TW202011227A - Method and system for intelligent learning word editing and multi-language translating - Google Patents

Abstract

A method and a system for intelligent learning word editing and multi-language translation are disclosed. The system includes a translation sentence acquisition module, a translation sentence analysis module, a word vocabulary correction module, a release module, a text layout module, a database module, and a word sentence recommendation module. The invention utilizes the word vocabulary correction module to replace translated sentences from more than three translation platforms into words close to that used in a particular field. Thus, it can solve the problem of translation correctness of the conventional translation platforms, and the problem that corresponding nouns used may not be familiar to the particular field.

Description

System and method for intelligently learning word editing and multilingual translation

The invention relates to a system and method for multi-language mutual translation, in particular to a system and method for intelligently learning word editing and multi-language mutual translation.

In today's society where information exchange is frequent, many news, published articles, and even public audio-visual information in foreign countries need to be sent back to the country in the fastest way for relevant people to receive, so that appropriate and timely responses can be made . Thanks to the popularity of the Internet, the aforementioned materials can be obtained on relevant websites or platforms. However, limited to people's understanding of foreign languages, these first-hand information may not be absorbed by correct understanding, but rather extend more problems. In order to solve this inconvenience, many online translation platforms came into being, such as google translation, Microsoft translation and Baidu translation. As the time for deep learning increases, these platforms can present the nearly correct translation on the user's terminal device after receiving foreign texts, sentences, paragraphs, or even articles input by the user, and after a few seconds of operation in the cloud.

Although the emergence of these translation platforms has solved many text translation problems encountered in today's life. However, the following problems still exist. First of all, the translation of the native semantics will occasionally produce differences from the original text. What's more, because of the different translation algorithms, even if the input sentence has the same meaning, different text or arrangement will still lead to different results after translation. Secondly, the corresponding nouns used by many translation platforms for translation may not be well known and commonly used in a certain country market, field or place. For example, when data is translated into Chinese, it is likely that data or information will be obtained. The former is common in mainland China, the latter is the mainstream of Taiwan. Moreover, receiving data in the field of science and technology is data, but traditional business transactions regard data as data. Although the differences between these translated texts are not large and can be understood by users, many commercial activities, such as marketing advertisements, hope that the translated sentences can use the text that is closest to the market acceptance, or even the most rammed synonyms. Therefore, the translation sentence obtained by the translation platform may not be directly usable, but needs further modification. With the increase of translation materials, this inconvenience has gradually changed from implicit cost to explicit cost.

Although there will be a certain degree of distortion in the correctness of the translated sentence (may be related to the unclear language of the original text), it can be alleviated by a certain degree of intelligent analysis and replacement. The text used in the translated sentence can also be adjusted appropriately. However, solutions to the above two problems are still lacking.

This paragraph extracts and compiles certain features of the invention. Other features will be revealed in subsequent paragraphs. Its purpose is to cover the spirit and scope of the additional patent application scope, various modifications and similar arrangements.

The purpose of the present invention is to provide a system and method for intelligent learning of word editing and multi-language mutual translation to solve the translation accuracy of traditional translation platforms, and the corresponding nouns used may not be a national market, field or place. Learn about common problems. The method includes the steps of: a) obtaining at least one first-party sentence from a remote server via a network; b) encoding the at least one first-party sentence through at least three translation platforms provided by API (Application Programming) Interface, application program interface), respectively sent to the at least three translation platforms for translation; c) respectively obtained from the at least three translation platforms, the first party sentence translated a second party sentence code to a servo The host; d) the servo host converts the encoding of the second-party sentences into at least 3 corresponding second-party sentences; e) the servo host selects the related keywords and the second-party sentences from the second-party sentences Analyze the respective used grammatical structure; f) replace a related keyword and a used grammatical structure in the servo host with a preset keyword and a preset grammatical structure in order to obtain a modified second-party sentence; and g) Send the modified second-party sentence to a designated end.

Preferably, the method may further include a step f1) after step f): applying a text layout format in the servo host to the modified second-party sentence.

