Auxiliary decision-making system and method based on text natural language understanding

Abstract

The invention discloses an auxiliary decision-making system and method based on text natural language understanding, and relates to the technical field of natural language processing, and the method comprises the steps: extracting three first-class coincident text sets according to three pieces of screened decision-making information, carrying out coincident text comparison on the target extended text information set and text information sets obtained by reversely deducing the three pieces of decision information to obtain three second-class coincident text sets, and carrying out comparative analysis processing on the three second-class coincident text sets and the three first-class coincident text sets to predict a new decision information set I; and performing respective step-by-step segmentation analysis processing on the three pieces of decision information to obtain three pre-processing step-by-step sets, re-integrating to obtain a new decision information set II, and determining corresponding optimal decision information from the three pieces of decision information according to the new decision information set I and the new decision information set II. According to the auxiliary decision-making system and method based on text natural language understanding provided by the invention, the efficiency of analysis decision-making processing work is improved.

Description

Technical Field

[0001] The present application relates to the technical field of natural language processing, and in particular to an auxiliary decision-making system and method based on natural language understanding of text. Background Art

[0002] Intelligent decision-making assistance is a complex process that covers technologies such as big data processing, machine learning, and artificial intelligence. Decision-making assistance based on text natural language understanding (NLU) is of great significance in many fields. It provides support and guidance for various decisions by enabling computers to understand and process human language.

[0003] In traditional technologies, there will be certain misjudgments in the auxiliary decisions generated by machine learning text natural language understanding, but the generated auxiliary decisions are not further verified, resulting in large errors in the final analysis results, reducing the efficiency of text processing and auxiliary decision analysis, and wasting resources. Summary of the invention

[0004] In order to overcome the above-mentioned deficiencies of the prior art, the present application provides a decision support system and method based on natural language understanding of text.

[0005] In a first aspect, the present application provides an auxiliary decision-making method based on natural language understanding of text, the method comprising:

[0006] Obtain target text information, and according to the target text information, screen out first decision information, second decision information, and third decision information, and according to the three types of decision information, respectively extract the overlapping text parts between the text information sets reversely derived from the first decision information, the second decision information, and the third decision information to obtain first-category overlapping text set 1, first-category overlapping text set 2, and first-category overlapping text set 3;

[0007] Performing overlapping text comparison on the target extended text information set obtained by performing text expansion on the target text information and the text information sets obtained by inversely deducing the first decision information, the second decision information and the third decision information to obtain second category overlapping text set one, second category overlapping text set two and second category overlapping text set three;

[0008] The second category overlap text set 1, the second category overlap text set 2 and the second category overlap text set 3 are compared with the first category overlap text set 1, the first category overlap text set 2 and the first category overlap text set 3 under various circumstances, and the overlap and difference parts are compared and statistically processed to predict the new decision information set 1;

[0009] The first decision information, the second decision information and the third decision information are processed by respective steps to obtain a first sub-step set, a second sub-step set and a third sub-step set, and a first pre-processing sub-step set, a second pre-processing sub-step set and a third pre-processing sub-step set are respectively selected from the first sub-step set, the second sub-step set and the third sub-step set to reintegrate to obtain a new decision information set 2;

[0010] The new decision information set one is respectively compared with the first decision information, the second decision information and the third decision information for similarity to obtain first-category similarity data one, first-category similarity data two and first-category similarity data three; the new decision information set two is respectively compared with the first decision information, the second decision information and the third decision information for similarity to obtain second-category similarity data one, second-category similarity data two and second-category similarity data three; the two categories and three types of similarity data are subjected to corresponding data comprehensive processing to determine the corresponding optimal decision information from the first decision information, the second decision information and the third decision information, and output the auxiliary decision analysis result.

[0011] Preferably, historical text record information is obtained, and a decision prediction model is trained based on the historical text record information;

[0012] Obtaining target text information, inputting the target text information into a decision prediction model for testing to obtain multiple decision information, prioritizing the multiple decision information, and outputting a decision information set;

[0013] Filter out the first three decision information with priority from the decision information set to obtain first decision information, second decision information and third decision information;

[0014] Inferring corresponding multiple text information according to the first decision information, and outputting a first text information set;

[0015] Inferring corresponding multiple text information according to the second decision information, and outputting a second text information set;

[0016] Inferring corresponding multiple text information according to the third decision information, and outputting a third text information set;

[0017] Extracting the overlapping text information parts between the first text information set and the second text information set, and outputting a first category of overlapping text set 1;

[0018] Extracting the overlapping text information parts between the first text information set and the third text information set, and outputting the first category overlapping text set 2;

[0019] The overlapping text information portion between the second text information set and the third text information set is extracted, and the first category overlapping text set three is output.

[0020] Preferably, text expansion is performed on the target text information to obtain a target expanded text information set;

[0021] Extracting the overlapping text information part between the target extended text information set and the first text information set, and outputting the second category overlapping text set 1;

[0022] Extracting the overlapping text information part between the target extended text information set and the second text information set, and outputting the second category overlapping text set 2;

[0023] The overlapping text information part between the target extended text information set and the third text information set is extracted, and the second category overlapping text set three is output.

