TWI759861B - Contract risk evaluation system and contract risk evaluation method - Google Patents

Abstract

A contract risk evaluation system comprises a template server, an input interface, and a risk control host. The template server stores a plurality of contract templates having different objects respectively. Each of the contract templates contains a plurality of types of agreed items, each of the agreed items has multiple keywords. Different types of agreed items correspond to different weights respectively. The input interface receives a contract draft and provide a user to select the object of the contract draft. The risk control host stores a machine learning model, and the machine learning model obtains the contract template whose object is the same as the object of the contract draft from the template server. The machine learning model compares the contract draft with the contract template to calculate the percentage of keywords of each of the agreed items which do not appear in the contract draft. The machine learning model calculates a sum of products of these percentages and the weights corresponding to the agreed items, and the sum of products is defined as the risk level of the contract draft.

Description

Contract risk assessment system and contract risk assessment method

The present invention relates to a risk assessment system, especially a contract risk assessment system.

With the vigorous development of modern commercial society and the frequent trading and cooperation, legal disputes may arise at any time, which makes contracts indispensable. However, how to formulate an effective contract and the matters that should be paid attention to when signing the contract, etc., clarifying these issues has long been a subject that modern practitioners must face. The contract is also the most favorable evidence. In addition to protecting and extending one's own rights, it is also conducive to the development of the cooperative relationship and minimizes unexpected and accidental damage.

At present, when evaluating the risk level of a contract, companies usually rely on the company's legal specialist or an attorney from an external firm. However, when external lawyers handle multiple client cases at the same time, the time that can be spent on evaluating contract risks is relatively less, making it easy to make mistakes when evaluating contracts. In addition, manual review of contracts is lengthy and costly.

In view of this, there is indeed a need for an improved contract risk assessment system in practice, which can at least address the above deficiencies

The present invention is to provide a contract risk assessment system and a contract risk assessment method, which can reduce the time and cost of contract review.

A contract risk assessment system disclosed according to an embodiment of the present invention includes a template server, an input interface, and a risk control host. The template server stores contract templates of different contract objects, each contract template includes multiple types of contract items, each type of contract item has multiple keywords, and different types of contract items correspond to different weights. The input interface is communicatively connected to the template server, and the input interface receives the draft contract and confirms the contract subject of the draft contract. The risk management host communicates with the input interface and the template server. The risk management host stores a keyword search engine and a machine learning model. The keyword search engine compares the draft contract and the contract template with the same contract object as the draft contract to calculate The machine learning model calculates the product of the missing percentage of each engagement and the weight of each engagement and the sum of these products, where the missing percentage is that the keyword of each engagement does not appear The percentage of the number of draft contracts relative to the total number of keywords for each contract, and the sum of these products is the risk level of the draft contract. The risk control host includes a processor and an external memory, the processor is electrically connected to the external memory, the processor also has an internal memory, and the internal memory is used to store the keyword search engine and the external memory. a neural network, the external memory is used to store the draft contract, the processor drives the keyword search engine to compare the draft contract text file with the contract template, the processor drives the neural network to calculate the risk level , and the missing percentage is the percentage of the number of keywords for each engagement that do not appear in the draft contract relative to the number of total keywords for each engagement.

A contract risk assessment method disclosed according to an embodiment of the present invention includes: Use the input interface to receive the draft contract; use the input interface to receive the draft contract; use the input interface to confirm the contract subject of the draft contract; use the risk control host to execute the keyword search engine and machine learning model; use the keyword search engine to obtain and The draft contract is a contract template with the same contract object. The contract template contains different types of agreed items. The contractual items in the contract template include the rights of both parties, the responsibilities of both parties, breach of contract conditions, termination conditions, declarations and warranties of both parties, and confidentiality obligations. Each agreement has multiple keywords, and different agreements correspond to different weights; use a keyword search engine to compare the draft contract with the contract template to calculate the percentage of omissions for each agreement; use a machine learning model to calculate each the product of the missing percentage of engagements and the weight of each engagement; and computing the sum of those products with a machine learning model, where the missing percentage is the number of each engagement keyword not appearing in the draft contract relative to each engagement The percentage of the number of total keywords of the matter, and the sum of these products is the risk level of the draft contract

According to the above structure, the contract risk assessment system and contract risk assessment method disclosed in the present invention calculates the percentage of omissions of each contract through keyword search, and calculates the product of the percentage of omissions and their corresponding weights through a machine learning model, so as to speculate and draft The risk level of the contract is for the company's reference. It not only saves the time of manual contract review and manual query of the regulatory database, but also saves capital costs. During the busy period of the legal staff's business, it can also assist the legal staff to assess the risk of the contract, so as to avoid the loss of the company due to the defects of the contract.