According to the present invention, the text layout format may be a length limitation of sentence, a sentence breaking method, a punctuation method, a specified font style, a specified non-text symbol, or a combination of any two or more of the foregoing. The related keyword may be a second-party word that translates all the homonyms corresponding to a first-party word. The usage grammatical structure can be formed by at least one of the following: sentence translation rules, noun singular and plural translation rules, noun masculine translation rules, and article translation rules. The preset keyword may be a second-party word corresponding to all translations of a first-party word, and a statistical operation algorithm or a machine learning algorithm is used to select the frequency of use in the previous translation data in a specific field The highest one may dynamically select one of the N highest before the frequency of use, or specify a specific second-party word corresponding to the first-party word. The predetermined grammatical structure may be a second-party sentence pattern corresponding to all translations of a first-party sentence pattern, selected from previously translated materials in a specific field through statistical operation algorithms or machine learning algorithms The one with the highest usage frequency or dynamically selects one of the N highest before the usage frequency, or specifies a specific second-party sentence pattern corresponding to the first-party sentence pattern. Where N is 2, 3, 4, or 5, the machine learning algorithm may be a TF-IDF algorithm.

In another embodiment, a system for intelligent learning word editing and multi-language translation can be installed in a server host. The server host is connected to an operation terminal through a network, and includes: a translation sentence acquisition module, which accepts the The operation command of the operation terminal is to obtain at least one first-party sentence from a remote server via the network, and encode the at least one first-party sentence to the at least 3 through the API provided by at least 3 translation platforms respectively Translation platforms for translation, and the at least three translation platforms respectively obtain the codes of a translated second-party sentence of the first-party sentence, and convert the codes of the second-party sentences to at least 3 Sentence second-party sentence; a translated sentence analysis module, which operates to select related keywords and analyze the respective usage grammatical structure among the second-party sentences from the translated sentence acquisition module; one-word lexical modification A module that operates to replace a related keyword and a grammatical structure with a predetermined keyword and a predetermined grammatical structure in order to obtain a modified second-party sentence; and a publishing module that operates to modify the second Fangwen sentence, accept the operation instruction of the operation end, and send it to a designated end through the network.

Preferably, the system may further include a text typesetting module, which is connected to the lexical grammar modification module, and operates to apply a text typesetting format to the modified second-party sentence, and return the updated modified second-party sentence Pass the word lexical correction module.

According to the present invention, the text layout format may be a length limitation of sentence, a sentence breaking method, a punctuation method, a specified font style, a specified non-text symbol, or a combination of any two or more of the foregoing.

Preferably, the system may further include a database module, the database module is connected to the lexical correction module and the text typesetting module for storing, setting and updating the default keywords and the pre-set Set the grammatical structure and the text layout format to provide related modules.

Preferably, the system may further include a one-word sentence recommendation module that operates to correspond to all translations of a first-party word to a second-party word through statistical calculation algorithms or machine learning algorithms. Among the translation materials in a specific field, select the one with the highest frequency of use or dynamically select one of the N highest before the frequency of use, a specific second-party word corresponding to the first-party word, and a first-party sentence pattern The second-party sentence patterns corresponding to all translations of the state, through statistical calculation algorithms or machine learning algorithms, in the previous translation data in a specific field, select the one with the highest frequency of use or dynamically select the top N of the frequency of use. One and/or a specific second-party sentence pattern corresponding to the first-party sentence pattern, and the operation result is stored or updated in the database module. Where N can be 2, 3, 4 or 5, and the machine learning algorithm can be a TF-IDF algorithm.

According to the present invention, the translated sentence acquisition module, the translated sentence analysis module, the word lexical correction module, the publishing module, the text typesetting module, the database module and the word sentence recommendation module It can be the software installed in the servo host, the external board installed in the servo host, or part of the software installed in the servo host is the external board installed in the servo host. The related keyword may be a second-party word that translates all the homonyms corresponding to a first-party word. The usage grammatical structure can be formed by at least one of the following: sentence translation rules, noun singular and plural translation rules, noun masculine translation rules, and article translation rules.