[0024] Preferably, the second type of overlapping text set 1 is compared with the first type of overlapping text set 1 to respectively select the overlapping text parts and the difference text parts, and output the overlapping comparison text set 1 and the difference comparison text set 1;

[0025] Compare the second type of overlapping text set 2 with the first type of overlapping text set 1 to respectively select the overlapping text parts and the difference text parts, and output the overlapping comparison text set 2 and the difference comparison text set 2;

[0026] Compare the second type of overlapping text set 1 with the first type of overlapping text set 2 to respectively select the overlapping text parts and the difference text parts, and output the overlapping comparison text set 3 and the difference comparison text set 3;

[0027] Compare the second type of overlapping text set three with the first type of overlapping text set two to respectively select the overlapping text parts and the difference text parts, and output the overlapping comparison text set four and the difference comparison text set four;

[0028] Compare the second type of overlapping text set 2 with the first type of overlapping text set 3 to respectively select the overlapping text parts and the difference text parts, and output the overlapping comparison text set 5 and the difference comparison text set 5;

[0029] Compare the second type of overlapping text set three with the first type of overlapping text set three to respectively select the overlapping text parts and the difference text parts, and output the overlapping comparison text set six and the difference comparison text set six;

[0030] Integrate the overlapping comparison text set 1, the overlapping comparison text set 2, the overlapping comparison text set 3, the overlapping comparison text set 4, the overlapping comparison text set 5 and the overlapping comparison text set 6 to obtain a total overlapping comparison text set;

[0031] Extracting the overlapping difference text parts among the difference comparison text set 1, the difference comparison text set 2, the difference comparison text set 3, the difference comparison text set 4, the difference comparison text set 5 and the difference comparison text set 6, and outputting the difference overlapping text set;

[0032] The total overlap comparison text set and the difference overlap text set are input into the decision prediction model for testing, and a new decision information set 1 is output.

[0033] Preferably, the first decision information is divided into a plurality of sub-steps, and a first sub-step set is output;

[0034] dividing the second decision information into a plurality of sub-steps and outputting a second sub-step set;

[0035] Dividing the third decision information into a plurality of sub-steps and outputting a third sub-step set;

[0036] Performing text inversion on each sub-step in the first sub-step set, the second sub-step set, and the third sub-step set to obtain a first sub-step text set, a second sub-step text set, and a third sub-step text set;

[0037] Extracting the overlapping text parts among the overlapping comparison text set 1, the overlapping comparison text set 2, the overlapping comparison text set 3, the overlapping comparison text set 4, the overlapping comparison text set 5 and the overlapping comparison text set 6, and outputting the common overlapping text set;

[0038] According to the co-overlapping text set, extract the sub-steps with the most overlapping text parts with the co-overlapping text set from the first sub-step set, the second sub-step set and the third sub-step set respectively, and output the first pre-processing sub-step set, the second pre-processing sub-step set and the third pre-processing sub-step set;

[0039] The first preprocessing sub-step set, the second preprocessing sub-step set and the third preprocessing sub-step set are reintegrated to obtain a new decision information set two.

[0040] Preferably, each new decision information in the new decision information set 1 is respectively subjected to a similarity comparison with the first decision information, the second decision information and the third decision information, and the data results are comprehensively counted to obtain first-category similarity data 1, first-category similarity data 2 and first-category similarity data 3;

[0041] The new decision information in the new decision information set 2 is respectively compared with the first decision information, the second decision information and the third decision information to obtain the second similarity data 1, the second similarity data 2 and the second similarity data 3 by comprehensive statistics of the data results;

[0042] The first type of similarity data 1 and the second type of similarity data 1 are averaged to obtain a first comprehensive similarity;

[0043] The first similarity data 2 and the second similarity data 2 are averaged to obtain a second comprehensive similarity;

[0044] The first similarity data 3 and the second similarity data 3 are averaged to obtain a third comprehensive similarity;

[0045] The maximum comprehensive similarity is extracted from the first comprehensive similarity, the second comprehensive similarity and the third comprehensive similarity, and the corresponding optimal decision information is determined from the first decision information, the second decision information and the third decision information according to the maximum comprehensive similarity, and the auxiliary decision analysis result is output.

[0046] In the second aspect, an auxiliary decision-making system based on natural language understanding of text includes:

[0047] The first text information analysis unit is used to obtain target text information, and according to the target text information, screen out the first decision information, the second decision information and the third decision information, and according to the three types of decision information, respectively extract the overlapping text parts between the text information sets obtained by inverting the first decision information, the second decision information and the third decision information to obtain the first category overlapping text set 1, the first category overlapping text set 2 and the first category overlapping text set 3;

[0048] The second text information analysis unit is used to perform overlapping text comparison on the target extended text information set obtained by performing text expansion on the target text information and the text information sets obtained by inferring the first decision information, the second decision information and the third decision information to obtain the second category overlapping text set one, the second category overlapping text set two and the second category overlapping text set three;

[0049] A new decision prediction unit 1 is used to perform statistical processing on the results of comparing the overlapped and different parts of the second category overlapped text set 1, the second category overlapped text set 2 and the second category overlapped text set 3 with the first category overlapped text set 1, the first category overlapped text set 2 and the first category overlapped text set 3 under various circumstances, so as to predict a new decision information set 1;

