CN111160345A

CN111160345A - Intelligent enterprise contract generation system and method

Info

Publication number: CN111160345A
Application number: CN201911413063.9A
Authority: CN
Inventors: 莫紫霄
Original assignee: Chongqing Mushe Technology Co ltd
Current assignee: Chongqing Mushe Technology Co ltd
Priority date: 2019-12-31
Filing date: 2019-12-31
Publication date: 2020-05-15

Abstract

The invention relates to the technical field of computers, in particular to an enterprise contract intelligent generation system, which comprises: the image acquisition unit is used for acquiring a handwritten contract image and sending the contract image; the image preprocessing unit is used for receiving the contract image, denoising the contract image and sending the processed image; a character extraction unit for receiving the processed image, extracting characters of the image, and transmitting the extracted characters; the character classification unit is used for receiving the extracted characters, classifying the contracts according to the meanings of the characters and sending a classification result; the character matching unit is used for matching the contract template base according to the classification result and sending the matching result; the text generation unit is used for receiving the matching result and calling a contract template according to the matching result to generate a contract text; and the text output unit is used for outputting the contract text. The invention converts the character record and the recording record of consultation and negotiation into the electronic contract text in a specific form, can save the procedure of contract signing, and thus improves the efficiency of contract signing and management.

Description

Intelligent enterprise contract generation system and method

Technical Field

The invention relates to the technical field of computers, in particular to an enterprise contract intelligent generation system and method.

Background

The contract is an agreement for establishing, changing and terminating civil relations between parties, is a product appearing along with the development of commodity economy, and is widely used in various commercial buying and selling, leasing, contracting, processing and other activities. Because the paper-based contracts are not easy to store and inconvenient to carry, many enterprises begin to adopt electronic contracts so as to improve the contract management efficiency.

In view of this, document CN108132926A discloses a contract generating apparatus and system, including: the information acquisition unit is used for generating a declaration form according to filling information of the form to be declared and acquiring all target information in the declaration form; the information sending unit is in communication connection with the information acquisition unit and is used for transmitting the declaration form; the multiple auditing unit is in communication connection with the information sending unit and is used for auditing all target information in the declaration form for multiple times so as to judge whether the target information meets a preset condition; the target generating unit is in communication connection with the multiple auditing units and is used for generating a target contract according to the target information; and the electronic seal signing unit is in communication connection with the target generating unit and is used for signing the electronic seal for the target contract.

With the improvement of electronic signature law and related laws and regulations, many contracts can be signed in the form of electronic texts at present. However, the practical situation is limited, the legal level of people is uneven, and the situation that the handwritten contract is used on a large scale still exists. The handwritten contract format is not standard, is not easy to store and is inconvenient to carry. In addition, before the contract is formally made, individuals or businesses have a similar bargaining and bargaining process such as consultation and conversation. In the process of consultation and conversation, corresponding character records are available about the intentions of both parties and the important terms of the contract; for some extremely important contracts, the parties may even record each other. After both parties reach the agreement, they sign the formal contract. However, these character records and sound records contain all the information of the parties in consultation, and the contract text can be directly generated by the information without the parties setting up the terms again. Therefore, the handwritten paper contract or the character record and the recording record of consultation and consultation are converted into the electronic contract text in a specific form, so that a contract signing program can be omitted, and the efficiency of contract signing and management is improved.

Disclosure of Invention

The invention provides an enterprise contract intelligent generation system and method, which convert the handwritten paper contract or the character record and the recording record of consultation and conversation into the electronic contract text in a specific form; the procedure of signing the contract is saved, so that the efficiency of signing and managing the contract is improved; the technical problem that the time is needed to sign formal contract texts after both parties negotiate and talk is solved.

