Visualization methods, apparatus, electronic devices, and storage media for claims sequences

Abstract

The present disclosure relates to a visualization method, device, electronic device and storage medium of a claim sequence, the method comprising: obtaining a to-be-processed text containing the claim sequence; generating a protection category and a reference number corresponding to each claim in the claim sequence according to the to-be-processed text; the protection category is used to distinguish between method claims and non-method claims; the reference number is the claim number of another claim cited by each claim; based on the protection category and reference number corresponding to each claim, generating a subordinate relationship corresponding to each claim; generating and displaying a hierarchical structure of the claim sequence according to the subordinate relationship corresponding to each claim. The disclosed embodiment can accurately provide a hierarchical structure diagram of the claim sequence to help users quickly understand the content of the claim sequence.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to a visualization method, device, electronic device and storage medium for claim sequence. Background Art

[0002] With the development of science and technology, people have begun to pay more and more attention to intellectual property rights, and patent documents are one of the most important means of protecting intellectual property rights. At present, the claims in patent documents are divided into independent claims and dependent claims. This text structure can easily cause readers to have difficulty understanding, especially when readers read a large number of patent documents. This problem is particularly obvious. Summary of the invention

[0003] According to one aspect of the present disclosure, a method for visualizing a claim sequence is provided, the method comprising: obtaining a text to be processed containing the claim sequence; generating, based on the text to be processed, a protection category and a reference number corresponding to each claim in the claim sequence; the protection category is used to distinguish between method claims and non-method claims; the reference number is the claim number of another claim referenced by each claim; based on the protection category and reference number corresponding to each claim, generating a subordinate relationship corresponding to each claim; and generating and displaying a hierarchical structure of the claim sequence according to the subordinate relationship corresponding to each claim.

[0004] In a possible implementation, the generation of a subordinate relationship corresponding to each claim based on the protection category and reference number corresponding to each claim includes: selecting a current claim from a plurality of claims in sequence, and performing the following steps; obtaining a first claim having the same claim number as the reference number according to the reference number corresponding to the current claim; determining whether the protection category corresponding to the first claim is consistent with the protection category corresponding to the current claim; and updating a parent node list of the current claim when it is determined that the protection category corresponding to the first claim is consistent with the protection category corresponding to the current claim; the parent node list is used to record the claim number of the first claim to which the current claim is subordinate.

[0005] In a possible implementation manner, generating a protection category and a reference number corresponding to each claim in the claim sequence according to the text to be processed also includes: selecting a current claim from a plurality of claims in sequence, and performing the following steps: determining whether the current claim has an OCR error; when it is determined that the current claim has an OCR error, updating the claim with the OCR error according to a preset OCR error correction library, wherein, according to the updated claim, generating a protection category and a reference number corresponding to each claim.

[0006] In a possible implementation, generating a protection category corresponding to each claim in the claim sequence based on the text to be processed includes: selecting a current claim from a plurality of claims in sequence, and performing the following steps: determining whether the current claim contains specific words; and in the case where it is determined that the current claim contains specific words, generating a protection category corresponding to the current claim based on the protection category corresponding to the specific words.

[0007] In a possible implementation, the visualization method further includes: dividing the claim sequence into at least one claim by using a preset paragraph segmentation model.

[0008] In a possible implementation, dividing the claim sequence into at least one claim by using a preset paragraph segmentation model includes: generating paragraph marks for each word in the claim sequence based on the preset paragraph segmentation model; and the paragraph marks are used to divide the claim sequence into at least one claim.

[0009] In a possible implementation, the paragraph segmentation model includes a convolutional layer, which performs the following operations during the training process of the paragraph segmentation model: obtaining multiple claim training sequences; classifying each claim training sequence into different text sets based on the text length value of each claim training sequence; filling each claim training sequence in each text set based on the maximum value of the text length value in each of the text sets, and generating a mask matrix based on all the filled claim training sequences; generating and updating the convolution value corresponding to each claim training sequence based on the mask matrix; the convolution value is used to represent the text feature of the claim sequence, wherein the paragraph tags generated by the paragraph segmentation model based on the convolution value are used to train the paragraph segmentation model.

[0010] In a possible implementation, dividing the claim sequence into at least one claim by using a preset paragraph segmentation model includes: acquiring a text language of the claim sequence; dividing the claim sequence into a plurality of words according to a preset word segmentation algorithm corresponding to the text language; and dividing the claim sequence after being divided into a plurality of words by using a preset paragraph segmentation model to obtain at least one claim.

[0011] In a possible implementation, dividing the claim sequence into at least one claim by using a preset paragraph segmentation model further includes: performing ambiguity resolution on each of the words divided by a preset word segmentation algorithm.

[0012] In a possible implementation, the obtaining of the to-be-processed text containing the claim sequence includes: obtaining the claim sequence number; and obtaining the to-be-processed text containing the claim sequence in a claim sequence library according to the claim sequence number.

[0013] In a possible implementation, the method further includes: generating a dependency relationship between each short sentence based on the short sentence type of each short sentence in each claim in the claim sequence through a preset short sentence dependency model; the short sentence types include: main clauses, subordinate clauses, and subordinate clause introductory sentences; and generating and displaying a hierarchical structure between the short sentences based on the dependency relationship between each of the short sentences.

[0014] In a possible implementation, the dependency relationship between each short sentence is generated based on the short sentence type of each short sentence in each claim in the claim sequence through a preset short sentence dependency model, including: selecting a current claim from multiple claims in sequence, and performing the following steps: generating a dependency parent number for each short sentence in the current claim based on the short sentence type of each short sentence in the short sentence sequence corresponding to the current claim according to the preset short sentence dependency model; wherein the short sentence sequence includes each short sentence in the current claim; and the dependency parent number is used to represent the short sentence numbers of other short sentences in the current claim that have a dependency relationship with each short sentence.

[0015] In a possible implementation, the method of generating a dependency relationship between each short sentence based on the short sentence type of each short sentence in each claim in the claim sequence through a preset short sentence dependency model also includes: selecting a current claim from multiple claims in sequence, and performing the following steps: generating a position mark corresponding to each word in the current claim based on a word sequence corresponding to the current claim according to a preset short sentence segmentation model; wherein the word sequence includes each word in the current claim; the position mark is used to indicate the position of each word in the current claim in the main clause, subordinate clause or clause introductory sentence to which it belongs; dividing the current claim into multiple short sentences with short sentence types according to the position of each word in the short sentence to which it belongs, wherein the multiple short sentences with short sentence types are used to input the short sentence dependency model to generate a dependency relationship between each short sentence.

[0016] In one possible implementation, the short sentence segmentation model includes an encoding layer and a decoding layer; the encoding layer includes a self-attention layer; the self-attention layer is provided with a detection threshold; the detection threshold is used to compare with the spacing value between any two words in the same claim to determine the position of each word in the sentence to which it belongs.

[0017] In a possible implementation, the short sentence segmentation model includes a convolutional layer, and the convolutional layer performs the following operations during the training process of the short sentence segmentation model: obtaining multiple training claims; classifying each training claim into different text sets based on the text length value of each training claim; filling each training claim in each text set based on the maximum value of the text length value in each of the text sets, and generating a mask matrix based on all the filled training claims; generating and updating the convolution value corresponding to each training claim based on the mask matrix; the convolution value is used to represent the text feature of the claim sequence, wherein the position mark generated by the short sentence segmentation model based on the convolution value is used to train the short sentence segmentation model.

[0018] According to one aspect of the present disclosure, a visualization device for a claim sequence is provided, the visualization device comprising: a text acquisition module, for acquiring a text to be processed containing the claim sequence; a category and number acquisition module, for generating, according to the text to be processed, a protection category and a reference number corresponding to each claim in the claim sequence; the protection category is used to distinguish between method claims and non-method claims; the reference number is the claim number of another claim referenced by each claim; a dependency generation module, for generating, according to a preset dependency matching algorithm, the dependency corresponding to each claim based on the protection category and the reference number corresponding to each claim; and a hierarchical structure generation module, for generating and displaying the hierarchical structure of the claim sequence according to the dependency corresponding to each claim.

[0019] According to one aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to call the instructions stored in the memory to execute the above method.

[0020] According to one aspect of the present disclosure, a computer-readable storage medium is provided, on which computer program instructions are stored, and the computer program instructions implement the above method when executed by a processor.