[0050] A second new decision prediction unit is used to perform step-by-step segmentation processing on the first decision information, the second decision information and the third decision information to obtain a first sub-step set, a second sub-step set and a third sub-step set, and to select a first pre-processing sub-step set, a second pre-processing sub-step set and a third pre-processing sub-step set from the first sub-step set, the second sub-step set and the third sub-step set, respectively, to reintegrate to obtain a second new decision information set;

[0051] The decision optimization unit is used to perform similarity comparison between the new decision information set 1 and the first decision information, the second decision information and the third decision information to obtain first-class similarity data 1, first-class similarity data 2 and first-class similarity data 3, perform similarity comparison between the new decision information set 2 and the first decision information, the second decision information and the third decision information to obtain second-class similarity data 1, second-class similarity data 2 and second-class similarity data 3, perform corresponding data comprehensive processing on the two types of three similarity data, so as to determine the corresponding optimal decision information from the first decision information, the second decision information and the third decision information, and output the auxiliary decision analysis result.

[0052] Compared with the prior art, the present invention has the following characteristics and beneficial effects:

[0053] By first using traditional technology to analyze the corresponding multiple decision information according to the target text information, three relatively optimal decision information are screened out from the multiple decision information, namely the first decision information, the second decision information and the third decision information, so as to facilitate the subsequent further verification and analysis of the three decision information to select the best decision information, so as to avoid the traditional technology of directly randomly selecting one decision information from multiple decision information for use, which increases the probability of misjudgment, leads to waste of resources and low efficiency of the entire data processing work, that is, by reversely inferring the text information of the three decision information, so as to statistically analyze the overlapping text parts of the three text information sets, so as to compare the comparison results obtained with the target extended text information set obtained by the subsequent expansion of the target text information and the overlapping text parts between the three text information sets, so as to statistically analyze the text information that can be extended. The characteristic law of continuity is conducive to improving the accuracy of subsequent decision information verification analysis and processing, and the new decision information one is predicted through the above processing steps, which is used as one of the objects for comparison and analysis of the above three decision information. Furthermore, by segmenting the multiple sub-steps of each of the above three decision information, three sub-step sets are obtained, and the text information of each of the multiple sub-steps is reversed, that is, three sub-step text sets are obtained, combined with the comparison results between the multiple overlapping text information sets in the above-mentioned processing process of predicting the new decision information one, three pre-processing sub-step sets are screened out from the three sub-step text sets, and integrated to obtain new decision information set two, the new decision information set one and the new decision information set two are respectively compared with the above three decision information for similarity and the comparison results are comprehensively counted, so as to select the best one from the above three decision information, and the resource utilization and the reliability of decision selection are improved through the above analysis process. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1It is a flowchart of a step-by-step decision-making method based on natural language understanding of text, which is mainly embodied in this embodiment.

[0055] Figure 2 It is a structural block diagram of an auxiliary decision-making system based on natural language understanding of text, which is mainly embodied in this embodiment. DETAILED DESCRIPTION

[0056] The present invention is further described in detail below in conjunction with the following examples.

[0057] Reference Figure 1 , an auxiliary decision-making system and method based on natural language understanding of text, the method comprising the following steps:

[0058] S1. Obtain target text information, and based on the target text information, screen out the first decision information, the second decision information, and the third decision information, and based on these three types of decision information, respectively extract the overlapping text parts between the text information sets reversely derived from the first decision information, the second decision information, and the third decision information to obtain the first category overlapping text set one, the first category overlapping text set two, and the first category overlapping text set three.

[0059] S2. Perform overlapping text comparison on the target extended text information set obtained by text expansion of the target text information and the text information sets inferred from the first decision information, the second decision information and the third decision information to obtain second category overlapping text set one, second category overlapping text set two and second category overlapping text set three.

[0060] S3. Perform statistical processing on the results of comparing the overlapping and different parts of the second category overlapping text set one, the second category overlapping text set two and the second category overlapping text set three with the first category overlapping text set one, the first category overlapping text set two and the first category overlapping text set three in various situations, so as to predict the new decision information set one.

[0061] S4. Divide the first decision information, the second decision information and the third decision information into their own steps to obtain a first sub-step set, a second sub-step set and a third sub-step set, and respectively select a first pre-processing sub-step set, a second pre-processing sub-step set and a third pre-processing sub-step set from the first sub-step set, the second sub-step set and the third sub-step set to reintegrate them to obtain a new decision information set two.

[0062] S5. Perform similarity comparison on the new decision information set one with the first decision information, the second decision information and the third decision information to obtain first-category similarity data one, first-category similarity data two and first-category similarity data three; perform similarity comparison on the new decision information set two with the first decision information, the second decision information and the third decision information to obtain second-category similarity data one, second-category similarity data two and second-category similarity data three; perform corresponding data comprehensive processing on these two categories and three types of similarity data to determine the corresponding optimal decision information from the first decision information, the second decision information and the third decision information, and output the auxiliary decision analysis result.