The invention uses the word and word grammatical modification module to replace the translated sentences from more than three translation platforms to replace words in a specific field, so that the foregoing two problems can be solved. At the same time, the word sentence recommendation module can learn from previous translation materials, dynamically provide replaceable words and use grammatical structures, make the operation of the system more intelligent, and reduce human intervention and modification.

The present invention will be described more specifically by referring to the following embodiments.

Please refer to FIG. 1, which is a flowchart of a method for intelligently learning word editing and multilingual translation according to an embodiment of the present invention. The method is divided into several steps including steps. First, obtain at least one first sentence from a remote server via the network (S01). In the embodiments of the present specification, the first party represents the original text usage area of a sentence, and the second party represents the translated text usage area of the sentence, and the two parties use different texts. For convenience of explanation, in the following example, the first party is North America and its text is English; the second party is Taiwan and its text is Chinese. The source of the first sentence in step S01 is a remote server on the network. In practice, the at least one first-party sentence may be a blog article, periodical text data, product specification text data, or advertisement text provided by a web server in the United States. The method for obtaining the at least one first-party sentence is usually a text crawler program, which is initiated by a server host and actively obtains data from a specific or non-specific URL (Uniform Resource Locator) of the server. Here, "at least one" means that the amount of the first-party sentence obtained at one time is more than one, and "text sentence" includes articles or sentences. That is to say, the subject to be translated by the method provided by the present invention is an article or sentence, and the article or sentence can be more than one paragraph at a time. In contrast, although the present invention can also translate words or words, it is not within the scope of the claims of the present invention.

The second step of the method is to encode the at least one first-party sentence to the at least three translation platforms for translation through APIs (Application Programming Interfaces) provided by at least three translation platforms respectively. (S02). One of the spirits of the present invention is to perform the second refined translation. Specifically, at least one first-party sentence to be translated is subject to preliminary translation by a third party before these translated sentences can be processed. In practice, the at least one first-party sentence can be encoded into an identifiable code, such as Unicode, UTF-8, Big5, etc., and the format required by the API can be applied to facilitate data transmission. Of course, the choice of encoding depends on the network protocol, and the running program of the device that sends the translation data (such as the server host) can be determined automatically or set by others. The invention requires that the translated second-party sentence must come from more than 3 translation platforms (the remote specific server and the translation software running on it, the number is 3, 4...). For the convenience of explanation, we use Google translation platform, Microsoft translation platform and Baidu translation platform as examples. Each of the aforementioned has a browser version of the interface, allowing users to directly copy the original text into the translation text box for translation, and the translated text will also be displayed on the browser in time. However, the present invention is an application automation operation, so the APIs provided by the translation platforms are adopted, and the encoding of the at least one first-party sentence is concatenated to each API, and the three translation platforms obtain the codes of the three translated texts. That is, each of the at least three translation platforms obtains a code of a translated second-party sentence from the first-party sentence to the servo host (S03).

Then, relying on the servo host, the codes of these second-party sentences are converted into at least 3 corresponding second-party sentences (S04). This step is to convert the substantial content of the second-party sentence into content that can be processed by the servo host. In practice, it can also be converted into another specific code or maintain the original code to allow a specific program to process the "text".