[0021] The visualization method, device, electronic device and storage medium of the claim sequence provided by the embodiments of the present disclosure can obtain the text to be processed containing the claim sequence, and generate the protection category and reference number corresponding to each claim in the claim sequence according to the text to be processed, and generate the subordinate relationship corresponding to each claim based on the protection category and reference number corresponding to each claim. Since the subordinate relationship corresponding to the claim is generated based on the protection category and reference number corresponding to each claim, the accuracy of its subordinate relationship is extremely high. In addition, the present disclosure can generate and display the hierarchical structure of the claim sequence according to the subordinate relationship corresponding to each claim, which is easy for readers to quickly understand the subordinate relationship of each claim.

[0022] It should be understood that the above general description and the following detailed description are exemplary and explanatory only and do not limit the present disclosure. Other features and aspects of the present disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and are used to illustrate the technical solutions of the present disclosure together with the specification.

[0024] Figure 1 A flowchart of a method for visualizing a claim sequence according to an embodiment of the present disclosure is shown.

[0025] Figure 2 A flowchart of a method for visualizing a claim sequence according to an embodiment of the present disclosure is shown.

[0026] Figure 3 Shows Figure 2 The flowchart of step S21 is shown.

[0027] Figure 4 Shows Figure 2 The flowchart of step S13 is shown.

[0028] Figure 5 The hierarchical structure diagram of the method for visualizing the claim sequence according to an embodiment of the present disclosure is shown.

[0029] Figure 6 The hierarchical structure diagram of the method for visualizing the claim sequence according to an embodiment of the present disclosure is shown.

[0030] Figure 7 A block diagram of a device for visualizing a claim sequence according to an embodiment of the present disclosure is shown.

[0031] Figure 8 A block diagram of an electronic device according to an embodiment of the present disclosure is shown.

[0032] Fig. 9 A block diagram of a storage medium according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0033] Various exemplary embodiments, features and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0034] In the description of the present disclosure, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the present disclosure and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present disclosure.

[0035] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present disclosure, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0036] In the present disclosure, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present disclosure can be understood according to specific circumstances.

[0037] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0038] The term "and / or" herein is only a description of the association relationship of the associated objects, indicating that there may be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the term "at least one" herein represents any combination of at least two of any one or more of a plurality of. For example, including at least one of A, B, and C can represent including any one or more elements selected from the set consisting of A, B, and C.

[0039] In addition, in order to better illustrate the present disclosure, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present disclosure can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present disclosure.

[0040] At present, the claims of patent documents are divided into independent claims and dependent claims. The reference relationship of some claims is relatively complex, which is not conducive to readers' reading. Especially in some complex patent documents, the number of claims is often more than dozens. If the claim dependency is too complicated, it will cause readers to be unable to understand the technical solutions in the patent documents. Related technologies generally determine the dependency of claims through simple string matching, but such methods often cannot accurately determine the dependency between claims, and incorrect dependency between claims will make it more difficult for readers to understand.

[0041] The visualization method of the claim sequence provided by the embodiment of the present disclosure can obtain the to-be-processed text containing the claim sequence, and generate the protection category and reference number corresponding to each claim in the claim sequence according to the to-be-processed text, and generate the subordinate relationship corresponding to each claim based on the protection category and reference number corresponding to each claim. Since the subordinate relationship corresponding to the claim is generated based on the protection category and reference number corresponding to each claim, the accuracy of the subordinate relationship is extremely high. In addition, the present disclosure can generate and display the hierarchical structure of the claim sequence according to the subordinate relationship corresponding to each claim, so that readers can quickly understand the subordinate relationship of each claim.

[0042] In an exemplary application scenario, a user can search in the patent search interface by searching information such as patent application number and publication number, and enter the display interface of the retrieved patent. In the display interface, a control for displaying the hierarchical structure of the claim sequence of the patent can be displayed. After the user triggers the control, the visualization method of the claim sequence according to the embodiment of the present disclosure can be run, so as to display the hierarchical structure of the claim sequence in the display interface. Exemplarily, the visualization method of the claim sequence provided by the embodiment of the present disclosure can realize the front-end display by constructing a progressive framework (such as vue.js) of the user interface, and the user can realize the storage and expansion of multiple dependent claims subordinate to the same independent claim by clicking a specific control, so that the user can quickly understand the subordinate relationship between the claims.

[0043] The visualization method of the claim sequence provided in the embodiments of the present disclosure can be executed by an electronic device such as a terminal device or a server. The terminal device can be a user equipment (User Equipment, UE), a mobile device, a user terminal, a terminal, a cellular phone, a cordless phone, a personal digital assistant (Personal Digital Assistant, PDA), a handheld device, a computing device, a vehicle-mounted device, a wearable device, etc. The method can be implemented by a processor calling computer-readable instructions stored in a memory.

[0044] See also Figure 1 , Figure 1 A flowchart showing a method for visualizing a claim sequence according to an embodiment of the present disclosure is shown.

[0045] like Figure 1 As shown, the method includes:

[0046] Step S11, obtaining the text to be processed containing the claim sequence.

[0047] Step S12, based on the text to be processed, generate the protection category and reference number corresponding to each claim in the claim sequence; the protection category is used to distinguish between method claims and non-method claims; the reference number is the claim number of another claim referenced by each claim; illustratively, the non-method claims may include a device and system with any one of a mechanical structure, an interactive module / unit, a storage medium, etc.

[0048] Step S13, generating a dependency relationship corresponding to each claim based on the protection category and reference number corresponding to each claim;

[0049] Step S14, generating and displaying the hierarchical structure of the claim sequence according to the subordinate relationship corresponding to each of the claims. The embodiment of the present disclosure does not limit the form of expression of the above hierarchical structure. For example, the hierarchical structure can be selected as a progressive framework, a tree diagram, a mind map, etc. When the method is executed by a server, the server can generate a hierarchical structure of the claim sequence and control the terminal device to display the hierarchical structure.

[0050] The visualization method of the claim sequence provided by the embodiment of the present disclosure can obtain the to-be-processed text containing the claim sequence, and generate the protection category and reference number corresponding to each claim in the claim sequence according to the to-be-processed text, and generate the subordinate relationship corresponding to each claim based on the protection category and reference number corresponding to each claim. Since the subordinate relationship corresponding to the claim is generated based on the protection category and reference number corresponding to each claim, the accuracy of the subordinate relationship is extremely high. In addition, the present disclosure can generate and display the hierarchical structure of the claim sequence according to the subordinate relationship corresponding to each claim, so that readers can quickly understand the subordinate relationship of each claim.

[0051] The text to be processed in step S11 may be a text to be processed in any document format, and the text may record a claim sequence, and each claim sequence of the text to be processed may belong to the same patent.

[0052] In a possible implementation, step S11, obtaining the text to be processed containing the claim sequence, may include: obtaining the claim sequence number, and obtaining the text to be processed containing the claim sequence in a claim sequence library according to the claim sequence number.

[0053] Exemplarily, the claim sequence library stores multiple claim sequences that do not belong to the same patent document and their corresponding claim sequence numbers. The reader can query the claim sequence that the reader wants to read in the claim sequence library through the claim sequence number corresponding to each claim sequence.

[0054] The claim sequence number may include the content of the search field of the patent to which the claim sequence belongs, such as the application number, publication number, etc. Exemplarily, the claim sequence number may be input by a user. For example, in the above application scenario, the to-be-processed text containing the claim sequence may be obtained through the application number input by the user.

[0055] See also Figure 2 , Figure 2 A flowchart showing a method for visualizing a claim sequence according to an embodiment of the present disclosure is shown.

[0056] like Figure 2 As shown, in a possible implementation, after obtaining the text to be processed containing the claim sequence, the method may further include step S21, dividing the claim sequence into at least one claim through a preset paragraph segmentation model.

[0057] In this way, the claim sequence can be automatically divided into individual claims. Those skilled in the art should understand that in some cases, such as when the claim sequence in the to-be-processed text has been divided into individual claims, step S21 can be omitted.

[0058] In a possible implementation, step S21 may be to generate a paragraph marker for each word in the claim sequence based on the claim sequence according to a preset paragraph segmentation model; the paragraph marker is used to divide the claim sequence into at least one claim.