[0063] Specifically, by first using traditional technology to analyze the corresponding multiple decision information according to the target text information, three relatively optimal decision information are screened out from the multiple decision information, namely the first decision information, the second decision information and the third decision information, so as to facilitate the subsequent further verification and analysis of the three decision information to select the best decision information, so as to avoid the traditional technology of directly randomly selecting one decision information from multiple decision information for use, which increases the probability of misjudgment, leads to waste of resources and low efficiency of the entire data processing work, that is, by reversely inferring the text information of the three decision information, so as to statistically analyze the overlapping text parts of the three text information sets, so as to compare the comparison results obtained with the target extended text information set obtained by subsequently expanding the target text information and the overlapping text parts between the three text information sets, so as to statistically analyze the text information that can be found. The characteristic law of continuity is beneficial to improving the accuracy of subsequent decision information verification analysis and processing, and the new decision information one is predicted through the above processing steps, which is used as one of the objects for comparison and analysis of the above three decision information. Furthermore, by segmenting the multiple sub-steps of each of the above three decision information, three sub-step sets are obtained, and the text information of each of the multiple sub-steps is reversed, that is, three sub-step text sets are obtained, combined with the comparison results between the multiple overlapping text information sets in the above-mentioned processing process of predicting the new decision information one, three pre-processing sub-step sets are screened out from the three sub-step text sets, and integrated to obtain new decision information set two, the new decision information set one and the new decision information set two are respectively compared with the above three decision information for similarity and the comparison results are comprehensively counted, so as to select the best one from the above three decision information, and the resource utilization and the reliability of decision selection are improved through the above analysis process.

[0064] The specific step S1 includes the following sub-steps:

[0065] Obtain historical text record information and train a decision prediction model based on the historical text record information.

[0066] Obtain target text information, input the target text information into a decision prediction model for testing to obtain multiple decision information, prioritize the multiple decision information, and output a decision information set.

[0067] The first three decision information with priority are screened out from the decision information set to obtain the first decision information, the second decision information and the third decision information.

[0068] Based on the first decision information, corresponding multiple text information are inferred and a first text information set is output.

[0069] According to the second decision information, the corresponding multiple text information are inferred and a second text information set is output.

[0070] According to the third decision information, the corresponding multiple text information are inferred and a third text information set is output.

[0071] The overlapping text information part between the first text information set and the second text information set is extracted, and a first category overlapping text set 1 is output.

[0072] The overlapping text information part between the first text information set and the third text information set is extracted, and the first category overlapping text set 2 is output.

[0073] The overlapping text information part between the second text information set and the third text information set is extracted, and the first category overlapping text set three is output.

[0074] Specifically, for example, training a decision prediction model: combining natural language processing (NLP) and machine learning, selecting appropriate machine learning algorithms (such as naive Bayes, support vector machine, random forest, deep learning model, etc.) to train the model. Use historical text data as a training set, and optimize performance by adjusting model parameters. Decision information set (i.e., directly input the acquired target text information, if it is text information A, into the trained decision prediction model for feature testing to match the corresponding decision information, and the decision information here is most likely multiple, i.e., one-to-many, and prioritize multiple decision information, i.e., prioritize the one that best meets the semantic understanding of the target text information), and select the three decision information with the highest ranking, i.e., the first decision information, the second decision information, and the third decision information. For example, if the target text information is "how to solve a quadratic equation", the corresponding decision information includes: formula method, matching method, factorization method, and Vieta theorem. The first text information set (that is, one decision information can be inferred from multiple targeted text information. The multiple text information inferred here may include the target text information or may not include it, but may include other text information similar to the target text information. If the first text information set contains: m1, m2, m3, m4, o5, m6), the second text information set: m1, n2, m3, m4, o5, o6 and the third text information set: o1, n2, m3, m4, o5, o6 (the explanation is the same as the first text information set, so no further explanation is given here), the first category overlapping text set one (such as m1, m3, m4), the first category overlapping text set two (m3, m4), the first category overlapping text set three (m4, o5, o6).

[0075] The specific step S2 includes the following sub-steps:

[0076] Perform text expansion on the target text information to obtain a target expanded text information set.

[0077] The overlapping text information part between the target extended text information set and the first text information set is extracted, and the second category overlapping text set one is output.

[0078] The overlapping text information part between the target extended text information set and the second text information set is extracted, and the second category overlapping text set 2 is output.

[0079] The overlapping text information part between the target extended text information set and the third text information set is extracted, and the second category overlapping text set three is output.

[0080] Specifically, such as the target extended text information set (for example, the target text information "how to solve a quadratic equation" mentioned above can be expanded to the following types of text information: the definition of a quadratic equation, a summary of solutions to a quadratic equation, a detailed explanation of the formula method, the application of the matching method, the direct square root method, the factorization method, the Vieta theorem, the historical background, the practical application, common misunderstandings and precautions; if expressed in symbols: the other text information extended by the target text information A is A1, A2, A3, A4, A5, A6, A7, A8, A9, A10), the second category overlapping text set one (if it is m3, m4, o5, m6, it means that the four text information m3, m4, o5, m6 are the same as some of the extended text information in the extended text information of A1-A10), the second category overlapping text set two and the second category overlapping text set three (the same explanation as the second category overlapping text set one, if they are n2, m3, m4, o5; m3, m4, o5, o6 respectively).

[0081] The specific step S3 includes the following sub-steps:

[0082] The second category overlapping text set 1 is compared with the first category overlapping text set 1 to respectively filter out the overlapping text parts and the difference text parts, and the overlapping comparison text set 1 and the difference comparison text set 1 are output.