The basic scheme provided by the invention is as follows: an intelligent enterprise contract generation system comprising: the image acquisition unit is used for acquiring a handwritten contract image and sending the contract image; the image preprocessing unit is used for receiving the contract image, denoising the contract image and sending the processed image; a character extraction unit for receiving the processed image, extracting characters of the image, and transmitting the extracted characters; the character classification unit is used for receiving the extracted characters, classifying the contracts according to the meanings of the characters and sending a classification result; the character matching unit is used for matching the contract template base according to the classification result and sending the matching result; the text generation unit is used for receiving the matching result and calling a contract template according to the matching result to generate a contract text; and the text output unit is used for outputting the contract text.

The working principle of the invention is as follows: firstly, extracting characters in the records of consultation and consultation, then classifying according to the meanings of the characters, matching the characters with corresponding contract templates, and finally generating a contract text with a preset format. The invention has the advantages that: after two parties of the contract negotiate and negotiate, a contract text can be generated immediately according to records of negotiation and negotiation, and a contract signing process is omitted; meanwhile, handwritten consultation and conversation records are converted into electronic texts in a specific form, so that the electronic texts are convenient to standardize and manage, easy to store and convenient to carry.

The contract text is directly generated according to the records of consultation and consultation, so that a contract simplifying program can be simplified, and consultation records can be quickly processed; if the two parties have new ideas. The modification can also be directly made on the generated electronic contract text.

Further, the image preprocessing unit adopts a gaussian filtering algorithm. Most of noise of the image belongs to Gaussian noise, the algorithm is used for carrying out weighted average on the whole image, the value of each pixel point is obtained by carrying out weighted average on the value of each pixel point and other pixel values in the neighborhood, and the algorithm is suitable for eliminating the Gaussian noise.

Further, the text extraction unit extracts the LBP features of the text, and the specific steps include: s1: obtaining a basic local binary pattern of each pixel of the window; s2: the realization of invariable rotation: circularly right-shifting the binary string to obtain all possible values, and taking the minimum value as the LBP value of the current window; s3: reducing feature dimensions by using an equivalent mode; s4: obtaining a histogram of the whole window by using the LBP value of each pixel; s5: judging whether to traverse the whole image, and if not: step S1 is returned to after the step moving window is connected; if yes, go to step S6; s6: connecting the histograms of all the windows to obtain LBP characteristics under the window scale, and scaling the window according to the window scaling factor; s7: judging whether the maximum window upper limit is exceeded, if so, judging that: outputting a result; otherwise: returning to step S1, the image is traversed from the beginning using the new window size. LBP (Local Binary Pattern) is an operator for describing Local texture characteristics of an image, has rotation invariance and gray scale invariance, and has a good identification effect under the condition of severe illumination change.

Further, the text classification unit comprises the following specific steps: s1: and word segmentation, namely segmenting a Chinese character sequence into separate words. The successful Chinese word segmentation can achieve the effect of improving the automatic recognition of the sentence meaning by the computer. S2: stop words without an actual meaning are removed. The words are removed, so that the index amount can be reduced, the retrieval efficiency is improved, and the retrieval effect is improved. The stop words mainly comprise English characters, numbers, mathematical characters, punctuation marks and single Chinese characters with high use frequency. S3: and constructing a bag-of-words space. Ignore, word order, grammar, syntax, consider it to be a set of words, the occurrence of each word in the text is independent. Thus, the natural language is simplified and is convenient for modeling. S4: the TF-IDF constructs word weights. TF-IDF is a statistical method to evaluate the importance of a word to one of a set of documents or a corpus, the importance of a word increasing in proportion to the number of times it appears in a document. S5: and (6) classifying. The algorithm is very easy to understand and easy to implement.

Further, the text matching unit adopts a KMP algorithm, and the method specifically comprises the following steps: s1: finding the longest common element length of each substring prefix and suffix of the pattern string; s2: solving a next array; s3: and matching according to the next array. Compared with BF, RK, KMP and BM algorithms, the KMP algorithm can improve the matching speed and reduce the memory requirement.