[0059] Exemplarily, the above paragraph markers include but are not limited to: claim-beg (the marked word is the beginning of a claim), claim-mid (the marked word is the middle of a claim), and claim-end (the marked word is the end of a claim). When a word marked as claim-end is adjacent to a word marked as claim-beg, it can be determined that the two words do not belong to the same claim. In other words, the space between the two words is a paragraph divider.

[0060] In a possible implementation, the paragraph segmentation model may be one of Dilated-CNN (i.e., Dilated Convolutional Neural Networks, dilated-CNN for short) + CRF (i.e., Conditional Random Field, natural language processing, hereinafter referred to as CRF), CNN (i.e., Convolutional Neural Networks, hereinafter referred to as CNN) + CRF, GRU (i.e., Gated Recurrent Unit networks, hereinafter referred to as GRU) + CRF, LSTM (i.e., Long Short-Term Memory, hereinafter referred to as LSTM) + CRF, Bi-GRU (i.e., bidirectional gated recurrent unit network, hereinafter referred to as Bi-GRU) + CRF, Bi-LSTM + CRF, Bi-GRU, and Bi-LSTM (i.e., bidirectional long short-term memory network, hereinafter referred to as Bi-LSTM), but is not limited thereto. At present, the convolution operators of deep learning frameworks such as Keras and TensorFlow do not support variable-length prediction and variable-length training. In other words, developers must expand the length of a text to a fixed length before using the convolution operator. However, in natural language processing, the lengths of input texts are very likely to be different. At present, without using masking algorithms, training samples are usually classified into different categories according to their length. Inevitably, there are categories with too many training samples or too few training samples, which is not conducive to the training effect of the model.

[0061] The embodiments of the present disclosure can effectively solve the above problems. Here, the CNN+CRF model is taken as an example. The model includes an input layer, a convolution layer, a pooling layer, a CRF layer, an output layer, etc. Among them, the convolution layer can perform the following operations during the training process of the paragraph segmentation model:

[0062] Obtain multiple claim training sequences.

[0063] Based on the text length value of each claim training sequence, each claim training sequence is classified into different text sets. Exemplarily, this step sorts each claim training sequence based on the text length value of each claim training sequence, and sorts the sorted claim training sequences into different text sets according to a predetermined number. With such a setting, on the one hand, the number of claim sequences in each text set can be close, reducing the probability of overfitting. On the other hand, the text length values ​​between the claim sequences in each text set can be similar, which facilitates the paragraph segmentation model to extract text features.

[0064] Based on the maximum value of the text length in each of the text sets, each claim training sequence in each text set is padded, and a mask matrix is ​​generated according to all the padded claim training sequences. The mask matrix is ​​used to distinguish the padded data from the original data. Exemplarily, the value corresponding to the padded data in the mask matrix is ​​0.

[0065] Exemplarily, the relevant parameters of the convolution layer are set as follows: strides (i.e., the convolution kernel moving step) = 1, padding (i.e., the filling method) = same.

[0066] According to the mask matrix, the convolution value corresponding to each claim training sequence is generated and updated; the convolution value is used to represent the text features of the claim sequence and is called by other layers of the convolutional neural network. In one example, the representation vector of the input sentence text can be updated by the mask matrix, and then convolution is performed according to the updated vector to generate the convolution value, and then the convolution value is updated by the mask matrix to obtain the convolution value corresponding to each claim training sequence.

[0067] Exemplarily, the calculation formula of the convolution value after the mask matrix is ​​updated is:

[0068] conv1d r (x*M 卷积前掩码 ,w)*M 卷积后掩码

[0069] Among them, conv1d is the convolution operator, and r is the convolution expansion coefficient (ie, dilation rate).

[0070] x is the representation vector of the input sentence text, and the tensor dimension of x (hereinafter referred to as shape) = (m, n, d), where m is the number of sentences, n is the number of words in the longest sentence, and d is the word vector dimension.

[0071] w is the three-dimensional tensor of the convolution kernel, and is also the parameter to be trained for the paragraph segmentation model. The shape of w = (l, d, d'), where l is the size of the convolution kernel moving window field of view, d is the word vector dimension, and d' is the output word vector dimension.

[0072] M 卷积前掩码 represents the mask matrix used before convolution, which is used to set unnecessary text data to 0, and the shape of the mask is (d, n, m). Exemplarily, the unnecessary text data is a representation vector of the filling data.

[0073] M 卷积后掩码Represents the mask matrix used after convolution, which is used to set unnecessary convolution values ​​to 0. The shape of the mask is (d', n, m), where d' is the output word vector dimension. Exemplarily, the above-mentioned unnecessary convolution values ​​are the convolution values ​​corresponding to the padding data. Among them, the values ​​of the mask matrix used before convolution are equal to those of the mask matrix used after convolution.

[0074] After using the convolution mask operator, the final output value of the masked convolution, that is, the convolution value matrix, will have its head or tail value set to 0 by the mask operator, which will not interfere with subsequent calculations.

[0075] Wherein, the paragraph segmentation model generates paragraph tags based on the convolution values ​​for training the paragraph segmentation model. Exemplarily, the convolution values ​​in the convolution value matrix are passed into the CRF network operator to obtain the sequence label category, and the output layer outputs the sequence label category (i.e., paragraph tag) of the training claim. By determining the accuracy of the output sequence label category, the various parameters to be trained in the short sentence segmentation model are adjusted to realize the training process of the paragraph segmentation model.

[0076] In the disclosed embodiment, the convolution layer performs one-dimensional masked convolution. Optionally, the convolution layer may also perform multiple convolutions to increase the training depth of the paragraph segmentation model.

[0077] See also Figure 3 As shown, in a possible implementation, the claim sequence may be first divided into a plurality of words according to the text language of the claim sequence, and optionally, the words may be disambiguated, and then the processed claim sequence may be divided into at least one claim. In this case, step S21, dividing the claim sequence into at least one claim by a preset paragraph segmentation model, includes:

[0078] Step S211, obtaining the text language of the claim sequence.

[0079] Step S212, dividing the claim sequence into a plurality of words according to a preset word segmentation algorithm corresponding to the text language. Exemplarily, different word segmentation algorithms have different word segmentation standards. For example: when the text language is English, word segmentation is performed according to punctuation marks and spaces. For another example: when the text language is Chinese, semantic word segmentation is performed according to a dual array search tree algorithm.

[0080] In a possible implementation, after step S212, the method further includes performing ambiguity resolution on each of the words divided by the preset word segmentation algorithm. For example, when the text language is Chinese, since some Chinese words are not composed of single characters, it is necessary to determine whether the semantics of each word conforms to the current context. In the disclosed embodiment, merging or splitting words by means of ambiguity resolution can effectively improve the accuracy of word segmentation of the word segmentation algorithm.

[0081] For example, in step S212, the word segmentation algorithm divides the word "tea cup" into two words, namely, "tea" and "cup", both of which can be independent words. Combined with the specific context of the context: "tea cup includes a cup cover", it can be seen that one subject includes two nouns, which does not conform to Chinese grammar, so "tea" and "cup" are merged into a tea cup. The above ambiguity resolution can be achieved by algorithms such as the "constraint" ambiguity resolution algorithm and the "optimization" ambiguity resolution algorithm in the relevant technology, and the embodiments of the present disclosure will not be repeated here.

[0082] Continue reading Figure 3 As shown, in step S213, the claim sequence divided into multiple words is divided by a preset paragraph segmentation model to obtain at least one claim.

[0083] In natural language, each word is usually composed of multiple characters, and a single character has no actual semantics. For example, when the text language is English, the word 'stand' has actual semantics, while the characters 's', 't', 'n', etc. have no actual semantics. In this case, the machine learning model should learn the semantics of 'stand', not the semantics of 's', 'n', and 't'. Therefore, in order to improve the accuracy of the machine learning model, the embodiment of the present disclosure divides the claim sequence into multiple words through a word segmentation algorithm, and trains the paragraph segmentation model based on the claim sequence divided into multiple words to ensure that the text features extracted by the paragraph segmentation model are generated based on the semantics of each word, thereby increasing the accuracy of the paragraph segmentation model in dividing paragraphs. In other words, each paragraph marker generated by the paragraph segmentation model can correspond to a complete word. When a word marked as claim-end is adjacent to a word marked as claim-beg, it can be determined that the two words do not belong to the same claim, that is, the paragraph division is between the two words.

[0084] When each claim in the claim sequence in the text to be processed is determined according to the above example (for example, by executing step S21), the claim number and / or reference number of the claim may optionally be corrected to further improve the accuracy of the obtained dependency relationship.