The fifth step of this method is that the servo host selects related keywords from the second-party sentences and analyzes the respective usage grammatical structure (S05). For a better explanation of this step, please refer to Figure 4, which is an example of translation. In Figure 4, the first sentence is "The client would like to take back his assets. However, his agent didn't agree.", the translated second sentence from google translation platform is marked as "1", the content "The customer wants to get his assets back. However, his agent disagrees."; The translated second sentence from the Microsoft translation platform is marked as "2", which reads "The customer wants to take his assets Take it back, but its agent rejected it."; The translated second sentence from the Baidu translation platform is marked as "3", and the content is "The guest wants to take their property back, but their agent does not agree." According to the present invention, the servo host sets the related keyword to translate all the homonymous second-party words corresponding to a first-party word. Here, "words" includes words and words, that is, related keywords can be a single string or text of an image, or a combination of several words. The second-party words of the same part of speech in Figure 4 are surrounded by dotted lines. For example, "The client" related keywords from each translation platform are "the customer", "customer" and "guest", part of speech is a noun; "would like" related keywords from each translation platform are "want to be" , Part-of-speech is a verb phrase; "However" Related keywords from various translation platforms are "however", "but" and "but". Of course, there are many synonymous second-party words relative to the first-party words, so the related keywords are not limited to those stated in this example. At the same time, the servo host is also set to analyze the respective usage grammatical structure. The usage grammatical structure is formed by at least one of the following: sentence translation rules, noun singular and plural translation rules, noun yin and masculine translation rules, and article translation rules. The rules of sentence translation are the positions of subject words and verbs. Certain words have their specific order, but some do not. If there is no text, it must be set before analysis. The singular and plural of nouns are extremely rigorously defined in Western scripts. However, Eastern scripts, especially Chinese, appear to be unclear. For example, "apples" should be translated into apples or apples. This kind of rule should also be fixed for analysis. Furthermore, although there are rules for the translation of noun yin and masculine in every text, some texts have yin and masculine objects. The rules for the translation of these words also affect the use of grammatical structures. Finally, the article translation rules involve the specificity of an object. This correspondence is emphasized in some words and not emphasized in some words, so it needs qualitative analysis later. The best example in Figure 4 is "his". "his" can be translated into "his" or "his"; and "his" is proposed twice, can it be translated into "his" both times ", "His", or once translated into "Qi" and "His", these must be analyzed in this step to find the intended corresponding grammatical structure in the servo host.

Next, replace a related keyword and a used grammatical structure with a preset keyword and a preset grammatical structure in the servo host, so as to obtain a modified second-party sentence (S06). After the related keywords are found and the grammatical structure is analyzed, the preset keywords and grammatical structures corresponding to the presets in the servo host can be found and replaced. As shown in Figure 4, the default keyword of the server host for the associated keywords "this customer", "customer" and "guest" is "guest", so in the final translation of the revised second sentence (marked as "4" ) Uses "guest"; the default keyword for the associated keywords "asset" and "property" is "asset", so "asset" is used in the revised second sentence of the final translation; the associated keyword "however", The default keyword for "but" and "but" is "however", so use "however" in the revised second-party sentence of the final translation; the default keyword for the related keywords "agent" and "agent" is "Agent", so "Agent" is used in the revised second sentence of the final translation; the default keywords for the related keywords "disagree", "rejected" and "disagree" are "disagree", Therefore, "disagree" is used in the revised second-party sentence of the final translation. Similarly, three kinds of grammatical structures are used: "Qi...his", "his...his" and "qi...qi", and finally unified into the default grammatical structure "his...his".

The selection of preset keywords is another technical feature of the present invention. The preset keyword may be a second-party word corresponding to all translations of a first-party word, and a statistical operation algorithm or a machine learning algorithm is used to select the frequency of use in the previous translation data in a specific field The highest one or the one with the highest frequency is selected dynamically. In practice, machine learning algorithms will be the mainstay, such as the TF-IDF algorithm. The so-called specific fields here, such as a certain area, such as Taiwan; such as a certain business field, such as shoe sales; such as a certain academic field, such as the physical world. In these specific fields, the first-party text to be translated will have a specific corresponding to the second-party text, so it is necessary to accumulate the translated sentences from the field and obtain it after repeated learning. Of course, it is best to use the most frequently used words as the default keywords. However, although this single text replacement meets market demand, it is not monotonous. It is best to dynamically select one of the top 2 high, top 3 high, top 4 high, or even top 5 high frequencies of use, so that there will be 2, 3, 4 or 5 dynamic changes. The above description is programmable to select the default keyword. The invention can be used for commercialization, so if some customers need marketing to attract the attention of the audience with certain words, the default keyword may be a specific second party word that corresponds to the first party word word. For example, in any case, "I" is converted into "I" after translation.