Further, the text generating unit further comprises a watermark module for receiving the contract text and generating the watermark at the specified position of the contract text. Through the watermark, the anti-counterfeiting effect can be achieved, and the legal risk caused by contract loss is effectively reduced.

Further, the text generation unit includes a date module for receiving the contract text and generating a signing date of the contract in a specified format at a specified position of the contract text. Management of the fulfillment period and fulfillment status of the contract is facilitated based on the date of generation.

Further, the text generation unit also comprises a page number module for receiving the contract text and generating the page number of each page of the contract according to the specified format at the specified position of the contract text. According to the page number, the arrangement, reading and binding of the contract are facilitated.

Further, the text generation unit also comprises a signature module which is used for receiving the contract text and generating the electronic signature and the seal of the contract party according to the specified format at the specified position of the contract text. Therefore, the signing of both parties is not needed, and the signing efficiency of the contract is improved.

Drawings

Fig. 1 is a block diagram of a system structure of an embodiment of an intelligent enterprise contract generation system according to the present invention.

Detailed Description

The following is further detailed by the specific embodiments:

example 1

The embodiment of the enterprise contract intelligent generation system is basically as shown in the attached figure 1, and comprises an image acquisition unit, an image preprocessing unit, a character extraction unit, a character classification unit, a character matching unit, a text generation unit and a text output unit. In this embodiment, the scanner is used as the image capturing unit, and the scanner is mounted on a DELLT440 tower server. The contract of the invention comprises all documents related to the contract, such as consultation, literal records of the conversation, and the like.

In this embodiment, both parties to the contract have completed negotiations and conversations, and a handwritten conversation record is formed in the form described below. The general contents are as follows:

"prescription A: i intend to buy 100-200 pieces of pattern steel plates with the specification of 1500mm x 4000mm x 20mm from the noble company.

B, prescription B: the patterned steel plate of the specification of our company is sufficient in source, and the expensive units buy 100 to 200 patterned steel plates and have the patterned steel plates at any time.

Prescription A: how much the price of the pattern steel plate of the noble unit is? Do you have a discount on a multi-purchase?

B, prescription B: generally, 4000 yuan of the specification is one piece, and if the size is large, eight folds can be made.

Prescription A: currently my company intends to buy 150, and if of good quality, 50.

B, prescription B: the method has no problem, and the company delivers goods in a bag, absolutely guarantees the quality and quantity, and delivers the goods by using a truck 7 days after payment.

Prescription A: well, do our company pay using the check of the building bank in china, and dispute what is you seeing is handled by the x-th middle people court in Chongqing?

B, prescription B: there is no problem. "

The image acquisition unit adopts a Fujitsu ix500 scanner which has a WIFI wireless transmission function. When the parties to the contract end the session, the handwritten session log is scanned into an image using the scanner and the image is sent to the DELLT440 tower server. The image preprocessing unit is provided with Adobe Photoshop software (or program codes programmed with filtering algorithms, source codes of various image filtering algorithms can be found in a CSDN technical community), after the image is received, the image is subjected to denoising processing by adopting a Gaussian filtering algorithm, and the processed image is sent to the character extraction unit.

And after receiving the preprocessed image, the character extraction unit extracts characters by extracting LBP characteristics of the characters in the image. The method comprises the following specific steps: step one, a basic local binary pattern of each pixel of a window is obtained. And step two, realizing the unchanged rotation, circularly right shifting the binary string to obtain all possible values, and taking the minimum value as the LBP value of the current window. And step three, reducing the characteristic dimension by using the equivalent mode. And step four, obtaining a histogram of the whole window by using the LBP value of each pixel. Step five, judging whether to traverse the whole image, and if not: step S1 is returned to after the step moving window is connected; if yes, step six is executed. And step six, connecting the histograms of all the windows to obtain the LBP characteristics under the window scale, and scaling the window according to the window scaling factor. Step seven, judging whether the maximum window upper limit is exceeded, if so: outputting a result; otherwise: return to step one to traverse the image from scratch using the new window size. The above steps can be implemented by programming corresponding programs (for example, the source codes of MATLAB, C language and C + + language of the algorithm for extracting LBP features in image texts can be found in the CSDN technology community). After the character extraction unit extracts the LBP characteristics of the characters in the image, the extracted characters are sent to the character classification unit.