[0085] Exemplarily, the current claim may be selected from multiple claims in sequence, and the following steps may be performed:

[0086] Determining whether the current claim has an OCR error;

[0087] When it is determined that the current claim has an OCR error, the claim with the OCR error is updated according to the preset OCR error correction library; wherein, according to the updated claim, a protection category and a reference number corresponding to each of the claims are generated. After the claim with the OCR error is updated, the subsequent step S13 is continued to generate a protection category and a reference number corresponding to each of the claims.

[0088] In one example, before determining whether the current claim has an OCR error, the following steps may be included:

[0089] According to the preset error library, determine whether there is an error in the claim number and / or reference number of the current claim. Exemplarily, the above-mentioned error library may include the error type and the corresponding algorithm for detecting the error type. For example, the error library may include an OCR (Optical Character Recognition, also known as optical character recognition, hereinafter referred to as OCR) error correction library. The error type targeted by the OCR error correction library is OCR error, and the corresponding algorithm can automatically identify whether there is an OCR error in the text. Exemplarily, the above-mentioned OCR error correction library includes typos that appear in OCR, possible combinations between typos, and corrected text of errors. The OCR error correction library can automatically correct typos to prevent the subordination of the claims from being affected by OCR errors. When the text content of the current claim is consistent with the content of the typos or the combination of typos, it is determined that the current claim has an OCR error. The above-mentioned OCR error correction library can identify errors in reference numbers.

[0090] For example: the original text is '6. A according to claim 5'. When an OCR error occurs, the original text will be mistakenly converted to 'A according to claim S', that is, "5" will be mistakenly recognized as "S". Since the reference number of the claim disappears, the parent node list of the dependent claim will not be updated, which will cause the dependent claim to be recognized as an independent claim. In the embodiment of the present disclosure, 'claim S' can be hit by the typo combination 'claim S' in the OCR error correction library, that is, the OCR error correction library can recognize that an OCR error has occurred in claim 6. The OCR error correction library records that the corrected text of 'claim S' should be 'claim 5', then the OCR error correction library updates claim 6 to '6. A according to claim 5', and the corrected text is consistent with the original text, that is, the OCR error correction of the claim is achieved.

[0091] Optionally, the above error library may include a serial number error library, the error type targeted by the serial number error library is a serial number error, and the serial number error library includes a corresponding algorithm for automatically identifying whether the claim number of each claim is generated in sequence. Exemplarily, the serial number error library determines whether a serial number error occurs in the current claim by calculating whether the difference between two adjacent claim numbers is 1.

[0092] When determining the subordination of each claim, the more important judgment factors are the claim number and the reference number. In the embodiment of the present disclosure, when errors occur in the claim number and the reference number, they are corrected first through the OCR error correction library, and secondly through the serial number error library. There are mainly two situations when errors occur in the claim number and the reference number: one is a human input error, and the other is a file conversion error. In the embodiment of the present disclosure, the file conversion error is corrected first through the OCR error correction library (which is more in line with the actual ideas of the patent writer). When the OCR error correction library cannot correct the error, the serial number error library is used to correct the claim number of the claim based on the adjacent claim number of the claim. Exemplarily, the updated claim can be marked in the user interface to remind the user that there may still be problems with the claim number and reference number of the claim.

[0093] Re-reference Figure 2 In step S12, the reference number corresponding to each of the claims can be identified from the text to be processed by OCR or other related technologies. Furthermore, the serial number of each claim can also be identified from the text to be processed by OCR or other related technologies.

[0094] In a possible implementation manner, in step S12, generating, based on the text to be processed, a protection category corresponding to each claim in the claim sequence includes:

[0095] To select a current claim from multiple claims in sequence, perform the following steps:

[0096] Determine whether the current claim contains a particular term.

[0097] When it is determined that the current claim contains a specific word, a protection category corresponding to the current claim is generated according to the protection category corresponding to the specific word.

[0098] Exemplarily, at least one of the non-method claims and the method claims corresponds to specific words, for example: the specific words corresponding to the non-method claims are: device, equipment, storage medium, component, product, system, mobile terminal, etc., and the specific words corresponding to the method claims are: method, process, algorithm, process, etc. The specific words of the method claims and non-method claims are recorded in a corpus, and the specific words of the method claims and non-method claims can be obtained by accessing the corpus. In one example, the non-method claim and one of the method claims correspond to specific words, for example: the method claim has specific words, and the non-method claim does not have specific words. If the current claim does not contain the specific words, the current claim is determined to be a non-method claim. In one example, if the current claim does not contain the specific words corresponding to the non-method claim and the method claim, the claim is marked in step S14 to remind the reader. Exemplarily, the protection category of the claim can be determined as a non-method claim or a method claim by judging which specific word is closest to the text before "wherein it is" in the current claim. Exemplarily, the subject before "wherein it is" can be extracted by the subject extraction method in the related art, and the current claim can be determined by determining whether the subject is in the specific words of the method / non-method claim.

[0099] Re-reference Figure 2 In step S13, based on the protection category and reference number corresponding to each of the claims, the dependency relationship corresponding to each of the claims is generated. At present, in the claim sequence, there are often "false dependencies" in which non-method claims refer to method claims. Readers often cannot determine whether such claims are independent claims, which leads to a certain degree of hierarchical misunderstanding when readers read the claims. The disclosed embodiment can solve the above-mentioned hierarchical misunderstanding problem by analyzing the protection category and reference number corresponding to each claim.

[0100] See also Figure 4 As shown, in a possible implementation, step S13, based on the protection category and reference number corresponding to each claim, generates a dependency relationship corresponding to each claim, which may include: selecting a current claim from multiple claims in sequence, and performing the following steps:

[0101] Step S131, according to the reference number corresponding to the current claim, obtain the first claim whose claim number is the same as the reference number.

[0102] Step S132, determining whether the protection category corresponding to the first claim is consistent with the protection category corresponding to the current claim.

[0103] Step S133, when it is determined that the protection category corresponding to the first claim is consistent with the protection category corresponding to the current claim, the parent node list of the current claim is updated; the parent node list is used to record the claim number of the first claim to which the current claim is subordinate. Optionally, when it is determined that the protection category corresponding to the first claim is inconsistent with the protection category corresponding to the current claim, the parent node list of the current claim is not updated.

[0104] For example: Claim 1 is 'a lamp, characterized in that the lamp includes a bracket, a light-emitting element...', and the current claim is 'a method for manufacturing the lamp as claimed in claim 1'. Then the first claim is claim 1. Since the protection category generated in step S12 for 'manufacturing method' is a method claim, and the protection category generated in step S12 for claim 1 is a non-method claim, then in step S133, the parent node list of the current claim is not updated. In other words, when the number of claim numbers in the parent node list is 0, the current claim can be determined to be an independent claim.

[0105] Exemplarily, each claim corresponds to a parent node list, which is set to an empty list by default. After steps S131 to S133, some claims update the parent node list, and each claim can be determined to be an independent claim or a dependent claim by judging whether the parent node list is an empty list, that is, when the parent node list is empty, the corresponding claim is an independent claim, and when the parent node list is not empty, the corresponding claim is a dependent claim. In one example, a directed network diagram can be generated through the above-mentioned multiple parent node lists, and the subsequent steps can generate and display the hierarchical structure of the claim sequence based on the subordinate relationship corresponding to each claim in the directed network diagram.

[0106] For example, if the claim sequence includes Figure 5 As shown in claims 1-5, the reference number corresponding to claim 2 is 1, and the first claim with the same reference number is claim 1. The protection categories of the two are the same, and both are method claims. Then, the serial number "1" is added to the parent node list of claim 2.

