A method and system for mathematical formula recognition and extraction

Abstract

The application discloses a kind of for mathematical formula identification and extraction method and system, method includes: obtaining input original text data;The original text data is preprocessed, and the text information after preprocessing is obtained;Formula identification extraction method is used to extract mathematical formula to the text information after preprocessing, and extraction result is obtained;Output extraction result.The method can accurately improve the accuracy and adaptability of mathematical formula extraction, and extraction efficiency is high, improve the efficiency of processing a large amount of text.

Description

Technical Field

[0001] This invention relates to the field of online education technology, specifically to a method and system for mathematical formula recognition and extraction. Background Technology

[0002] Mathematical formula extraction is a challenging yet practically significant task. In the field of online education, identifying and extracting mathematical formulas from the plain text of user responses is a core component of intelligent grading. For example, the formula "20×2 beginning year equity value of Company B = 20×1 end year equity value of Company B = 350×90%×(1+7%) / (11%-7%) = 8426.25 (ten thousand yuan)" is correct. Here, "20×2" represents the year, not "20 multiplied by 2," and "=20×1" represents the year, not "20 multiplied by 1." Therefore, "=350×90%×(1+7%) / (11%-7%) = 8426.25 (ten thousand yuan)" is correct. Furthermore, user responses are often non-standard. While the system can determine the meaning of "20×1" by using the word "year" in "20×1 year," it is quite difficult to determine the meaning of "20×1" when the user does not specify keywords such as "year".

[0003] The main solutions currently available can be categorized into three types: rule-based methods, template-based methods, and machine learning-based methods.

[0004] 1. Rule-based methods: Rule-based methods primarily rely on predefined rules to extract mathematical formulas by matching these rules. For example, formulas can be identified by matching specific patterns in LaTeX code, or by using regular expressions to match formulas that conform to predefined patterns. Rule-based methods are also the most commonly used approach.

[0005] Advantages: This method works well when processing formulas that conform to predefined rules, is relatively simple to implement, and has a fast matching speed.

[0006] Disadvantages: Due to the diversity and complexity of mathematical formulas, defining comprehensive rules is extremely difficult. Furthermore, this method is poorly adaptable to non-standard formulas or new formula forms, and cannot handle them well.

[0007] 2. Template-based approach: This approach relies primarily on predefined templates to extract formulas that conform to the template structure. This method typically requires experts to conduct in-depth analysis of the formula structure and create refined formula templates.

[0008] Advantages: This method can accurately extract and parse formulas that conform to the template.

[0009] Disadvantages: The template creation process requires expert input, resulting in a large workload and significant time consumption. Furthermore, this method is ineffective at handling formulas that do not conform to the template. Additionally, newly emerging formula formats require template redesign, leading to poor flexibility and adaptability.

[0010] 3. Machine Learning-Based Methods: Machine learning-based methods primarily utilize machine learning algorithms to automatically learn the features and structure of formulas for formula extraction. These methods mostly employ deep learning techniques, such as using Named Entity Recognition (NAME) to identify the location of formulas.

[0011] Advantages: This method can automatically learn the characteristics of formulas without the need for manual definition of rules or templates, can handle more complex formulas, and has a certain generalization ability to new formula forms.

[0012] Disadvantages: This type of method requires a large amount of labeled data for training, and the training process is computationally complex, time-consuming, and resource-intensive. Furthermore, deep learning models often require more precise parameter tuning and optimization.

[0013] In summary, although existing technologies have achieved mathematical formula extraction to a certain extent, some problems and shortcomings remain, such as the accuracy and efficiency of extraction, and the adaptability of the formulas. Therefore, it is essential to research and develop a new method for mathematical formula extraction. Summary of the Invention

[0014] In view of the shortcomings of the prior art, the present invention provides a method and system for mathematical formula recognition and extraction, which can accurately improve the accuracy and adaptability of mathematical formula extraction, and has high extraction efficiency, thereby improving the efficiency of processing large amounts of text.

[0015] In a first aspect, embodiments of the present invention provide a method for mathematical formula recognition and extraction, comprising:

[0016] Obtain the raw text data from the input;

[0017] The original text data is preprocessed to obtain preprocessed text information;

[0018] Mathematical formulas are extracted from the preprocessed text information using a formula recognition and extraction method to obtain the extraction results.

