry in a timely and accurate manner.

[0094] Embodiment 2:

[0095] The present invention also proposes an energy policy monitoring and analysis system based on natural language processing, please refer to Figure 4 , which shows a structure diagram of an energy policy monitoring and analysis system based on natural language processing provided by an embodiment of the present invention, the system includes: an acquisition module 10, a vocabulary category acquisition module 20, a node influence degree acquisition module 30, an adjustment influence degree acquisition module 40, a centralized adjustment degree acquisition module 50 and an energy policy change evaluation text acquisition module 60.

[0096] The acquisition module 10 is used to acquire policy vocabulary from the energy policy text and industry node vocabulary from the energy industry structure diagram.

[0097] The vocabulary category acquisition module 20 is used to uniformly divide the policy vocabulary and the industry node vocabulary based on the vocabulary vectors corresponding to the policy vocabulary and the industry node vocabulary to obtain multiple vocabulary categories.

[0098] The node influence degree acquisition module 30 is used to acquire the node influence degree of each vocabulary category according to the distribution of policy vocabulary in each vocabulary category and the distribution of industrial node vocabulary edges in the energy industry structure diagram.

[0099] The adjustment impact degree acquisition module 40 is used to screen out adjustment vocabulary categories in the vocabulary categories based on the policy vocabulary of the revised part in the energy policy text; use the industrial node vocabulary in the adjustment vocabulary category as the adjustment vocabulary, and obtain the adjustment impact degree of each adjustment vocabulary based on the number of industrial node vocabulary in the preset closed loop in the energy industry structure diagram and the node impact degree of the adjustment vocabulary category in which each adjustment vocabulary is located.

[0100] The centralized adjustment degree acquisition module 50 is used to select the focused adjustment words based on the adjustment impact degree, and acquire the centralized adjustment degree of the energy policy text based on the adjustment impact degree and quantity of the focused adjustment words.

[0101] The energy policy change evaluation text acquisition module 60 is used to input the short sentences in the energy policy text where the policy words in the adjustment vocabulary category are located, the adjustment influence degree and the concentrated adjustment degree of the focus adjustment words into the GPT model, and obtain the energy policy change evaluation text through the GPT model.

[0102] It should be noted that: the system provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device is divided into different functional modules to complete all or part of the functions described above. In addition, the energy policy monitoring and analysis system based on natural language processing and the energy policy monitoring and analysis method based on natural language processing provided in the above embodiment belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0103] Embodiment 3:

[0104] The present invention also proposes an energy policy monitoring and analysis device based on natural language processing, the device includes a memory and a processor, wherein the memory stores an executable program code, and the processor is used to call and execute the executable program code to execute an energy policy monitoring and analysis method based on natural language processing provided in an embodiment of the present application. The device can be a chip, a component or a module, and the chip may include a connected processor and a memory; wherein the memory is used to store instructions, and when the processor calls and executes the instructions, the chip can execute an energy policy monitoring and analysis method based on natural language processing provided in the above embodiment.

[0105] In addition, the present application embodiment also protects a computer device, see Figure 5 The computer device includes a memory 401, a processor 402, and a computer program 403 stored in the memory 401 and running on the processor 402, wherein when the processor 402 executes the computer program 403, the computer device can execute any one of the energy policy monitoring and analysis methods based on natural language processing introduced above.

[0106] Embodiment 4:

[0107] This embodiment also provides a computer-readable storage medium, which stores computer program code. When the computer program code runs on a computer, the computer executes the above-mentioned related method steps to implement an energy policy monitoring and analysis method based on natural language processing provided in the above embodiment.

[0108] Embodiment 5:

[0109] This embodiment also provides a computer program product. When the computer program product runs on a computer, it enables the computer to execute the above-mentioned related steps to implement an energy policy monitoring and analysis method based on natural language processing provided in the above embodiment.

[0110] Among them, the device, computer-readable storage medium, computer program product or chip provided in this embodiment is used to execute the corresponding method provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding method provided above, and will not be repeated here.