[0107] If there is also claim 6, which is a non-method claim, it refers to claims 1-5, but the protection category of claim 6 is inconsistent with that of claims 1-5, so the parent node list of claim 6 is empty, that is, claim 6 is identified as an independent claim. Thus, the incorrect identification of "false dependent rights" is avoided. For example: claim 6 is "a storage medium on which computer program instructions are stored, characterized in that when the computer program instructions are executed by a processor, the method described in any one of claims 1 to 5 is implemented". According to the specific word "storage medium" closest to 'characterized in that', it can be known that the claim is a non-method claim. Since claims 1 to 5 are method claims, the parent node list corresponding to claim 6 does not record the claim numbers of claims 1 to 5. If the parent node list corresponding to claim 6 is finally empty, claim 6 is considered to be an independent claim. Another example: claim 4 is "an oxygen preparation method according to any one of claims 1 to 3, characterized in that..." According to the specific word 'method' closest to 'characterized in that', it can be known that the claim is a method claim, and claims 1 to 3 are all method claims, so the claim numbers of claims 1 to 3 are recorded in the parent node list corresponding to claim 4. Since the parent node list of claim 4 is not empty, it can be known that claim 4 is a dependent claim, and its dependent claim number is recorded in the parent node list to facilitate subsequent retrieval of the dependent relationship of the claim.

[0108] Continue reading Figure 2 As shown, in a possible implementation, the visualization method of the claim sequence may further include: step S31, generating a dependency relationship between each short sentence based on the short sentence type of each short sentence in each claim in the claim sequence through a preset short sentence dependency model; the short sentence type includes a main clause, a subordinate clause, and a subordinate clause introductory sentence; step S32, generating and displaying a hierarchical structure between the short sentences according to the dependency relationship between each short sentence.

[0109] Thus, the dependency relationship of each short sentence in the same claim can be provided to facilitate the reader to understand each claim. Determining the dependency relationship of each short sentence in the claim and determining the subordination relationship of the claim can be performed simultaneously, or in any order, or one or both of them can be performed according to the user's choice.

[0110] In a possible implementation, step S31 generates a dependency relationship between each short sentence based on the short sentence type of each short sentence in each claim in the claim sequence through a preset short sentence dependency model, including:

[0111] To select a current claim from multiple claims in sequence, perform the following steps:

[0112] According to a preset short sentence segmentation model, based on the word sequence corresponding to the current claim, a position mark corresponding to each word in the current claim is generated; wherein the word sequence includes each word in the current claim; the position mark is used to indicate the position of each word in the current claim in the corresponding main clause or subordinate clause or clause introductory sentence.

[0113] According to the position of each word in the short sentence, the current claim is divided into a plurality of short sentences with short sentence types, wherein the plurality of short sentences with short sentence types are used to input a short sentence dependency model to generate dependency relations between each short sentence.

[0114] Exemplarily, the above position markers may include: subject-beg (the marked word is the beginning of a main sentence), subject-mid (the marked word is the middle of a main sentence), subject-end (the marked word is the end of a main sentence), subordinate-beg (the marked word is the beginning of a subordinate clause), subordinate-mid (the marked word is the middle of a subordinate clause), subordinate-end (the marked word is the end of a subordinate clause). For example: the word marked as subject-end is adjacent to the word marked as subordinate-beg, which means that the two words belong to different short sentences, and the short sentence division is between the two words.

[0115] For example, the claim is "According to the file update method of claim 1, characterized in that the method further comprises: updating the file path." In the claim, the word 'according to' is marked as the beginning of the main clause; the word 'method' in 'update method' is marked as the end of the main clause; 'it' is marked as the beginning of the subordinate clause introduction; 'in' is marked as the end of the subordinate clause introduction; 'the method' is marked as the beginning of the subordinate clause; 'path' is marked as the end of the subordinate clause. Exemplarily, the subordinate clause part can be expanded or collapsed by a control to facilitate the reader's reading.

[0116] In one possible implementation, the short sentence segmentation model includes one or more self-attention layers (i.e., self-attention layers); the self-attention layer is used to perform a dot product operation on the word vectors of any two words in the same claim to determine the position of each word in the sentence to which it belongs. For example, if the input sentence length is 10,000 words, the size of the self-attention matrix (e.g., similarity matrix) is 10,000*10,000, which greatly reduces the operational efficiency of the short sentence segmentation model and takes up a lot of storage space. Attached here is the calculation formula for the semantic representation of words in the self-attention layer (the relevant parameters will be explained below):

[0117] z i =softmax(a) i ×V

[0118] From this formula, it can be seen intuitively that the formula has not been interfered with by the mask operator. In other words, when the number of words is too large, this formula will lead to the problem of the self-attention matrix being too large as mentioned above. The relevant parameters will be explained below.

[0119] In the disclosed embodiment, the size of the self-attention matrix is ​​reduced by setting the detection threshold, and the size of the self-attention matrix can be simplified from 10000*10000 to 10000*(2*detection threshold+1), which greatly improves the operational efficiency of the short sentence segmentation model and can save storage space. Exemplarily, if the distance between two words is greater than the detection threshold, the attention value of the two is set to 0, thereby reducing the size of the self-attention matrix. Exemplarily, the calculation formula of each word semantic representation (also called word vector) after adjustment by the detection threshold is:

[0120] z i =softmax(a′) i ×V

[0121] The parameters are defined as follows:

[0122] z is the output sentence tensor, and i is the word number.

[0123] V=x×W V , V is the input representation of the sentence text in the self-attention operator, x is the representation value of the input sentence text (a high-dimensional vector), W V is a parameter to be trained.

[0124] a′=a-∞(1-BandPart l,u(1)), a represents the self-attention matrix, l and u are two integer parameters of the BandPart operator, l is the following word vector, and u is the previous word vector. a' is the area of ​​the self-attention matrix expanded along the diagonal, and the remaining areas will not affect subsequent operations. For example: when u = l = 0, only the values ​​corresponding to the diagonal in the self-attention matrix are obtained, and the remaining values ​​are masked. When u = -1 and l = 1, the values ​​on the diagonal in the self-attention matrix and the two lines obtained by translating the diagonal up and down by one unit are obtained. The self-attention matrix is ​​reduced by setting the values ​​of u and l to the detection threshold. In other words, if the distance between two words is greater than the detection threshold, it is considered that there is no connection between the two words.

[0125] Among them, define BandPart l,u (x)=x*[j:n,i:m]|ij∈[-u;l+1)|, that is: the difference between i and j is a valid input in the interval [-u;l+1), j is the word number, that is, the values ​​in the self-attention matrix are screened.

[0126] In a possible implementation, the above-mentioned short sentence segmentation model may be one of Dilated-CNN+Self-Attention (i.e., self-attention model)+CRF, CNN+Self-Attention+CRF, GRU+CRF, LSTM+CRF, Bi-GRU+CRF, Bi-LSTM+CRF, Bi-GRU, and Bi-LSTM, but is not limited thereto. At present, the convolution operators of deep learning frameworks such as Keras and TensorFlow do not support variable-length prediction and variable-length training. In other words, developers must expand the length of a text to a fixed length before using the convolution operator. However, in natural language processing, the text lengths between the input texts are extremely likely to be unequal. At present, without using a masking algorithm, training samples are usually classified into different categories according to length. Inevitably, there are categories with too many training samples or too few training samples, which is not conducive to the training effect of the model.

[0127] The embodiments of the present disclosure can effectively solve the above problems. Here, the CNN+CRF model is taken as an example. The model may include an input layer, a convolution layer, a pooling layer, a CRF layer, an output layer, etc. Among them, the convolution layer performs the following operations during the training process of the short sentence segmentation model:

[0128] Get multiple training claims.

[0129] Based on the text length value of each training claim, each training claim is classified into different text sets. Specifically, based on the text length value of each training claim, each training claim is sorted, and the sorted training claims are classified into different text sets according to a predetermined number. Such a setting can, on the one hand, make the number of training claims in each text set close to reduce the probability of overfitting, and on the other hand, make the text length values ​​between the training claims in each text set similar, which is convenient for the short sentence segmentation model to extract text features.

[0130] Based on the maximum value of the text length value in each of the text sets, each training claim in each text set is filled, and a mask matrix is ​​generated based on all the filled training claims; the above-mentioned mask matrix is ​​used to distinguish between the filled data and the original text data. Exemplarily, the value corresponding to the filled data in the mask matrix is ​​0.

[0131] Exemplarily, the relevant parameters of the convolution layer are set as follows: strides (i.e., the convolution kernel moving step) = 1, padding (i.e., the filling method) = same.

[0132] According to the mask matrix, the convolution value corresponding to each training claim is generated and updated; the convolution value is used to represent the text features of the claim sequence and is called by other layers of the convolutional neural network. In one example, the representation vector of the input sentence text can be updated by the mask matrix, and then convolution is performed according to the updated vector to generate the convolution value, and then the convolution value is updated by the mask matrix to obtain the convolution value corresponding to each claim training sequence.