[0019] Output the extraction results.

[0020] Secondly, the present invention provides a system for mathematical formula recognition and extraction, comprising: a data acquisition module, a preprocessing module, a formula extraction module, and a result output module;

[0021] The data acquisition module is used to acquire the input raw text data;

[0022] The preprocessing module is used to preprocess the original text data to obtain preprocessed text information;

[0023] The formula extraction module uses a formula recognition and extraction method to extract mathematical formulas from the preprocessed text information to obtain the extraction results.

[0024] The result output module is used to output the extraction results.

[0025] The beneficial effects of this invention are:

[0026] The method for mathematical formula recognition and extraction provided in this invention has the following significant effects and advantages:

[0027] High accuracy: It can effectively identify and extract mathematical formulas from text, including complex mathematical symbols and structures, greatly improving the accuracy of formula extraction.

[0028] High adaptability: It can handle not only common types and formats of mathematical formulas, but also unseen or non-standard formulas. By introducing machine learning techniques, the model has stronger generalization ability and can adapt to various complex situations.

[0029] High efficiency: The extraction process has been optimized, redundant calculations have been reduced, and extraction efficiency has been effectively improved. Especially when processing large-scale text, it can significantly reduce processing time.

[0030] The present invention provides a system for mathematical formula recognition and extraction, which has the same concept as the above-described method for mathematical formula recognition and extraction. In addition to having the same effect, it also has the advantage of easy expansion: the design has good modularity and can be easily combined with other systems or modules, such as combining with a natural language processing system for in-depth text analysis and processing. Attached Figure Description

[0031] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0032] Figure 1 A flowchart of a method for mathematical formula recognition and extraction provided in the first embodiment of the present invention is shown;

[0033] Figure 2This diagram illustrates a structural block diagram of a mathematical formula recognition and extraction system provided by another embodiment of the present invention;

[0034] Figure 3 It shows Figure 2 The structural block diagram of the preprocessing module and the formula extraction module in the program. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0037] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0038] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0039] As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrases "if determined" or "if [described condition or event] is detected" may be interpreted, depending on the context, as "once determined," "in response to determination," "once [described condition or event] is detected," or "in response to detection of [described condition or event]."

[0040] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application should have the ordinary meaning as understood by those skilled in the art to which this invention pertains.

[0041] like Figure 1 As shown, the first embodiment of the present invention provides a method for mathematical formula recognition and extraction, comprising:

[0042] Obtain the input raw text data, which in this embodiment is the user's answer;

[0043] The original text data is preprocessed to obtain preprocessed text information;

[0044] Mathematical formulas are extracted from the preprocessed text information using a formula recognition and extraction method to obtain the extraction results.

[0045] Output the extraction results.

[0046] Specifically, the preprocessing of raw text data includes text cleaning, formatting, and symbol recognition. The input raw text data is plain text, and the case of Chinese character formulas is not considered. Specific methods for preprocessing raw text data include:

[0047] Segmentation: Combining consecutive numbers or symbols in the original text data into a group to obtain segmented text;

[0048] Merge: Merge the split text, restore the Chinese characters and add "=" in the corresponding positions to obtain the merged data;

[0049] Grouping: Group the formulas in the merged data to determine the relationship between the calculation process and the result in the merged data;

[0050] Symbol conversion: Converting symbols in formulas into mathematical operators that computers can recognize.

[0051] For example: "The equity value of Company B at the beginning of 20×3 = the equity value of Company B at the end of 20×1 + the equity value of Company B at the end of 20×2". In this embodiment, "the equal annuity of Plan A in 20*1" will be used.

[0052] Taking the example of "60.58 / (P / A, 8%, 5) = 60.58 / 3.9927 = 151,733 yuan\nThe equivalent annuity of Plan B in 20X1 years = 83 / (P / A, 10%, 8) = 83 / 5.3349 = 155,579 yuan", each step will be explained to clearly illustrate the preprocessing method of this embodiment. In the following description, "formula example" will be used to replace the example.