[0083] The second category overlapping text set 2 is compared with the first overlapping text set 1 to respectively select the overlapping text parts and the difference text parts, and the overlapping comparison text set 2 and the difference comparison text set 2 are output.

[0084] The second category overlapping text set one and the first overlapping text set two are compared to respectively select the overlapping text parts and the difference text parts, and output the overlapping comparison text set three and the difference comparison text set three.

[0085] The second category overlapping text set three is compared with the first overlapping text set two to respectively filter out the overlapping text parts and the difference text parts, and the overlapping comparison text set four and the difference comparison text set four are output.

[0086] The second category overlapping text set two and the first overlapping text set three are compared to respectively screen out the overlapping text parts and the different text parts, and the overlapping comparison text set five and the different comparison text set five are output.

[0087] The second category overlapping text set three is compared with the first category overlapping text set three to respectively filter out the overlapping text parts and the difference text parts, and the overlapping comparison text set six and the difference comparison text set six are output.

[0088] The overlapping comparison text set one, the overlapping comparison text set two, the overlapping comparison text set three, the overlapping comparison text set four, the overlapping comparison text set five and the overlapping comparison text set six are integrated to obtain the total overlapping comparison text set.

[0089] The overlapping difference text parts among difference comparison text set 1, difference comparison text set 2, difference comparison text set 3, difference comparison text set 4, difference comparison text set 5 and difference comparison text set 6 are extracted, and the difference overlapping text set is output.

[0090] The total overlap comparison text set and the difference overlap text set are input into the decision prediction model for testing, and a new decision information set 1 is output.

[0091] Specifically, such as the overlapping comparison text set 1 (m3, m4) and the difference comparison text set 1 (m1, o5, m6), the overlapping comparison text set 2 (m3, m4) and the difference comparison text set 2 (m1, n2, o5), the overlapping comparison text set 3 (m3, m4) and the difference comparison text set 3 (o5, m6), the overlapping comparison text set 4 (m3, m4) and the difference comparison text set 4 (o5, o6), the overlapping comparison text set 5 (m4, o5) and the difference comparison text set 5 (n2, m3, o6), the overlapping comparison text set 6 (m4, o5, o6) and the difference comparison text set 6 (m3), the total Overlapping comparison text set (m3, m4, o5, o6), difference overlapping text set (o5, it should be noted that when not all difference comparison text sets have the same difference text part, the difference text part with the most identical data is extracted, and the difference overlapping text may have the same difference text part as the total overlapping comparison text), new decision information set one (if it is M, the two feature information sets of the extracted total overlapping comparison text set and the difference overlapping text set are input into the model for feature testing and matching to obtain corresponding multiple new decision information, and M includes other new decision information: M1, M2, M3, M4, M5, M6).

[0092] The specific step S4 includes the following sub-steps:

[0093] The first decision information is divided into a plurality of sub-steps, and a first sub-step set is output.

[0094] The second decision information is divided into a plurality of sub-steps, and a second sub-step set is output.

[0095] The third decision information is divided into a plurality of sub-steps, and a third sub-step set is output.

[0096] The text of each sub-step in the first sub-step set, the second sub-step set and the third sub-step set is reversed to obtain the first sub-step text set, the second sub-step text set and the third sub-step text set.

[0097] The overlapping text parts among the overlapping comparison text set 1, the overlapping comparison text set 2, the overlapping comparison text set 3, the overlapping comparison text set 4, the overlapping comparison text set 5 and the overlapping comparison text set 6 are extracted, and the common overlapping text set is output.

[0098] According to the co-overlapping text set, the sub-steps having the most overlapping text parts with the co-overlapping text set are extracted from the first sub-step set, the second sub-step set and the third sub-step set respectively, and the first preprocessing sub-step set, the second preprocessing sub-step set and the third preprocessing sub-step set are output.

[0099] The first preprocessing sub-step set, the second preprocessing sub-step set and the third preprocessing sub-step set are reintegrated to obtain a new decision information set two.

[0100] Specifically, such as the first sub-step set (if the first decision information is X, the sub-steps included are: x1, x2, x3, x4, x5), the second sub-step set (if the second decision information is Y, the sub-steps included are: y1, y2, y3, y4, y5, y6, y7), the third sub-step set (if the third decision information is Z, the sub-steps included are: z1, z2, z3, z4, z5, z6, z7, z8, z9), the first sub-step text set, the second sub-step text set and the third sub-step text set (taking the first sub-step text set as an example, and taking the sub-steps x1 and x2 as an example, x1: the multiple text information corresponding to the sub-push are t1, t2, m3, m4;

[0101] x2: Multiple text information corresponding to the split push are r1, t2, m3, r4, r5), the overlapping text set (m3, m4), the first preprocessing sub-step (then it is sub-step x1), the second preprocessing sub-step and the third preprocessing sub-step (all have the same explanation as the first preprocessing sub-step, no further explanation here, if they are: y3, y4; z5, z8, z9), new decision information set two (if it is N, then one of the new decision information sub-steps is composed of x1, y3, y4; z5, z8, z9, N contains other new decision information: N1, N2, N3, N4).