[0133] Exemplarily, the calculation formula of the convolution value after the mask matrix is ​​updated is:

[0134] conv1d r (x*M 卷积前掩码 ,w)*M 卷积后掩码

[0135] Among them, conv1d is the convolution operator, and r is the convolution expansion coefficient (ie, dilation rate).

[0136] x is the representation vector of the input sentence text, and the tensor dimension of x (hereinafter referred to as shape) = (m, n, d), where m is the number of sentences, n is the number of words in the longest sentence, and d is the word vector dimension.

[0137] w is the three-dimensional tensor of the convolution kernel, and is also the parameter to be trained for the short sentence segmentation model. The shape of w = (l, d, d'), where l is the field of view of the convolution kernel moving window, d is the word vector dimension, and d' is the output word vector dimension.

[0138] M 卷积前掩码 Represents the mask before convolution, the shape of the mask = (d, n, m).

[0139] M 卷积后掩码 Represents the mask after convolution, which is used to set unnecessary convolution values ​​to 0. The shape of the mask is (d', n, m), where d' is the output word vector dimension.

[0140] After using the convolution mask operator, the final output value of the masked convolution, that is, the convolution value matrix, has its head and tail values ​​set to 0 by the mask operator, which will not interfere with subsequent calculations.

[0141] The position mark generated by the convolution value of the short sentence segmentation model is used to train the short sentence segmentation model. Exemplarily, the convolution value in the convolution value matrix is ​​passed into the CRF network operator to obtain the sequence label category, and the output layer outputs the sequence label category (i.e., the position mark) of the training claim. By determining the accuracy of the output sequence label category, the various parameters to be trained in the short sentence segmentation model are adjusted to realize the training process of the short sentence segmentation model.

[0142] In the disclosed embodiment, the convolution layer performs one-dimensional masked convolution. Optionally, the convolution layer may also perform multiple masked convolutions to increase the training depth of the paragraph segmentation model.

[0143] In a possible implementation, a dependency relationship between each short sentence is generated based on the short sentence type of each short sentence in each claim in the claim sequence through a preset short sentence dependency model, including:

[0144] To select a current claim from multiple claims in sequence, perform the following steps:

[0145] According to the preset short sentence dependency model, based on the short sentence type of each short sentence in the short sentence sequence corresponding to the current claim, the dependency parent number of each short sentence in the current claim is generated; exemplarily, the short sentence dependency model described in the embodiment of the present disclosure may be a short sentence dependency model in any related technology, and the embodiment of the present disclosure is not limited here. Exemplarily, the short sentence sequence corresponding to the current claim is input into the short sentence dependency model, and the short sentence dependency model compresses the short sentences in the short sentence sequence into text vectors, and generates a directed graph spanning tree through the Deep Biaffine Attention for Neural Dependency Parsing model (i.e., the deep biaffine attention mechanism neural dependency analysis model) in the short sentence dependency model. By reading the directed graph spanning tree, the dependency parent number of each short sentence can be obtained.

[0146] The short sentence sequence includes each short sentence in the current claim; the dependent parent number is used to indicate the short sentence numbers of other short sentences in the current claim that have a dependent relationship with each short sentence. For example: if the dependent parent number of the short sentence sequence is 2, then the short sentence is a clause of the short sentence with the short sentence number 2. If the dependent parent number of the short sentence sequence is -1, then the short sentence is not a clause of any short sentence.

[0147] Exemplarily, for the purpose of simplifying the user interface, each independent claim can be simply divided into a main clause (i.e., the preamble of the claim, i.e., the text before 'It is characterized by'), a clause introductory sentence (i.e., 'It is characterized by'), and a clause (i.e., the characteristic part of the claim, i.e., the text after 'It is characterized by'), and each dependent claim can be divided into a main clause (i.e., the reference part of the claim, i.e., the text before 'It is characterized by'), a clause introductory sentence (i.e., 'It is characterized by'), and a clause (i.e., the characteristic part of the claim, i.e., the text after 'It is characterized by').

[0148] The visualization method of claim sequence provided by the embodiment of the present disclosure can generate a hierarchical structure based on the dependency relationship between claims, so as to facilitate readers to understand the dependency relationship between claims. In addition, the embodiment of the present disclosure can generate a hierarchical structure based on the dependency relationship between each short sentence in each claim, so as to facilitate readers to understand the dependency relationship between each short sentence in the same claim.

[0149] See also Figure 5 , 6 As shown, Figure 5 , 6 The diagram is a hierarchical structure diagram of a method for visualizing a claim sequence according to an embodiment of the present disclosure.

[0150] Figure 5 The left side shows the hierarchical structure of the claim sequence, with independent claims 1 and 5 marked in different colors, and the dependent relationship between the claims is expressed by the connection relationship of line segments, and the dependent claims are displayed in adjacent positions, such as claims 1, 2, and 4. The clause part of each claim can be collapsed or expanded to facilitate reading.

[0151] Figure 6 The left side shows the hierarchical structure of the claim sequence, and the connection relationship of different line segments expresses the dependent relationship of the claims. For example, claims 1 and 2 are independent claims, and claims 8, 9, 10, 12, and 13 are dependent claims of claims 1 and 2. The clauses of each claim can be collapsed or expanded to facilitate reading.

[0152] See also Figure 7 , Figure 7 A block diagram of a device for visualizing a claim sequence according to an embodiment of the present disclosure is shown.

[0153] See also Figure 7 As shown, the embodiment of the present disclosure also provides a visualization device 100 for a claim sequence, and the visualization device includes: a text acquisition module 110, used to acquire a text to be processed containing the claim sequence; a category and number acquisition module 120, used to generate a protection category and a reference number corresponding to each claim in the claim sequence according to the text to be processed; the protection category is used to distinguish between method claims and non-method claims; the reference number is the claim number of another claim referenced by each claim; a subordination generation module 130, used to generate a subordination corresponding to each claim based on the protection category and the reference number corresponding to each claim according to a preset subordination matching algorithm; a hierarchical structure generation module 140, used to generate and display the hierarchical structure of the claim sequence according to the subordination corresponding to each claim.

[0154] In a possible implementation, the generation of a subordinate relationship corresponding to each claim based on the protection category and reference number corresponding to each claim includes: selecting a current claim from a plurality of claims in sequence, and performing the following steps; obtaining a first claim having the same claim number as the reference number according to the reference number corresponding to the current claim; determining whether the protection category corresponding to the first claim is consistent with the protection category corresponding to the current claim; and updating a parent node list of the current claim when it is determined that the protection category corresponding to the first claim is consistent with the protection category corresponding to the current claim; the parent node list is used to record the claim number of the first claim to which the current claim is subordinate.

[0155] In a possible implementation manner, generating a protection category and a reference number corresponding to each claim in the claim sequence according to the text to be processed also includes: selecting a current claim from a plurality of claims in sequence, and performing the following steps: determining whether the current claim has an OCR error; when it is determined that the current claim has an OCR error, updating the claim with the OCR error according to a preset OCR error correction library, wherein, according to the updated claim, generating a protection category and a reference number corresponding to each claim.

[0156] In a possible implementation, generating a protection category corresponding to each claim in the claim sequence based on the text to be processed includes: selecting a current claim from a plurality of claims in sequence, and performing the following steps: determining whether the current claim contains specific words; and in the case where it is determined that the current claim contains specific words, generating a protection category corresponding to the current claim based on the protection category corresponding to the specific words.

[0157] In a possible implementation, the visualization device further includes a claim division module for dividing the claim sequence into at least one claim through a preset paragraph segmentation model.

[0158] In a possible implementation, dividing the claim sequence into at least one claim by using a preset paragraph segmentation model includes: generating paragraph marks for each word in the claim sequence based on the preset paragraph segmentation model; and the paragraph marks are used to divide the claim sequence into at least one claim.

[0159] In a possible implementation, the paragraph segmentation model includes a convolutional layer, which performs the following operations during the training process of the paragraph segmentation model: obtaining multiple claim training sequences; classifying each claim training sequence into different text sets based on the text length value of each claim training sequence; filling each claim training sequence in each text set based on the maximum value of the text length value in each of the text sets, and generating a mask matrix based on all the filled claim training sequences; generating and updating the convolution value corresponding to each claim training sequence based on the mask matrix; the convolution value is used to represent the text feature of the claim sequence, wherein the paragraph tags generated by the paragraph segmentation model based on the convolution value are used to train the paragraph segmentation model.