[0053] Segmentation. The plain text is segmented according to Chinese characters. Each character in the text is traversed, and consecutive non-Chinese characters are grouped together. Chinese characters are replaced with empty strings. After segmentation, the formula example becomes: "['20*1',",'A',",",",",",",",","'=60.58 / (P / A,8%,5)=60.58 / 3.9927=15.1733',",",",'\n20X1',","'B',",",",",",",",","'=83 / (P / A,10%,8)=83 / 5.3349=15.5579',","]" As you can see, consecutive numbers or symbols in the text are now grouped together.

[0054] Merging. The purpose of merging is to restore the Chinese characters and add "=" in the appropriate positions. Since mathematical formulas often involve units, such as "100 (ten thousand yuan), 10 million, 10 pieces / minute," etc., during the merging process, two pointers are used: one pointing to each element of the string before grouping (CA) (pointer A), and the other pointing to each element of the list after grouping (LB) (pointer B). A new empty string (ConcatS) is initialized to store the merged text. If the element pointed to by pointer B is empty, it means that there is a Chinese character at the corresponding position in CA, and ConcatS needs to concatenate the current character in CA. If the element pointed to by pointer B is not empty, and the element contains valid calculation symbols such as +, -, *, / , and the element does not start with "=", then the element of pointer B needs to have "=" added at the beginning before ConcatS concatenates it with the element. The formula example becomes: "Equal annuity for Plan A (20 * 1 year) = 60.58 / (P / A, 8%, 5) = 60.58 / 3.9927 = 151,733 yuan\nEqual annuity for Plan B (20 * 1 year) = 83 / (P / A, 10%, 8) = 83 / 5.3349 = 155,579 yuan"

[0055] Grouping. User answers often contain many mathematical formulas. When these formulas yield the same result or have the same calculation process, it can lead to confusion in subsequent intelligent grading. Therefore, it's necessary to group the formulas in user answers to clarify the relationship between the calculation process and the result. In grouping, "=" is used as the basis for separating the formula parts, and "\n,;.:" is used as the basis for separating each group. The text is separated using "=", and each segment is normalized using a special formula. Special formulas mainly refer to annuity present value calculation formulas, such as "Calculate (P / F, 10%, 4), (P / F, i, 4) = 1000, calculate i, (P / F, 10%, n), calculate n", etc. In these formulas, "P / F" refers to the annuity present value factor, which has four types: "F / P, P / F, F / A, P / A", each corresponding to a different calculation formula, as shown in Table 1.

[0056] F / P=(1+i)**n P / F = (1 + i)**(-n) F / A=((i+i)**(n-1)) / i P / A=(1-(1+i)**-n) / i

[0057] Table 1. Formulas for Calculating the Present Value of Annuities

[0058] As shown in Table 1, the coefficients need to be converted into calculable formulas first. The normalized text is then divided using "\n,,。;;::". The formula example becomes "[['Equal Annuity of Plan A in 20*1 Year','60.58 / (P / A,8%,5)','60.58 / 3.9927','15.1733 million yuan'],['Equal Annuity of Plan B in 20X1 Year','83 / (P / A,10%,8)','83 / 5.3349','15.5579 million yuan']]". It can be seen that the formulas and their components are divided into several groups.

[0059] Symbol Conversion. Since users input various symbols in their answers, most of which are not directly calculable by computers, these symbols need to be converted into mathematical operators that computers can recognize. ① Replace special symbols such as "\t", "\r", "\xa0", and "\u3000" in the text with empty strings; ② Process percent signs and thousandth signs, replacing symbols with different forms such as "%" and "° / 0" with normal "%", and then with " / 100" and " / 1000"; ③ Replace symbols with different forms such as addition, subtraction, multiplication, and division with normal symbols, such as "÷", "+", "*", "X", "x", "×", "\u031F", "\u1429", and "-", replacing them with "+-* / ", and "^" with "**" to represent exponentiation; ④ Remove the parentheses from operators with parentheses, such as "(÷), (+), (**), and (=); ⑤ Replace various parentheses with English parentheses, such as "()【】[]{}". ⑥ Replace Greek letter symbols with computable forms, such as: "\\(\x08eta\\), \\(\beta\\), β, \delta, Δ", etc. ⑦ Replace LaTeX symbols with basic operators, such as: "\times", "\div", etc.