[0102] The specific step S5 includes the following sub-steps:

[0103] The similarity comparison data results of each new decision information in the new decision information set one with the first decision information, the second decision information and the third decision information are comprehensively counted to obtain the first category similarity data one, the first category similarity data two and the first category similarity data three.

[0104] The similarity comparison data results of each new decision information in the new decision information set 2 and the first decision information, the second decision information and the third decision information are respectively statistically analyzed to obtain the second category similarity data 1, the second category similarity data 2 and the second category similarity data 3.

[0105] The first type of similarity data 1 and the second type of similarity data 1 are averaged to obtain a first comprehensive similarity.

[0106] The second comprehensive similarity is obtained by averaging the first type similarity data 2 and the second type similarity data 2.

[0107] The first type of similarity data three and the second type of similarity data three are averaged to obtain a third comprehensive similarity.

[0108] The maximum comprehensive similarity is extracted from the first comprehensive similarity, the second comprehensive similarity and the third comprehensive similarity, and the corresponding optimal decision information is determined from the first decision information, the second decision information and the third decision information according to the maximum comprehensive similarity, and the auxiliary decision analysis result is output.

[0109] Specifically, such as the first type of similarity data one, the first type of similarity data two and the first type of similarity data three (taking the first type of similarity data one as an example for main explanation: the similarity comparison here refers to the similarity of the overall decision implementation process, if the similarities of M1, M2, M3, M4, M5, and M6 to the first decision information are respectively 80%, 60%, 60%, 70%, 90%, and 80%), the second type of similarity data one, the second type of similarity data two and the second type of similarity data three (similarly, the second type of similarity data one is explained by example: if the similarities of N1, N2, N3, and N4 to the first decision information are respectively 80%, 80%, 60%, and 85%), the first comprehensive similarity (then it is 74.5%), the second comprehensive similarity (if it is 70%), the third comprehensive similarity (if it is 80%), the third comprehensive similarity is the maximum comprehensive similarity, and the corresponding third decision information is judged to be the optimal decision information.

[0110] An auxiliary decision-making system based on natural language understanding of text, by applying the above-mentioned auxiliary decision-making method based on natural language understanding of text, including a text information analysis unit 1, a text information analysis unit 2, a new decision prediction unit 1, a new decision prediction unit 2 and a decision optimization unit, referring to Figure 2By, a text information analysis unit 1 obtains target text information, and according to the target text information, screens out the first decision information, the second decision information and the third decision information, and according to the three types of decision information, extracts the overlapping text parts between the text information sets inferred from the first decision information, the second decision information and the third decision information, respectively, to obtain the first category overlapping text set 1, the first category overlapping text set 2 and the first category overlapping text set 3; by a text information analysis unit 2 performing text expansion on the target text information to obtain the target extended text information set, and respectively performing overlapping text comparison with the text information sets inferred from the first decision information, the second decision information and the third decision information, to obtain the second category overlapping text set 1, the second category overlapping text set 2 and the second category overlapping text set 3; by a new decision prediction unit 1, the second category overlapping text set 1, the second category overlapping text set 2 and the second category overlapping text set 3 are compared with the first category overlapping text set 1, the first category overlapping text set 2 and the first category overlapping text set 3 under multiple circumstances, and the results of statistical processing are performed to predict a new decision information set 1. ; The first decision information, the second decision information and the third decision information are divided and processed in steps by the new decision prediction unit 2 to obtain the first sub-step set, the second sub-step set and the third sub-step set, and the first pre-processing sub-step set, the second pre-processing sub-step set and the third pre-processing sub-step set are respectively screened out from the first sub-step set, the second sub-step set and the third sub-step set to reintegrate to obtain the new decision information set 2; The new decision information set 1 is respectively compared with the first decision information, the second decision information and the third decision information by the decision optimization unit to obtain the first type of similarity data 1, the first type of similarity data 2 and the first type of similarity data 3, and the new decision information set 2 is respectively compared with the first decision information, the second decision information and the third decision information to obtain the second type of similarity data 1, the second type of similarity data 2 and the second type of similarity data 3, and the two types of three similarity data are subjected to corresponding data comprehensive processing to determine the corresponding optimal decision information from the first decision information, the second decision information and the third decision information, and output the auxiliary decision analysis result.

[0111] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A decision-making assistance method based on natural language understanding of text, characterized in that: The following steps are involved: Obtain target text information, and according to the target text information, screen out first decision information, second decision information, and third decision information, and according to the three types of decision information, respectively extract the overlapping text parts between the text information sets reversely derived from the first decision information, the second decision information, and the third decision information to obtain first-category overlapping text set 1, first-category overlapping text set 2, and first-category overlapping text set 3; Performing overlapping text comparison on the target extended text information set obtained by performing text expansion on the target text information and the text information sets obtained by inversely deducing the first decision information, the second decision information and the third decision information to obtain second category overlapping text set one, second category overlapping text set two and second category overlapping text set three; The second category overlap text set 1, the second category overlap text set 2 and the second category overlap text set 3 are compared with the first category overlap text set 1, the first category overlap text set 2 and the first category overlap text set 3 under various circumstances, and the overlap and difference parts are compared and statistically processed to predict the new decision information set 1; The first decision information, the second decision information and the third decision information are processed by respective steps to obtain a first sub-step set, a second sub-step set and a third sub-step set, and a first pre-processing sub-step set, a second pre-processing sub-step set and a third pre-processing sub-step set are respectively selected from the first sub-step set, the second sub-step set and the third sub-step set to reintegrate to obtain a new decision information set 2; The new decision information set one is respectively compared with the first decision information, the second decision information and the third decision information for similarity to obtain first-category similarity data one, first-category similarity data two and first-category similarity data three; the new decision information set two is respectively compared with the first decision information, the second decision information and the third decision information for similarity to obtain second-category similarity data one, second-category similarity data two and second-category similarity data three; the two categories and three types of similarity data are subjected to corresponding data comprehensive processing to determine the corresponding optimal decision information from the first decision information, the second decision information and the third decision information, and output the auxiliary decision analysis result.