The above description of the present disclosure and the following description of the embodiments are used to demonstrate and explain the spirit and principle of the present invention, and provide further explanation of the scope of the patent application of the present invention.

The detailed features and advantages of the present invention are described in detail below in the embodiments, and the content is sufficient to enable any person skilled in the relevant art to understand the technical content of the present invention and implement it accordingly, and according to the content disclosed in this specification, the scope of the patent application and the drawings , any person skilled in the related art can easily understand the related objects and advantages of the present invention. The following examples further illustrate the viewpoints of the present invention in detail, but do not limit the scope of the present invention in any viewpoint.

FIG. 1 is a functional block diagram of a contract risk assessment system according to a first embodiment of the present invention. As shown in FIG. 1 , the contract risk assessment system 100 includes a template server 10 , an input interface 12 and a risk control host 14 . The template server 10 stores contract templates of different contract objects, and the format of the contract templates is a text file. For example, the subject matter of the contract may include, but not limited to, financial cards attached to demand deposits, rental of safe deposit boxes in financial institutions, customer funds retained in special delivery accounts of securities firms, business of electronic payment institutions, and advertising agencies. Each contract template contains various types of covenants. For example, covenants may include the rights of both parties, obligations of both parties, conditions of breach, termination conditions, representations and warranties of both parties, and confidentiality obligations, but not in this regard. limited. Different types of commitments correspond to different weights, and each type of commitment has a plurality of keywords, and the data of the weights and keywords are stored in the template server 10 .

The input interface 12 is communicated with the template server 10 through a wired or wireless network, and the user can use an electronic device S (personal computer, mobile communication device or tablet computer) that can be connected to the network to communicate with the input interface 12 and draw up a contract Uploaded to the input interface 12, and the format of the uploaded draft contract is a text file. The input interface 12 has a contract target menu for the user to select the same contract target as the contract target of the draft contract. After the user selects the contract target, the input interface 12 generates a confirmation signal.

The risk management host 14 includes a processor 142 and an external memory 144. The processor 142 is electrically connected to the external memory 144, and the processor 142 is connected to the template server 10 through a wired or wireless network and the input interface 12 in communication. External memory 144 is used to store draft contracts. The processor 142 also has an internal memory 146, and the internal memory 146 is used to store the keyword search engine 16 and the machine learning model 18, and the machine learning model 18 includes a neural network-like, psycholinguistic programming (Neuro-Linguistic Programming) , Fuzzy Logic Models, Hidden Markov Models, Decision Trees, Bayesian Algorithms, Conditional Random Fields or Support Vector Machines. The processor 142 receives the confirmation signal to know the contract subject to which the draft contract belongs, and the processor 142 executes the keyword search engine 16 to drive the keyword search engine 16 to obtain the contract with the same contract subject as the draft contract from the template server 10 according to the confirmation signal The weight of each contractual item in the template, the contract template, and the information of the key word of each contractual matter, the contractual matter with a higher weight indicates a higher influence on the risk degree of the contract. The processor 142 executes the keyword search engine 16 to drive the keyword search engine 16 to compare the two text files of the draft contract and the contract template to calculate the missing percentage for each contractual item, and the missing percentage is each contractual term of the contract template The percentage of the number of keywords not appearing in the draft contract relative to the number of total keywords per contract. The processor 142 executes the machine learning model 18 to drive the machine learning model 18 to calculate the product of the missing percentage of each engagement and the weight of each engagement and to calculate the sum of the products, the sum of which is the risk of drafting the contract degree. For example, as listed in Table 1 below, the contract template has stipulations A to F, the draft contract has stipulations A to F, and the stipulations A to F are the rights of the contracting parties, the obligations of the contracting parties, the breach conditions, and the termination conditions. , the parties' declarations and warranties, and confidentiality obligations. Agreement A Covenant B contract C contract D Agreement E Covenant F Weights 25% 25% 15% 10% 10% 15% Total number of keywords 4 4 4 4 4 4 Number of keywords missing from draft contracts 2 2 2 2 2 2 (Table 1) Each of the contract items A~F of the draft contract lacks 2 keywords relative to each of the contract items A~F of the contract template, so the percentage of missing items A to F of each contract is 50%. The calculation formula of the product sum is 25%*50%+25%*50%+15%*50%+10%*50%+10%*50%+15%*50%. The higher the risk level of the contract.