[0060] After the above preprocessing steps, each non-empty element in the list may be the process or result of a formula. Therefore, after preprocessing, it is necessary to further determine whether each part of the preprocessed text information is a calculation process, calculation result, and unit of a formula.

[0061] Unit Recognition: Units have a significant impact on the results, for example, "1 million yuan" and "1,000,000 yuan". Unit recognition uses a fixed pattern recognition method. Units typically exhibit the following patterns: "1. Following a number, 2. May exist within parentheses, 3. May contain / , 4. The Chinese character part has a maximum of 3 characters". Regular expressions are used to extract segments that match these patterns, and then "ten thousand" and "hundred million" are converted into a calculable numerical form. After unit recognition, it can be seen that the formula examples ['20*1 year', '15.1733 million yuan', '20X1 year', '15.5579 million yuan'] can all be considered to contain units, yielding the first recognition result.

[0062] Calculation process identification: In process identification, the condition of "being able to calculate a result" must be met. If the preprocessed part cannot calculate a result, then that part is not the calculation process of the formula. Using the sympify and evalf methods of the Python scientific computing library sympy, the sympify method can convert the input computable string into a format that the computer can calculate, and the evalf method can perform operations on the input expression and calculate the result. After the above grouping process, the formula example becomes: "[

[0063] ['Equal annuity under Plan A with 20*1 years', '60.58 / (P / A, 8%, 5)',

[0064] '60.58 / 3.9927','151,733 yuan'],

[0065] ['Equal annuity of Plan B in 20X1', '83 / (P / A, 10%, 8)', '83 / 5.3349', '15.5579 million yuan']. Now, taking the first group as an example, we will introduce the specific process of identification in the calculation process.

[0066] First, after identifying the units, we can determine which units are likely to exist and their specific locations. We then extract the expressions preceding the units, considering them as possible calculation processes, such as:

[0067] ['20*1','60.58 / (P / A, 8%, 5)','60.58 / 3.9927','15.1733'].

[0068] Secondly, these expressions are passed to the sympify method, which converts them into a format that the computer can calculate. The result of the sympify method is then passed to the evalf method to calculate the specific value of these expressions.

[0069] Finally, record the results and their positions in each element of the group. For example, '20*1' is located in the first segment '20*1 year Plan A equal annuity' in the first group of the grouped results, so it is recorded as ('20*1', 0). (In computers, 0 is usually used as the marker for the first element).

[0070] Each part is judged and its position in each group is recorded, which is recorded as the second recognition result. For ease of description, the second recognition result is denoted as P = "[[('20*1',0),

[0071] ('60.58 / 3.9927100370780884',1),('60.58 / 3.9927',2),(15.1733,3)],[('20*1',0),('83 / 5.33492619790267',1),('83 / 5.3349',2),

[0072] (15.5579,3)]]”.

[0073] Calculation Result Recognition: Based on process recognition, result recognition assesses each part to determine if it can be converted into a pure number. If so, the part is likely the result, resulting in a third recognition result. This third recognition result is denoted as R = "[[(20.0,'yuan'),(15.17265201765895,None),(15.172690159541162,None),(15.1733,'ten thousand')],[(20.0,'yuan'),(15.557853458709506,None),(15.557929858104183,None),(15.5579,'ten thousand')]]". As you can see, after result recognition, the calculation result and unit are saved separately.

[0074] The calculation process and results are mapped. After the calculation process and results are identified separately, they need to be combined. The result corresponding to a certain process is determined. First, the third identification result is filtered, matching each element in P with each element in R according to their position. If the value at a certain position in P is exactly equal to the corresponding value at the same position in R, then the element at that position in P and R must be the result in the formula. After filtering, R = "[[(15.1733,'ten thousand')],[(15.5579,'ten thousand')]]", where "(20.0,'yuan')" actually represents "20*1 year" in the original text, but it is filtered here because the correspondence between the calculation process and the result cannot be found.