And after receiving the extracted characters, the character classification unit classifies the characters according to the meanings of the characters. The method comprises the following specific steps: step one, word segmentation, namely segmenting a Chinese character sequence into independent words. For example, the phrase "generally 4000 yuan of the specification can be folded into eight if the amount is large" is cut into "generally 4000 yuan of the specification can be folded into eight if the amount is large". And step two, removing stop words without actual meanings. For example, the three terms "generally, if the amount of such specification is large" have no too large practical meaning and can be removed; obtain '4000 yuan one piece, can make eight folds'. And step three, constructing a bag-of-words space. Neglect, word order, grammar and syntax, and regard three words of '4000 yuan, can be eight-folded' as a word set, and the appearance of each word is independent. And step four, the TF-IDF constructs word weight. TF-IDF is a statistical method to evaluate the importance of a word to one of a set of documents or a corpus, the importance of a word increasing in proportion to the number of times it appears in a document. For example, the weight of two words of "4000 yuan and eight folds" is very large. And step five, classifying. For example, the terms "4000 yuan with eight folds" are divided into categories of buying and selling contracts. And after classification, sending the classification result of the contract belonging to the buying and selling contract category to a character matching unit.

After the character matching unit receives the classification result that the contract belongs to the buying and selling contract, the KMP algorithm (the KMP algorithm finds the source codes in the forms of MATLAB, C language and C + + language in the CSDN technical community) is adopted to match the contract template base according to the classification result. The method comprises the following specific steps: step one, searching the longest common element length of each substring prefix and suffix of a pattern string; step two, solving a next array; and step three, matching according to the next array. Various contract templates such as buying and selling contracts, loan contracts, leasing contracts and the like are stored in the contract template library, and the KMP algorithm matches the conversation contents to the categories of the buying and selling contracts in the contract template library according to the steps. After the matching is completed, the character matching unit sends a matching result (position information of a contract template) to the text generation unit.

The text generation unit receives the matching result, namely the position information of the trading contract template, calls the trading contract template, and fills the conversation content into the trading contract text according to a keyword matching algorithm. The buying and selling contract template comprises the items of parties, targets, quantity, quality, price or reward, fulfillment period, place and mode, default responsibility, dispute solving method and the like, and the contract text can be generated by filling the contents of the conversation into the corresponding items. If the column has no corresponding content, a word of 'to be determined' is generated at the column.

After the contract text is generated, the text generation unit sends the generated contract text to the output unit. After receiving the contract text, the text output unit generates a watermark in the body part of the contract text, generates a signing date in a format of XXXXXXXXXXMXXXXday at the rightmost position of a footer of the contract text, and generates a page number in an Arabic numeral format in the middle of the footer of the contract text; and generating a blank position for signing and stamping by the party of the contract at the last position where the text content of the contract ends so as to directly sign when the party reads without objection.

The final output deal contract text is as follows:

contract for buying and selling pattern steel plate

The party: prescription A: company a, party b: company B

The target is as follows: patterned steel plate with 1500mm x 4000mm x 20mm specification

Quantity: 150 blocks

Quality: to be determined

Price or consideration: 4000 x 150 x 0.8 ═ 48 ten thousand yuan

Term, place and mode of fulfillment: 7 days, the seller delivers the goods to the door and the goods are transported by truck

The method for breach responsibility and dispute resolution: appeal to the first X middle-grade national institute of Chongqing

····

Prescription A: the signature seal of company A, year, month, day

B, prescription B: signature seal of company B, XX year XX month XX day "

Example 2

The only difference from example 1 is that: the DELLT440 tower server is also loaded with Praat voice analysis software.