Furthermore, for the consideration of information security, after logging in the input interface 12 , the user must also input a personal password before uploading the draft contract to the input interface 12 , thereby controlling the data uploaded to the input interface 12 .

FIG. 2 is a functional block diagram of a contract risk assessment system according to a second embodiment of the present invention. As shown in FIG. 2 , the contract risk assessment system 200 of the second embodiment further includes an image recognition host 20 compared to the contract risk assessment system 100 of the first embodiment. The image recognition host 20 communicates with the input interface 12 and the processor 142 of the risk management host 14 via a wireless or wired network. Since the draft contract may be an image file, the text on the image file is not pure text and cannot be edited and copied directly. The text on the image file must be recognized to obtain the text file. First, the user provides the image file of the draft contract to the input interface 12 , and then the input interface 12 transmits the image file to the image recognition host 20 . The image recognition host 20 includes a color selection interface 201 , a capture circuit 202 , a pixel conversion circuit 203 , a background removal circuit 204 and a character recognition circuit 205 . The user clicks the first RGB value of the text of the image file through the color selection interface 201, the text is red, for example, the first RGB value of the text is (255, 0, 0), which is intercepted according to the first RGB value of the text out the range of text. After the user clicks the first RGB value, the capture circuit 202 reads the RGB value of each pixel in the image file. When the RGB value of the pixel in the read image file is equal to the first RGB value, capture The fetching circuit 202 forms the first block including the area that matches the first RGB value. Then, the pixel conversion circuit 203 converts the pixels in the first block that are not of the first RGB value into a second RGB value, and the second RGB value is different from the first RGB value. The first block after the pixel conversion is completed includes the first RGB value and the second RGB value. Next, the background removal circuit 204 will remove all pixels in the first block that are equal to the second RGB value, leaving the first RGB value. The pixels form a second block, and the second block contains only pixels of the first RGB value. Next, the text recognition circuit 205 recognizes the text in the second block, and recognizes the text in the second block through font trimming, segmentation of single characters, thinning of single characters, or extraction of feature points, and outputs the recognition result after recognition to obtain the recognized text. files. After the image recognition host 20 converts the draft file of the image file into the draft contract of the text file through the aforementioned image recognition algorithm, the text recognition circuit 205 of the image recognition host 20 transmits the draft contract of the text file to the risk control host 14. Processor 142.

FIG. 3 is a functional block diagram of a contract risk assessment system according to a third embodiment of the present invention. As shown in FIG. 3 , compared with the contract risk assessment system 100 of the first embodiment, the contract risk assessment system 300 of the third embodiment further includes an image recognition host 20 and a law server 30 , and the law server 30 stores the latest For financial regulations, such as the Electronic Ticket Issuance Regulations, the Financial Holding Company Act, and the Banking Act, the template server 10 and the input interface 12 communicate with the regulation server 30 through a wired or wireless network, thereby allowing the user to perform a Financial regulations amend the contract model to avoid invalidity due to conflict between the contract model and existing regulations.

FIG. 4 is a flowchart of a contract risk assessment method according to the first embodiment of the present invention. As shown in FIG. 4 , in step S401 , a draft contract uploaded by the user is received through the input interface 12 , and the uploaded draft contract is a text file. In step S402, the input interface 12 has a contract subject menu for the user to select the same contract subject as the draft contract subject, so that the input interface 12 confirms the draft contract subject and generates a confirmation signal. In step S403, the processor 142 of the risk control host 14 executes the keyword search engine 16 and the machine learning model 18. In step S404, the keyword search engine 16 obtains from the template server 10 a contract template that is the same as the contract subject of the draft contract. The contract template contains different types of covenants, each covenant in the contract template has multiple keywords, and different covenants correspond to different weights. In step S405, the keyword search engine 16 compares the draft contract uploaded by the user with the contract template obtained from the template server 10 to calculate the missing percentage of each contractual item. The greater the degree of risk. In step S406, the machine learning model 18 is used to calculate the product of the missing percentage of each engagement and the weight of each engagement. In step S407, the machine learning model 18 is used to calculate the sum of the products corresponding to the agreements, wherein the percentage of missing is that the keywords of each agreement in the contract template do not appear in the number of draft contracts relative to the number of each agreement. The percentage of the total number of keywords, and the sum of these products is the risk level of contract drafting. The larger the sum of the products, the higher the risk level of contract drafting.