[0075] The final extraction result: The process part of the calculation formula is: [60.58 / 3.9927, 83 / 5.3349], and the result part is: [(15.1733, 'ten thousand'), (15.5579, 'ten thousand')]. The formula extraction process is now complete.

[0076] The present invention provides a novel mathematical formula extraction method with the following significant effects and advantages:

[0077] High accuracy: The embodiments of the present invention can effectively identify and extract mathematical formulas in text, including complex mathematical symbols and structures, and greatly improve the accuracy of formula extraction.

[0078] High adaptability: The embodiments of this invention can handle not only common types and formats of mathematical formulas, but also unseen or non-standard formulas. By introducing machine learning techniques, the model has stronger generalization ability and can adapt to various complex situations.

[0079] High efficiency: This embodiment of the invention optimizes the extraction process, reduces redundant calculations, and effectively improves extraction efficiency. Especially when processing large-scale text, it can significantly reduce processing time.

[0080] In the first embodiment described above, a method for mathematical formula recognition and extraction is provided. Correspondingly, this application also provides a system for mathematical formula recognition and extraction. Please refer to... Figure 2 This is a structural block diagram of a mathematical formula recognition and extraction system provided in the second embodiment of the present invention. Since the device embodiment is basically similar to the method embodiment, it is described simply; relevant details can be found in the description of the method embodiment. The device embodiment described below is merely illustrative.

[0081] like Figure 2 , 3 As shown, another embodiment of the present invention provides a system for mathematical formula recognition and extraction, comprising: a data acquisition module, a preprocessing module, a formula extraction module, and a result output module; the data acquisition module is used to acquire input raw text data; the preprocessing module is used to preprocess the raw text data to obtain preprocessed text information; the formula extraction module uses a formula recognition and extraction method to extract mathematical formulas from the preprocessed text information to obtain extraction results; and the result output module is used to output the extraction results.

[0082] This invention provides a system for mathematical formula recognition and extraction, comprising: a data acquisition module, a preprocessing module, a formula extraction module, and a result output module. The main core modules are the preprocessing module and the formula extraction module. These four modules work together to form the system architecture of this invention, enabling efficient and accurate extraction of mathematical formulas. The functions of each module and the interaction methods between them are briefly described below.

[0083] Preprocessing Module: The main task of this module is to preprocess the raw input data (such as scientific papers, textbooks, web pages, etc.) to facilitate processing by subsequent modules. Preprocessing includes, but is not limited to, text cleaning, formatting, and symbol recognition, resulting in preprocessed text information. The preprocessing module includes a segmentation unit, a merging unit, a grouping unit, and a symbol conversion unit. The data processing methods for the segmentation, merging, grouping, and symbol conversion units are described in the above method embodiments and will not be repeated here.

[0084] Formula extraction module: Utilizing preprocessed text and a formula recognition algorithm, this module identifies and extracts mathematical formulas. It is the core of the entire system, and its effectiveness directly impacts overall system performance. After preprocessing, each non-empty element in the list could be a formula's process or result. Therefore, based on the preprocessing module, the formula extraction module further determines whether each preprocessed part represents a formula's calculation process, result, and units. The formula extraction module includes a unit identification unit, a calculation process identification unit, and a calculation result identification unit. The data processing methods for these units are the same as those for unit identification, calculation process identification, and calculation result identification in the above method embodiments, and will not be repeated here.

[0085] The present invention provides a novel mathematical formula extraction system with the following significant effects and advantages:

[0086] High accuracy: The embodiments of the present invention can effectively identify and extract mathematical formulas in text, including complex mathematical symbols and structures, and greatly improve the accuracy of formula extraction.

[0087] High adaptability: The embodiments of this invention can handle not only common types and formats of mathematical formulas, but also unseen or non-standard formulas. By introducing machine learning techniques, the model has stronger generalization ability and can adapt to various complex situations.

[0088] High efficiency: This embodiment of the invention optimizes the extraction process, reduces redundant calculations, and effectively improves extraction efficiency. Especially when processing large-scale text, it can significantly reduce processing time.