2. The auxiliary decision-making method based on text natural language understanding according to claim 1, characterized in that: The steps of obtaining target text information, screening out first decision information, second decision information and third decision information according to the target text information, respectively extracting the overlapping text parts between the text information sets inferred from the first decision information, the second decision information and the third decision information according to the three types of decision information to obtain first category overlapping text set 1, first category overlapping text set 2 and first category overlapping text set 3 are specifically as follows: Acquire historical text record information, and train a decision prediction model based on the historical text record information; Obtaining target text information, inputting the target text information into a decision prediction model for testing to obtain multiple decision information, prioritizing the multiple decision information, and outputting a decision information set; Filter out the first three decision information with priority from the decision information set to obtain first decision information, second decision information and third decision information; Inferring corresponding multiple text information according to the first decision information, and outputting a first text information set; Inferring corresponding multiple text information according to the second decision information, and outputting a second text information set; Inferring corresponding multiple text information according to the third decision information, and outputting a third text information set; Extracting the overlapping text information parts between the first text information set and the second text information set, and outputting a first category of overlapping text set 1; Extracting the overlapping text information parts between the first text information set and the third text information set, and outputting the first category overlapping text set 2; The overlapping text information portion between the second text information set and the third text information set is extracted, and the first category overlapping text set three is output.

3. The auxiliary decision-making method based on natural language understanding of text according to claim 2, characterized in that: The target extended text information set obtained by performing text expansion on the target text information is compared with the text information sets obtained by inversely deducing the first decision information, the second decision information, and the third decision information to obtain the second category overlapping text set 1, the second category overlapping text set 2, and the second category overlapping text set 3, specifically comprising: Performing text expansion on the target text information to obtain a target expanded text information set; Extracting the overlapping text information part between the target extended text information set and the first text information set, and outputting the second category overlapping text set 1; Extracting the overlapping text information part between the target extended text information set and the second text information set, and outputting the second category overlapping text set 2; The overlapping text information part between the target extended text information set and the third text information set is extracted, and the second category overlapping text set three is output.

4. The auxiliary decision-making method based on natural language understanding of text according to claim 3, characterized in that: The steps of comparing the overlapped and different parts of the second category overlapped text set 1, the second category overlapped text set 2 and the second category overlapped text set 3 with the first category overlapped text set 1, the first category overlapped text set 2 and the first category overlapped text set 3 under various circumstances and predicting the new decision information set 1 are as follows: Compare the second category overlapped text set 1 with the first category overlapped text set 1 to respectively select the overlapped text parts and the difference text parts, and output the overlapped comparison text set 1 and the difference comparison text set 1; Compare the second type of overlapping text set 2 with the first type of overlapping text set 1 to respectively select the overlapping text parts and the difference text parts, and output the overlapping comparison text set 2 and the difference comparison text set 2; Compare the second type of overlapping text set 1 with the first type of overlapping text set 2 to respectively select the overlapping text parts and the difference text parts, and output the overlapping comparison text set 3 and the difference comparison text set 3; Compare the second type of overlapping text set three with the first type of overlapping text set two to respectively select the overlapping text parts and the difference text parts, and output the overlapping comparison text set four and the difference comparison text set four; Compare the second type of overlapping text set 2 with the first type of overlapping text set 3 to respectively select the overlapping text parts and the difference text parts, and output the overlapping comparison text set 5 and the difference comparison text set 5; Compare the second type of overlapping text set three with the first type of overlapping text set three to respectively select the overlapping text parts and the difference text parts, and output the overlapping comparison text set six and the difference comparison text set six; Integrate the overlapping comparison text set 1, the overlapping comparison text set 2, the overlapping comparison text set 3, the overlapping comparison text set 4, the overlapping comparison text set 5 and the overlapping comparison text set 6 to obtain a total overlapping comparison text set; Extracting the overlapping difference text parts among the difference comparison text set 1, the difference comparison text set 2, the difference comparison text set 3, the difference comparison text set 4, the difference comparison text set 5 and the difference comparison text set 6, and outputting the difference overlapping text set; The total overlap comparison text set and the difference overlap text set are input into the decision prediction model for testing, and a new decision information set 1 is output.