FIG. 5 is a flowchart of a contract risk assessment method according to a second embodiment of the present invention. As shown in Figure 5, since the format of the draft contract is not necessarily a text file, but may be an image file, when the draft contract is an image file, since the file format of the draft contract and the contract template are different, the keyword search engine 16 The keyword comparison procedure between the draft contract and the contract template cannot be performed. In view of this, the difference between the contract risk assessment method of the second embodiment and the contract risk assessment method of the first embodiment is that it further includes step S502 : before confirming the contract subject of the draft contract with the input interface 12 , first identify the host 20 with an image The draft contract of the image file is converted into the draft contract of the text file, and the draft contract of the text file is transmitted to the external memory 144 of the risk management host 14 . Steps S501 and S503 to S508 are the same as steps S401 to S407, respectively. In this way, regardless of whether the draft contract is an image file or a text file, the risk level can be evaluated by the contract risk evaluation method of the second embodiment.

In other embodiments, when the risk level calculated by the machine learning model 18 is greater than the risk upper limit, the machine learning model 18 further sends a text message to the user's mobile device to remind the user that the draft contract does not meet the standard and must be re-formulated .

In the actual operation of the contract risk assessment system, the machine learning model 18 must be continuously updated in order to improve the accuracy of predicting the degree of risk. Therefore, the contract risk assessment method of the present invention may further include that after each time the risk level of a draft contract is calculated by the machine learning model 18, using the machine learning model 18 to manually review the draft contract according to the calculated risk level and experts. Errors in the estimated reference risk level are debugged and corrected for the machine learning model 18, thereby reducing the difference between machine review and manual review. For example, the machine learning model 18 used is a neural network, the percentage of the keywords of each covenant of the contract template that does not appear in the draft contract and the weight of each covenant of the draft contract as the input of the neural network , the risk level of the draft contract is taken as the output of the neural network, and the weight of the neural network is modified according to the error between the output of the neural network and the reference risk level reviewed by experts manually. When the weights of the neural network are modified for different draft contracts, the neural network can accurately predict the draft when receiving the real-time data of the keyword missing percentage of the draft contracts and the weights of the covenants of the draft contracts. The risk level of the contract.

Based on the above, the contract risk assessment system and contract risk assessment method disclosed in the present invention calculates the percentage of each contractual item that does not appear in the draft contract through keyword searching and calculates each contractual item through a machine learning model The product of the missing percentage of the keywords and the weight of each contract item is used to estimate the risk level of the draft contract for the company's reference. It not only saves the time of manual contract review and manual query of the regulatory database, but also saves capital costs. During the busy period of the legal staff's business, it can also assist the legal staff to assess the risk of the contract, so as to avoid the loss of the company due to the defects of the contract.

Although the present invention is disclosed in the foregoing embodiments, it is not intended to limit the present invention. Changes and modifications made without departing from the spirit and scope of the present invention belong to the scope of patent protection of the present invention. For the protection scope defined by the present invention, please refer to the attached patent application scope.

100, 200, 300: Contract Risk Assessment System 10: Template server 12: Input interface 14: Risk Control Host 142: Processor 144: External memory 146: Internal memory 16: Keyword Search Engines 18: Machine Learning Models 20: Image recognition host 201: Color selection interface 202: Capture circuit 203: Pixel conversion circuit 204: Background removal circuit 205: Text recognition circuit 30: Regulatory Server S: electronic device

FIG. 1 is a functional block diagram of a contract risk assessment system according to a first embodiment of the present invention. FIG. 2 is a functional block diagram of a contract risk assessment system according to a second embodiment of the present invention. 3 is a functional block diagram of a contract risk assessment system according to a third embodiment of the present invention. FIG. 4 is a flowchart of a contract risk assessment method according to the first embodiment of the present invention. FIG. 5 is a flowchart of a contract risk assessment method according to a second embodiment of the present invention.