[0089] Easy to expand: The design of this invention has excellent modularity and can be easily combined with other systems or modules, such as natural language processing systems for in-depth text analysis and processing.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for mathematical formula recognition and extraction, characterized in that, include: Obtain the raw text data from the input; The original text data is preprocessed to obtain preprocessed text information; Mathematical formulas are extracted from the preprocessed text information using a formula recognition and extraction method to obtain the extraction results. Output the extraction results; The specific methods for preprocessing the original text data include: The consecutive numbers or symbols in the original text data are combined into a group to obtain the segmented text; The segmented text is merged, Chinese characters are restored, and "=" is added in the corresponding positions to obtain the merged data; Group the formulas in the merged data to determine the relationship between the calculation process and the result in the merged data; Convert the symbols in the formula into mathematical operators that can be recognized by computers; The specific method for merging the segmented text includes: During the merging process, the first pointer points to each element of the string before grouping, and the second pointer points to each element of the list after grouping. A new empty string is initialized to store the merged text. If the element pointed to by the second pointer is empty, it means that there is a Chinese character at the corresponding position in the string before grouping. In this case, the new empty string is concatenated with the current character in the string before grouping. If the element pointed to by the second pointer is not empty, and the element contains a valid calculation symbol and does not start with "=", then "=" is added before the element pointed to by the second pointer, and the new empty string is concatenated with the element.

2. The method as described in claim 1, characterized in that, The specific method for grouping consecutive numbers or symbols in the original text data into a group includes: The original text data is segmented according to Chinese characters. Each character in the text is traversed, and consecutive non-Chinese characters are grouped together. Chinese characters are replaced with empty strings to obtain the segmented text.

3. The method as described in claim 1, characterized in that, The specific method for grouping the formulas in the merged data includes: The merged data is split into multiple segments using "=". If a segment contains coefficients, the coefficients are converted into calculable formulas before being split into several formula groups.

4. The method as described in claim 1, characterized in that, The specific methods for extracting mathematical formulas from preprocessed text information using formula recognition and extraction methods include: The unit, the calculation process, and the calculation result are identified sequentially to obtain the first identification result, the second identification result, and the third identification result, respectively. Map the second and third identification results to find the one-to-one corresponding elements; Extract the calculation process and the calculation result of the formula from the one-to-one corresponding elements.

5. The method as described in claim 4, characterized in that, The specific methods for identifying the calculation results include: Determine whether the second recognition result can be converted into a pure number. If so, it means that this part is a calculation result, and the third recognition result is obtained.

6. The method as described in claim 4, characterized in that, The specific method for mapping the second and third identification results includes: Each element in the second and third identification results is matched one-to-one according to its position; If the value at one position in the second identification result is exactly equal to the value at the corresponding position in the third identification result, it means that the element at the corresponding position in the second and third identification results is the result in the formula. If the value of a position in the second recognition result is not equal to the value of the corresponding position in the third recognition result, it needs to be filtered out.

7. A system for recognizing and extracting mathematical formulas, characterized in that, include: The module includes a data acquisition module, a preprocessing module, a formula extraction module, and a result output module. The data acquisition module is used to acquire the input raw text data; The preprocessing module is used to preprocess the original text data to obtain preprocessed text information; The formula extraction module uses a formula recognition and extraction method to extract mathematical formulas from the preprocessed text information to obtain the extraction results. The result output module is used to output the extraction results; The specific methods for preprocessing the original text data include: The consecutive numbers or symbols in the original text data are combined into a group to obtain the segmented text; The segmented text is merged, Chinese characters are restored, and "=" is added in the corresponding positions to obtain the merged data; Group the formulas in the merged data to determine the relationship between the calculation process and the result in the merged data; Convert the symbols in the formula into mathematical operators that can be recognized by computers; The specific method for merging the segmented text includes: During the merging process, the first pointer points to each element of the string before grouping, and the second pointer points to each element of the list after grouping. A new empty string is initialized to store the merged text. If the element pointed to by the second pointer is empty, it means that there is a Chinese character at the corresponding position in the string before grouping. In this case, the new empty string is concatenated with the current character in the string before grouping. If the element pointed to by the second pointer is not empty, and the element contains a valid calculation symbol and does not start with "=", then "=" is added before the element pointed to by the second pointer, and the new empty string is concatenated with the element.

Images