[0160] In a possible implementation, dividing the claim sequence into at least one claim by using a preset paragraph segmentation model includes: acquiring a text language of the claim sequence; dividing the claim sequence into a plurality of words according to a preset word segmentation algorithm corresponding to the text language; and dividing the claim sequence after being divided into a plurality of words by using a preset paragraph segmentation model to obtain at least one claim.

[0161] In a possible implementation, dividing the claim sequence into at least one claim by using a preset paragraph segmentation model further includes: performing ambiguity resolution on each of the words divided by a preset word segmentation algorithm.

[0162] In a possible implementation, the obtaining of the to-be-processed text containing the claim sequence includes: obtaining the claim sequence number; and obtaining the to-be-processed text containing the claim sequence in a claim sequence library according to the claim sequence number.

[0163] In a possible implementation, the visualization device further includes a dependency generation module for generating dependencies between each short sentence based on the short sentence type of each short sentence in each claim in the claim sequence through a preset short sentence dependency model; the short sentence types include: main clauses, subordinate clauses, and clause introductory sentences; and according to the dependency between each short sentence, a hierarchical structure between the short sentences is generated and displayed.

[0164] In a possible implementation, the dependency relationship between each short sentence is generated based on the short sentence type of each short sentence in each claim in the claim sequence through a preset short sentence dependency model, including: selecting a current claim from multiple claims in sequence, and performing the following steps: generating a dependency parent number for each short sentence in the current claim based on the short sentence type of each short sentence in the short sentence sequence corresponding to the current claim according to the preset short sentence dependency model; wherein the short sentence sequence includes each short sentence in the current claim; and the dependency parent number is used to represent the short sentence numbers of other short sentences in the current claim that have a dependency relationship with each short sentence.

[0165] In a possible implementation, the method of generating a dependency relationship between each short sentence based on the short sentence type of each short sentence in each claim in the claim sequence through a preset short sentence dependency model also includes: selecting a current claim from multiple claims in sequence, and performing the following steps: generating a position mark corresponding to each word in the current claim based on a word sequence corresponding to the current claim according to a preset short sentence segmentation model; wherein the word sequence includes each word in the current claim; the position mark is used to indicate the position of each word in the current claim in the main clause, subordinate clause or clause introductory sentence to which it belongs; dividing the current claim into multiple short sentences with short sentence types according to the position of each word in the short sentence to which it belongs, wherein the multiple short sentences with short sentence types are used to input the short sentence dependency model to generate a dependency relationship between each short sentence.

[0166] In one possible implementation, the short sentence segmentation model includes an encoding layer and a decoding layer; the encoding layer includes a self-attention layer; the self-attention layer is provided with a detection threshold; the detection threshold is used to compare with the spacing value between any two words in the same claim to determine the position of each word in the sentence to which it belongs.

[0167] In a possible implementation, the short sentence segmentation model includes a convolutional layer, and the convolutional layer performs the following operations during the training process of the short sentence segmentation model: obtaining multiple training claims; classifying each training claim into different text sets based on the text length value of each training claim; filling each training claim in each text set based on the maximum value of the text length value in each of the text sets, and generating a mask matrix based on all the filled training claims; generating and updating the convolution value corresponding to each training claim based on the mask matrix; the convolution value is used to represent the text feature of the claim sequence, wherein the position mark generated by the short sentence segmentation model based on the convolution value is used to train the short sentence segmentation model.

[0168] In some embodiments, the functions or modules included in the device provided by the embodiments of the present disclosure can be used to execute the method described in the above method embodiments. The specific implementation can refer to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0169] The embodiment of the present disclosure also provides a computer-readable storage medium on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the above method is implemented. The computer-readable storage medium may be a non-volatile computer-readable storage medium.

[0170] An embodiment of the present disclosure further provides an electronic device, comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to call the instructions stored in the memory to execute the above method.

[0171] The embodiments of the present disclosure also provide a computer program product, including a computer-readable code, or a non-volatile computer-readable storage medium carrying the computer-readable code. When the computer-readable code runs in a processor of an electronic device, the processor in the electronic device executes the above method.

[0172] The electronic device may be provided as a terminal, a server, or a device in other forms.

[0173] See also Figure 8 , Figure 8 A block diagram of an electronic device according to an embodiment of the present disclosure is shown.

[0174] For example, the electronic device 800 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, or the like.

[0175] Reference Figure 8, the electronic device 800 may include one or more of the following components: a processing component 802 , a memory 804 , a power component 806 , a multimedia component 808 , an audio component 810 , an input / output (I / O) interface 812 , a sensor component 814 , and a communication component 816 .

[0176] The processing component 802 generally controls the overall operation of the electronic device 800, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the above-mentioned method. In addition, the processing component 802 may include one or more modules to facilitate the interaction between the processing component 802 and other components. For example, the processing component 802 may include a multimedia module to facilitate the interaction between the multimedia component 808 and the processing component 802.

[0177] The memory 804 is configured to store various types of data to support operations on the electronic device 800. Examples of such data include instructions for any application or method operating on the electronic device 800, contact data, phone book data, messages, pictures, videos, etc. The memory 804 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0178] The power supply component 806 provides power to the various components of the electronic device 800. The power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the electronic device 800.

[0179] The multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front camera and / or a rear camera. When the electronic device 800 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera may receive external multimedia data. Each front camera and rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

[0180] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC), and when the electronic device 800 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 804 or sent via the communication component 816. In some embodiments, the audio component 810 also includes a speaker for outputting audio signals.

[0181] I / O interface 812 provides an interface between processing component 802 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include but are not limited to: home button, volume button, start button, and lock button.

[0182] The sensor assembly 814 includes one or more sensors for providing various aspects of status assessment for the electronic device 800. For example, the sensor assembly 814 can detect the open / closed state of the electronic device 800, the relative positioning of the components, such as the display and keypad of the electronic device 800, and the sensor assembly 814 can also detect the position change of the electronic device 800 or a component of the electronic device 800, the presence or absence of contact between the user and the electronic device 800, the orientation or acceleration / deceleration of the electronic device 800, and the temperature change of the electronic device 800. The sensor assembly 814 may include a proximity sensor configured to detect the presence of a nearby object without any physical contact. The sensor assembly 814 may also include an optical sensor, such as a complementary metal oxide semiconductor (CMOS) or a charge coupled device (CCD) image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0183] The communication component 816 is configured to facilitate wired or wireless communication between the electronic device 800 and other devices. The electronic device 800 can access a wireless network based on a communication standard, such as a wireless network (WiFi), a second generation mobile communication technology (2G) or a third generation mobile communication technology (3G), or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0184] In an exemplary embodiment, the electronic device 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above methods.

[0185] In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory 804 including computer program instructions, which can be executed by a processor 820 of an electronic device 800 to perform the above method.

[0186] See also Fig. 9 , Fig. 9 A block diagram of an electronic device according to an embodiment of the present disclosure is shown.

[0187] For example, the electronic device 1900 may be provided as a server. Fig. 9 , the electronic device 1900 includes a processing component 1922, which further includes one or more processors, and a memory resource represented by a memory 1932 for storing instructions executable by the processing component 1922, such as an application. The application stored in the memory 1932 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 1922 is configured to execute instructions to perform the above method.

[0188] The electronic device 1900 may also include a power supply component 1926 configured to perform power management of the electronic device 1900, a wired or wireless network interface 1950 configured to connect the electronic device 1900 to a network, and an input / output (I / O) interface 1958. The electronic device 1900 may operate based on an operating system stored in the memory 1932, such as a Microsoft Server operating system (Windows Server 2003). TM ), a graphical user interface operating system launched by Apple (Mac OSX TM ), a multi-user, multi-process computer operating system (Unix TM ), a free and open source Unix-like operating system (Linux TM ), an open source Unix-like operating system (FreeBSD TM ) or similar.

[0189] In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory 1932 including computer program instructions, which can be executed by the processing component 1922 of the electronic device 1900 to perform the above method.

[0190] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0191] Computer readable storage medium can be a tangible device that can hold and store instructions used by an instruction execution device. Computer readable storage medium can be, for example, (but not limited to) an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. More specific examples (non-exhaustive list) of computer readable storage medium include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, for example, a punch card or a convex structure in a groove on which instructions are stored, and any suitable combination thereof. The computer readable storage medium used here is not interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagated by a waveguide or other transmission medium (for example, a light pulse by an optical fiber cable), or an electrical signal transmitted by a wire.