100: Contract Risk Assessment System

10: Template server

12: Input interface

14: Risk Control Host

142: Processor

144: External memory

146: Internal memory

16: Keyword Search Engines

18: Machine Learning Models

S: electronic device

Claims (5)

A contract risk assessment system, including: a template server, storing contract templates of different contract objects, each contract template contains multiple types of contract items, each type of contract item has multiple keywords, and different types of contract items correspond to different weights; an input interface, communicatively connected to the template server, the input interface receives a draft contract image file and confirms the contract object of the draft contract image file; an image recognition host, comprising: a color selection interface , for selecting a first RGB value of the text portion of the draft contract image file and extracting the range of the text portion according to the first RGB value; a capture circuit for reading the draft contract image the RGB value of each pixel on the image file to form a first block, wherein the first block includes an area conforming to the first RGB value; a pixel conversion circuit is used for the first block Converting pixels not conforming to the first RGB value into a second RGB value; a background removal circuit for removing all pixels in the first block that conform to the second RGB value to form a second block; and a character recognition circuit for recognizing the character part in the second block through font trimming, dividing single characters, thinning single characters, or extracting feature points to generate a draft contract text file; and a risk control host, communication connection In the input interface, the template server and the text recognition circuit, the risk management host stores a keyword search engine and a machine learning model, and the machine learning model is a type of neural network, and the keyword search engine compares The draft contract and the contract template with the same contract object as the draft contract are used to calculate the missing percentage of each contract. The neural network is used to calculate the missing percentage of the contract and the weight of the contract. drafting a risk level of the contract text file and correcting a plurality of weights of the neural network according to the error between the risk level and a reference risk level; wherein the risk control host includes a processor and an external memory, the The processor is electrically connected to the external memory, the processor also has an internal memory, the internal memory is used for storing the keyword search engine and the neural network, and the external memory is used for storing the draft contract , the processor drives the keyword search engine to compare the draft contract text file with the contract template, the processor drives the neural network to calculate the risk level, and the missing percentage is that the keyword for each contract does not appear The percentage of the number of draft contracts relative to the number of total keywords for each contract. The contract risk assessment system according to claim 1, wherein the agreed items of the contract template include the rights of both parties, the responsibilities of both parties, breach conditions, termination conditions, declarations and warranties of both parties, and confidentiality obligations. The contract risk assessment system according to claim 1, wherein the contract risk assessment system further comprises a regulation server, the regulation server stores current regulations, and the template server is connected to the regulation server in communication. A contract risk assessment method, comprising: receiving a draft contract image file through an input interface; confirming the contract subject of the draft contract image file through the input interface; clicking the text of the draft contract image file through a color selection interface a first RGB value of a part; use the color selection interface to extract the range of the text part according to the first RGB value; use a capture circuit to read the RGB value of each pixel on the draft contract image file to form a first block, wherein the first block includes an area that conforms to the first RGB value; a pixel conversion circuit is used to convert the pixels in the first block that do not conform to the first RGB value into a second RGB value; a removing circuit is used to remove all pixels in the first block that conform to the second RGB value to form a second block; a character recognition circuit is used to trim, segment, and thin a single character through fonts , or extracting feature points to identify the text part in the second block to generate a draft contract text file; use the text recognition circuit to send the draft contract text file to a risk management host; execute with the risk management host a keyword search engine and a machine learning model, wherein the machine learning model is a type of neural network; using the keyword search engine to obtain a contract template with the same contract subject as the draft contract text file from a template server, the contract template There are different types of covenants. These covenants in this contract template include the rights of both parties, responsibilities of both parties, conditions of breach of contract, termination conditions, declarations and warranties of both parties, and confidentiality obligations. Each covenant has multiple keywords. , and different agreements correspond to different weights; Use the keyword search engine to compare the draft contract text file with the contract template to calculate the missing percentage of each contract; The percentage of gaps and the weights of the engagements to calculate a risk level of the draft contract document; and a plurality of weights to the neural network based on the error between the risk level and a reference risk level with the neural network Make corrections. The contract risk assessment method as claimed in claim 4, wherein the missing percentage is the percentage of the number of keywords of each contract that do not appear in the draft contract relative to the number of all keywords of each contract, and the The sum of the percentages of omissions of the covenants and one of the products of the weights of the covenants is the risk level of the draft contract.

Images