When the two parties talk, the recording pen is used for recording the sound in real time. And after the conversation is finished, importing the sound recording file into a server. After the sound recording file is imported into the server, the Praat software analyzes and labels the voice signals in the sound recording. After the Praat software reads the audio file, a pitch curve (pitch curve), a formant curve (formant curve), an intensity curve (intensity curve), and the like can be displayed, and various text information can be obtained by calculating the signal data.

In this embodiment, the key word a with higher pitch, the key word B with longer speaking duration, and the key word C with more occurrences are mainly obtained during the conversation process of both parties. For example, when the number of riffled steel plates is 150 during the conversation, the pitch of the person is higher, for example, 10% higher than the normal pitch, and the "number" belongs to the key a. When the price of the figured steel plate is 4000 yuan/block, the speaking time of a person is long, for example, 0.1 second is delayed, and then the price belongs to the key word B. If the patterned steel plate is transported by truck for more than 3 times, the "transportation" belongs to the keyword C.

After the contract text is generated, the system prompts the parties to perform key point check on related contents corresponding to the A, B and the C key words in the contract text and marks the corresponding contents in red. The client checks according to the prompt field, and modifies, deletes or adds the problematic part on the server, and the contents are input by keyboard typing.

When a party modifies, deletes or adds to the synthetic text, the system captures keyboard typing actions and marks content changes resulting from the actions as features. Such as: when for "price: 4000 cells/block "and" number: when 150 blocks of' two sentences are deleted, if the two sentences are deleted at a time, the two sentences are regarded as a whole as the characteristic I. The characteristic I corresponds to the B and A keywords, and when the input content is detected to contain the B keyword (price) or the A keyword (quantity) after being deleted and modified, the characteristic I is popped up (price: 4000 yuan/block; quantity: 150 blocks). If the popped feature I is considered to have no problem by the party, directly inputting the feature I by one key; if the popped feature I is considered by the principal to be problematic, the modification is made on the basis of the feature I.

For another example: when for "price: 4000 cells/block "and" number: when 150 blocks of' two sentences are deleted, if the deletion is carried out twice, the price is deleted for the first time: 4000 cells/block ", second delete" quantity: 150 blocks ". Then, the "price: 4000 yuan/block "as a feature I, the feature I corresponds to the key word B (price); the "number: 150 blocks "as feature II, feature II corresponds to the A keyword (number). When modification is performed after deletion, if it is detected that the input content contains a B keyword (price), a feature I (price: 4000 yuan/block) is popped up. If the popped feature I is considered to have no problem by the party, directly inputting the feature I by one key; if the popped feature I is considered by the principal to be problematic, the modification is made on the basis of the feature I. Similarly, if it is detected that the input content contains the A keyword (number), the feature II (number: 150 blocks) is popped up. If the popped feature II is considered to have no problem by the principal, directly inputting the feature II by one key; if the popped feature II is considered by the principal to be problematic, the modification is made on the basis of feature II. Therefore, the method can lead the party to quickly modify, delete or add the output contract text on site, thereby shortening the contract period and improving the contract signing efficiency.

The foregoing is merely an example of the present invention, and common general knowledge in the field of known specific structures and characteristics is not described herein in any greater extent than that known in the art at the filing date or prior to the priority date of the application, so that those skilled in the art can now appreciate that all of the above-described techniques in this field and have the ability to apply routine experimentation before this date can be combined with one or more of the present teachings to complete and implement the present invention, and that certain typical known structures or known methods do not pose any impediments to the implementation of the present invention by those skilled in the art. It should be noted that, for those skilled in the art, without departing from the structure of the present invention, several changes and modifications can be made, which should also be regarded as the protection scope of the present invention, and these will not affect the effect of the implementation of the present invention and the practicability of the patent. The scope of the claims of the present application shall be determined by the contents of the claims, and the description of the embodiments and the like in the specification shall be used to explain the contents of the claims.