[0192] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0193] The computer program instructions for performing the operation of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed completely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be customized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present disclosure.

[0194] Various aspects of the present disclosure are described herein with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0195] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0196] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0197] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of the module, program segment or instruction includes one or more executable instructions for realizing the specified logical function. In some alternative implementations, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of special hardware and computer instructions.

[0198] The computer program product may be implemented in hardware, software or a combination thereof. In one optional embodiment, the computer program product is embodied as a computer storage medium, and in another optional embodiment, the computer program product is embodied as a software product, such as a software development kit (SDK) and the like.

[0199] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of the module, program segment or instruction includes one or more executable instructions for realizing the specified logical function. In some alternative implementations, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of special hardware and computer instructions.

[0200] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for visualizing a claim sequence, characterized in that: The method comprises: Obtaining the text to be processed containing the claim sequence; Generate, according to the text to be processed, a protection category and a reference number corresponding to each claim in the claim sequence; the protection category is used to distinguish between method claims and non-method claims; the reference number is the claim number of another claim referenced by each claim; Based on the protection category and reference number corresponding to each of the claims, generate a dependency relationship corresponding to each of the claims; According to the dependency relationship corresponding to each of the claims, a hierarchical structure of the claim sequence is generated and displayed.

2. The visualization method according to claim 1, characterized in that: The generating of the dependency relationship corresponding to each claim based on the protection category and reference number corresponding to each claim includes: Selecting a current claim from a plurality of claims in sequence, performing the following steps; According to the reference number corresponding to the current claim, obtain the first claim with the same claim number as the reference number; Determining whether the protection category corresponding to the first claim is consistent with the protection category corresponding to the current claim; When it is determined that the protection category corresponding to the first claim is consistent with the protection category corresponding to the current claim, the parent node list of the current claim is updated; the parent node list is used to record the claim number of the first claim to which the current claim is subordinate.

3. The visualization method according to claim 2, characterized in that: The step of generating, according to the text to be processed, a protection category and a reference number corresponding to each claim in the claim sequence, further comprises: To select a current claim from multiple claims in sequence, perform the following steps: Determining whether the current claim has an OCR error; In the case where it is determined that the current claim has an OCR error, the claim having the OCR error is updated according to a preset OCR error correction library, Among them, according to the updated claims, the protection category and reference number corresponding to each claim are generated.

4. The visualization method according to claim 1, characterized in that: Generating the protection category corresponding to each claim in the claim sequence according to the text to be processed includes: To select a current claim from multiple claims in sequence, perform the following steps: Determine whether the current claim contains a specific wording; When it is determined that the current claim contains a specific word, a protection category corresponding to the current claim is generated according to the protection category corresponding to the specific word.

5. The visualization method according to claim 1, characterized in that: The visualization method further comprises: The claim sequence is divided into at least one claim by using a preset paragraph segmentation model.

6. The visualization method according to claim 5, characterized in that: The method of dividing the claim sequence into at least one claim by using a preset paragraph segmentation model comprises: generating a paragraph mark for each word in the claim sequence based on the preset paragraph segmentation model; and the paragraph mark is used to divide the claim sequence into at least one claim.

7. The visualization method according to claim 6, characterized in that: The paragraph segmentation model includes a convolutional layer, and the convolutional layer performs the following operations during the training of the paragraph segmentation model: Acquire multiple claim training sequences; Based on the text length value of each claim training sequence, classify each claim training sequence into different text sets; Fill each claim training sequence in each text set based on the maximum value of the text length value in each of the text sets, and generate a mask matrix according to all the filled claim training sequences; According to the mask matrix, a convolution value corresponding to each claim training sequence is generated and updated; the convolution value is used to represent the text feature of the claim sequence, Wherein, the paragraph tags generated by the paragraph segmentation model based on the convolution values ​​are used to train the paragraph segmentation model.

8. The visualization method according to claim 5, characterized in that: The step of dividing the claim sequence into at least one claim by using a preset paragraph segmentation model includes: Obtaining the text language of the claim sequence; Dividing the claim sequence into a plurality of words according to a preset word segmentation algorithm corresponding to the text language; The claim sequence divided into a plurality of words is divided by a preset paragraph segmentation model to obtain at least one claim.

9. The visualization method according to claim 8, characterized in that: The method of dividing the claim sequence into at least one claim by using a preset paragraph segmentation model further includes: performing ambiguity resolution on each of the words divided by a preset word segmentation algorithm.

10. The visualization method according to claim 1, characterized in that: The step of obtaining the text to be processed containing the claim sequence includes: Get the claim serial number; According to the claim sequence number, a text to be processed containing the claim sequence is obtained from a claim sequence library.

11. The visualization method according to claim 1, characterized in that: The method further comprises: Generate dependency relationships between each short sentence based on the short sentence type of each short sentence in each claim in the claim sequence through a preset short sentence dependency model; the short sentence types include: main clauses, subordinate clauses, and subordinate clause introductory sentences; According to the dependency relationship between each of the short sentences, a hierarchical structure between the short sentences is generated and displayed.

12. The visualization method according to claim 11, characterized in that: The method of generating the dependency relationship between each short sentence based on the short sentence type of each short sentence in each claim in the claim sequence by using a preset short sentence dependency model includes: To select a current claim from multiple claims in sequence, perform the following steps: According to a preset short sentence dependency model, based on the short sentence type of each short sentence in the short sentence sequence corresponding to the current claim, a dependency parent number of each short sentence in the current claim is generated; wherein the short sentence sequence includes each short sentence in the current claim; the dependency parent number is used to represent the short sentence numbers of other short sentences in the current claim that have a dependency relationship with each short sentence.

13. The visualization method according to claim 12, characterized in that: The method of generating the dependency relationship between each short sentence based on the short sentence type of each short sentence in each claim in the claim sequence by using a preset short sentence dependency model further includes: To select a current claim from multiple claims in sequence, perform the following steps: According to a preset short sentence segmentation model, based on the word sequence corresponding to the current claim, a position mark corresponding to each word in the current claim is generated; wherein the word sequence includes each word in the current claim; the position mark is used to indicate the position of each word in the main clause or subordinate clause or clause introductory sentence to which it belongs in the current claim; According to the position of each word in the corresponding short sentence, the current claim is divided into a plurality of short sentences with short sentence types, wherein the plurality of short sentences with short sentence types are used to input a short sentence dependency model to generate a dependency relationship between each short sentence.

14. The visualization method according to claim 13, characterized in that: The short sentence segmentation model includes an encoding layer and a decoding layer; the encoding layer includes a self-attention layer; the self-attention layer is provided with a detection threshold; the detection threshold is used to compare with the spacing value of any two words in the same claim to determine the position of each word in the sentence to which it belongs.

15. The visualization method according to any one of claims 13 to 14, characterized in that: The short sentence segmentation model includes a convolutional layer, and the convolutional layer performs the following operations during the training process of the short sentence segmentation model: Obtain multiple training claims; Based on the text length value of each training claim, classify each training claim into different text sets; Fill each training claim in each text set based on the maximum value of the text length value in each of the text sets, and generate a mask matrix according to all the filled training claims; According to the mask matrix, a convolution value corresponding to each training claim is generated and updated; the convolution value is used to represent the text features of the claim sequence, wherein the position markers generated by the short sentence segmentation model based on the convolution value are used to train the short sentence segmentation model.

16. A device for visualizing a claim sequence, characterized in that: The visualization device comprises: A text acquisition module, used to acquire the text to be processed containing the claim sequence; A category and number acquisition module, used to generate a protection category and a reference number corresponding to each claim in the claim sequence according to the text to be processed; the protection category is used to distinguish between method claims and non-method claims; the reference number is the claim number of another claim referenced by each claim; A dependency generation module, for generating a dependency corresponding to each of the claims based on a preset dependency matching algorithm and the protection category and reference number corresponding to each of the claims; The hierarchical structure generating module is used to generate and display the hierarchical structure of the claim sequence according to the subordinate relationship corresponding to each of the claims.

17. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method for visualizing the claim sequence described in any one of claims 1 to 15.

18. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method for visualizing the claim sequence described in any one of claims 1 to 15 is implemented.

Images