5. The auxiliary decision-making method based on text natural language understanding according to claim 4, characterized in that: The first decision information, the second decision information and the third decision information are processed in steps to obtain a first sub-step set, a second sub-step set and a third sub-step set, and a first pre-processing sub-step set, a second pre-processing sub-step set and a third pre-processing sub-step set are respectively selected from the first sub-step set, the second sub-step set and the third sub-step set to reintegrate to obtain a new decision information set 2, specifically: Dividing the first decision information into a plurality of sub-steps and outputting a first sub-step set; dividing the second decision information into a plurality of sub-steps and outputting a second sub-step set; Dividing the third decision information into a plurality of sub-steps and outputting a third sub-step set; Performing text inversion on each sub-step in the first sub-step set, the second sub-step set, and the third sub-step set to obtain a first sub-step text set, a second sub-step text set, and a third sub-step text set; Extracting the overlapping text parts among the overlapping comparison text set 1, the overlapping comparison text set 2, the overlapping comparison text set 3, the overlapping comparison text set 4, the overlapping comparison text set 5 and the overlapping comparison text set 6, and outputting the common overlapping text set; According to the co-overlapping text set, extract the sub-steps with the most overlapping text parts with the co-overlapping text set from the first sub-step set, the second sub-step set and the third sub-step set respectively, and output the first pre-processing sub-step set, the second pre-processing sub-step set and the third pre-processing sub-step set; The first preprocessing sub-step set, the second preprocessing sub-step set and the third preprocessing sub-step set are reintegrated to obtain a new decision information set two.

6. The auxiliary decision-making method based on natural language understanding of text according to claim 5, characterized in that: The steps of performing a similarity comparison between the new decision information set 1 and the first decision information, the second decision information and the third decision information to obtain first-category similarity data 1, first-category similarity data 2 and first-category similarity data 3, performing a similarity comparison between the new decision information set 2 and the first decision information, the second decision information and the third decision information to obtain second-category similarity data 1, second-category similarity data 2 and second-category similarity data 3, performing corresponding data comprehensive processing on the two categories and three types of similarity data to determine the corresponding optimal decision information from the first decision information, the second decision information and the third decision information, and outputting the auxiliary decision analysis result are specifically as follows: Performing a comprehensive statistical analysis of the similarity comparison data results of each new decision information in the new decision information set 1 with the first decision information, the second decision information and the third decision information to obtain first-category similarity data 1, first-category similarity data 2 and first-category similarity data 3; The new decision information in the new decision information set 2 is respectively compared with the first decision information, the second decision information and the third decision information to obtain the second similarity data 1, the second similarity data 2 and the second similarity data 3 by comprehensive statistics of the data results; The first type of similarity data 1 and the second type of similarity data 1 are averaged to obtain a first comprehensive similarity; The first similarity data 2 and the second similarity data 2 are averaged to obtain a second comprehensive similarity; The first similarity data 3 and the second similarity data 3 are averaged to obtain a third comprehensive similarity; The maximum comprehensive similarity is extracted from the first comprehensive similarity, the second comprehensive similarity and the third comprehensive similarity, and the corresponding optimal decision information is determined from the first decision information, the second decision information and the third decision information according to the maximum comprehensive similarity, and the auxiliary decision analysis result is output.

7. An auxiliary decision-making system based on natural language understanding of text, characterized in that: The system is used to implement the auxiliary decision-making method based on text natural language understanding as described in any one of claims 1 to 6, comprising: The first text information analysis unit is used to obtain target text information, and according to the target text information, screen out the first decision information, the second decision information and the third decision information, and according to the three types of decision information, respectively extract the overlapping text parts between the text information sets obtained by inverting the first decision information, the second decision information and the third decision information to obtain the first category overlapping text set 1, the first category overlapping text set 2 and the first category overlapping text set 3; The second text information analysis unit is used to perform overlapping text comparison on the target extended text information set obtained by performing text expansion on the target text information and the text information sets obtained by inferring the first decision information, the second decision information and the third decision information to obtain the second category overlapping text set one, the second category overlapping text set two and the second category overlapping text set three; A new decision prediction unit 1 is used to perform statistical processing on the results of comparing the overlapped and different parts of the second category overlapped text set 1, the second category overlapped text set 2 and the second category overlapped text set 3 with the first category overlapped text set 1, the first category overlapped text set 2 and the first category overlapped text set 3 under various circumstances, so as to predict a new decision information set 1; A second new decision prediction unit is used to perform step-by-step segmentation processing on the first decision information, the second decision information and the third decision information to obtain a first sub-step set, a second sub-step set and a third sub-step set, and to select a first pre-processing sub-step set, a second pre-processing sub-step set and a third pre-processing sub-step set from the first sub-step set, the second sub-step set and the third sub-step set, respectively, to reintegrate to obtain a second new decision information set; The decision optimization unit is used to perform similarity comparison between the new decision information set 1 and the first decision information, the second decision information and the third decision information to obtain first-class similarity data 1, first-class similarity data 2 and first-class similarity data 3, perform similarity comparison between the new decision information set 2 and the first decision information, the second decision information and the third decision information to obtain second-class similarity data 1, second-class similarity data 2 and second-class similarity data 3, perform corresponding data comprehensive processing on the two types of three similarity data, so as to determine the corresponding optimal decision information from the first decision information, the second decision information and the third decision information, and output the auxiliary decision analysis result.

Images