Claims

1. An enterprise contract intelligent generation system is characterized in that: the method comprises the following steps: the image acquisition unit is used for acquiring a paper contract image and sending the contract image; the image preprocessing unit is used for receiving the contract image, denoising the contract image and sending the processed image; a character extraction unit for receiving the processed image, extracting characters of the image, and transmitting the extracted characters; the character classification unit is used for receiving the extracted characters, classifying the contracts according to the meanings of the characters and sending a classification result; the character matching unit is used for matching the contract template base according to the classification result and sending the matching result; the text generation unit is used for receiving the matching result and calling a contract template according to the matching result to generate a contract text; and the text output unit is used for outputting the contract text.

2. The intelligent enterprise contract generation system of claim 1, wherein: the image preprocessing unit adopts a Gaussian filtering algorithm.

3. The intelligent enterprise contract generation system of claim 2, wherein: the character extraction unit extracts LBP characteristics of characters, and the specific steps comprise:

s1: obtaining a basic local binary pattern of each pixel of the window;

s2: the realization of invariable rotation: circularly right-shifting the binary string to obtain all possible values, and taking the minimum value as the LBP value of the current window;

s3: reducing feature dimensions by using an equivalent mode;

s4: obtaining a histogram of the whole window by using the LBP value of each pixel;

s5: judging whether to traverse the whole image, and if not: step S1 is returned to after the step moving window is connected; if yes, go to step S6;

s6: connecting the histograms of all the windows to obtain LBP characteristics under the window scale, and scaling the window according to the window scaling factor;

s7: judging whether the maximum window upper limit is exceeded, if so, judging that: outputting a result; otherwise: returning to step S1, the image is traversed from the beginning using the new window size.

4. The intelligent enterprise contract generation system of claim 3, wherein: the text classification unit comprises the following specific steps:

s1: word segmentation, namely segmenting a Chinese character sequence into separate words;

s2: removing stop words without actual meanings;

s3: constructing a word bag space;

s4: TF-IDF constructs word weight;

s5: and (6) classifying.

5. The intelligent enterprise contract generation system of claim 4, wherein: the text matching unit adopts a KMP algorithm, and the method specifically comprises the following steps:

s1: finding the longest common element length of each substring prefix and suffix of the pattern string;

s2: solving a next array;

s3: and matching according to the next array.

6. The intelligent enterprise contract generation system of claim 5, wherein: the text generation unit also comprises a watermark module which is used for receiving the contract text and generating the watermark at the specified position of the contract text.

7. The intelligent enterprise contract generation system of claim 6, wherein: the text generation unit also includes a date module for receiving the contract text and generating a signing date of the contract in a specified format at a specified position of the contract text.

8. The intelligent enterprise contract generation system of claim 7, wherein: the text generation unit also comprises a page number module which is used for receiving the contract text and generating the page number of each page of the contract according to the specified format at the specified position of the contract text.

9. The intelligent enterprise contract generation system of claim 8, wherein: the text generation unit also comprises a signature module which is used for receiving the contract text and generating the electronic signature and the seal of the contract party according to the specified format at the specified position of the contract text.

10. An intelligent enterprise contract generating method is characterized in that: the method comprises the following specific steps:

s1: acquiring a handwritten contract image and sending the contract image;

s2: receiving a contract image, denoising the contract image, and sending a processed image;

s3: receiving the processed image, extracting characters of the image, and sending the extracted characters;

s4: receiving the extracted characters, classifying the contracts according to the meanings of the characters, and sending a classification result;

s5: matching the contract template base according to the classification result, and sending the matching result;

s6: receiving a matching result, and calling a contract template according to the matching result to generate a contract text;

s7: and outputting a contract text.