[0111] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0112] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for monitoring and analyzing energy policies based on natural language processing, characterized in that: The method comprises the following steps: Obtain policy vocabulary from energy policy texts and industry node vocabulary from energy industry structure diagrams; Based on the vocabulary vectors corresponding to policy vocabulary and industrial node vocabulary, the policy vocabulary and industrial node vocabulary are uniformly divided to obtain multiple vocabulary categories; According to the distribution of policy words in each vocabulary category and the distribution of industrial node words on the edges of the energy industry structure diagram, the node influence degree of each vocabulary category is obtained; According to the policy vocabulary of the revised part in the energy policy text, the adjustment vocabulary category is screened out in the vocabulary category; the industrial node vocabulary in the adjustment vocabulary category is used as the adjustment vocabulary, and the adjustment influence degree of each adjustment vocabulary is obtained according to the number of industrial node vocabulary in the preset closed loop in the energy industry structure diagram and the node influence degree of each adjustment vocabulary category; Based on the degree of adjustment impact, the words that are focused on adjustment are selected, and based on the degree of adjustment impact and quantity of the words that are focused on adjustment, the degree of concentrated adjustment of the energy policy text is obtained; The short sentences in the energy policy text where the policy words in the adjustment vocabulary category are located, the adjustment influence degree and the concentration adjustment degree of the focus adjustment words are input into the GPT model, and the energy policy change evaluation text is obtained through the GPT model; The method for obtaining the energy policy change evaluation text is as follows: The adjustment influence value of the focused adjustment vocabulary is used as the local weight in the GPT model; The centralized adjustment degree value is used as the global weight in the GPT model; The short sentences, local weights and global weights of the policy words in the adjustment vocabulary category in the energy policy text are input into the GPT model, and the energy policy change evaluation text is output through the GPT model.

2. The energy policy monitoring and analysis method based on natural language processing according to claim 1, characterized in that: The method for obtaining the node influence degree is: For any vocabulary category, obtaining the number of edges of each industrial node vocabulary in the vocabulary category in the energy industry structure graph as the first number of the corresponding industrial node vocabulary; The sum of all the first quantities is taken as the number of related edges within the vocabulary category; The node influence degree of the vocabulary category is obtained according to the normalized result of the number of policy words in the vocabulary category and the normalized result of the number of related edges.

3. The energy policy monitoring and analysis method based on natural language processing according to claim 1, characterized in that: The method for obtaining the adjustment vocabulary category is: The vocabulary categories of the policy vocabulary in the revised part of the energy policy text shall be regarded as the adjustment vocabulary categories.

4. The energy policy monitoring and analysis method based on natural language processing according to claim 1, characterized in that: The method for obtaining the adjustment impact degree is: For any adjustment vocabulary, the product of the number of industrial node vocabulary in the preset closed loop where the adjustment vocabulary is located in the energy industry structure diagram and the node influence degree value of the adjustment vocabulary category where the adjustment vocabulary is located is obtained as the adjustment influence degree value of the adjustment vocabulary.

5. The energy policy monitoring and analysis method based on natural language processing according to claim 1, characterized in that: The acquisition method of the focus adjustment vocabulary is: Arrange the adjustment words in order from low to high according to the adjustment influence degree to obtain the adjustment word sequence; Obtaining the difference in adjustment influence between any two adjacent adjustment words in the adjustment word sequence, all of which are taken as the first difference; The adjustment vocabulary with the highest adjustment influence among the two adjustment vocabulary corresponding to the largest first difference is used as the target adjustment vocabulary; All adjustment words between the target adjustment word and the last adjustment word in the adjustment word sequence from left to right are regarded as emphasis adjustment words; wherein the emphasis adjustment words include the target adjustment word and the last adjustment word.

6. The energy policy monitoring and analysis method based on natural language processing according to claim 5, characterized in that: The method for obtaining the concentration adjustment degree is: The ratio of the number of adjusted words to the number of focused adjusted words is used as the reference value in the first set; The result of normalizing the largest first difference is used as the reference value in the second set; According to the first centralized reference value and the second centralized reference value, the centralized adjustment degree of the energy policy text is obtained; wherein, the first centralized reference value and the second centralized reference value are both positively correlated with the centralized adjustment degree.

7. The energy policy monitoring and analysis method based on natural language processing according to claim 1, characterized in that: The method for obtaining the short sentence is: The energy policy text is traversed from front to back, and each time a Chinese punctuation mark or period is detected, it is divided, and each divided text is regarded as a short sentence.

8. The energy policy monitoring and analysis method based on natural language processing according to claim 1, characterized in that: The method for obtaining the vocabulary category is: All policy words and industry node words are unified into a word set, and the word vector of each word in the word set is obtained by using the word2vec algorithm; All vocabulary vectors are converted into vector space, and the corresponding vocabulary vectors in the vector space are clustered by an iterative self-organizing clustering algorithm to obtain multiple vocabulary vector clusters; For any vocabulary vector cluster, the words corresponding to all the vocabulary vectors in the vocabulary vector cluster are constructed as a vocabulary category.

9. The energy policy monitoring and analysis method based on natural language processing according to claim 1, characterized in that: The method for obtaining the policy vocabulary and industry node vocabulary is as follows: Set the jieba segmentation mode to the precise mode, segment the energy policy text, and obtain the individual segmentations of the energy policy text as policy vocabulary; Each industry node name in the energy industry structure diagram is directly used as the industry node vocabulary.

Images