Similarly, the preset grammatical structure can also be a second-party sentence pattern corresponding to all translations of a first-party sentence pattern, and a translation data in a specific field in the past through a statistical operation algorithm or a machine learning algorithm Among them, select the one with the highest frequency of use or dynamically select one of the top few frequencies of use, or specify a specific second-party sentence pattern corresponding to the first-party sentence pattern. Among them, the machine learning algorithm can be the TF-IDF algorithm, and the "first few highs" can refer to "top 2 highs", "top 3 highs", "top 4 highs" or "top 5 highs".

Finally, the revised second-party sentence is sent to a designated end (S07). According to the present invention, the designated end may be the URL of any host device on the network for others to access. The designated end may also be a storage device on the Internet or on the local machine (servo host) to temporarily or permanently store the modified second-party sentence.

It should be noted that the method may further include a step S06' after step S06: applying a text layout format in the servo host to the modified second-party sentence. That is, typesetting and editing the revised second-party sentence. The text layout format can be, but not limited to, the limited length of the sentence (sentences exceeding a certain number of words must be broken), the way of breaking, the use of punctuation, the specified font style, and the insertion of non-text symbols (such as adding †, ∫ Symbols unrelated to the original translation), or a combination of any two or more of the foregoing.

In another embodiment of the present invention, a system 120 for intelligently learning word editing and multilingual translation is proposed. Please see Figure 2 and Figure 3. 2 is an explanatory diagram of the operation of the system 120, and FIG. 3 is a block diagram of components of the system 120. The system 120 can be installed in a servo host 100, and the servo host 100 can be connected to several operating terminals through a network 200. The operation terminal may be any hardware device that can interact with the server host 100 and control the system 120 through the network, such as a desktop computer 201, a notebook computer 202, a tablet computer 203, or a smartphone 204. For the convenience of operation, a mobile application or software package may be installed on the operation terminal, or the operation interface provided by the system 120 may be displayed in a browser, and the operation of these software may be performed through the API provided by the system 120.

As shown in FIG. 3, the system 120 includes a translated sentence acquisition module 121, a translated sentence analysis module 122, a word lexical modification module 123, a publishing module 124, a text typesetting module 125, and a database Module 126, and one-word word sentence recommendation module 127. The connection in the figure indicates that data is transmitted through the connected or related hardware. According to the present invention, the aforementioned modules may all be software installed in the servo host 100, or may be an external board (hardware) installed in the servo host 100, or may be part of the software installed in the servo host 100 Part is an external board installed in the servo host 100. Taking software as an example, please refer to Figure 2. The hardware architecture of the servo host 100 includes a network unit 101 (including network cards, RJ45 connectors and other hardware devices), an output input unit 102 (including IO control chip set, related IO connectors, etc.) a control unit 103 (Such as a CPU and related control circuits), a memory 104 (such as a DRAM module), and a storage unit 105 (such as a hard drive, solid state drive, or disk array). The program code of the system 120 is stored in the storage unit 105 when it is not in operation. When it is to be operated, the necessary data is temporarily stored in the memory 104 through the operating system 110, and the control unit 103 reads and operates as required.

The translated sentence acquisition module 121 can accept the operation commands of the operation terminal (the unidirectional dotted arrows on the desktop computer 201, notebook computer 202, tablet computer 203 or smartphone 204 in FIG. 2 represent the command is issued), through The network 20 executes to obtain at least one first party sentence from a remote server (such as a foreign blog server 400) and encode the at least one first party sentence through at least three translation platforms (first translation The platform 401, the second translation platform 402 and the third translation platform 403), and the API provided respectively, are transferred to the at least three translation platforms for translation, and the at least three translation platforms respectively obtain the translation of the first party sentence The coding of the second second-party sentence and the coding of these second-party sentences are converted into at least 3 second-party sentences. That is, the translated sentence acquisition module 121 can complete steps S01 to S04 of the method in the previous embodiment.

The translated sentence analysis module 122 is operable to select related keywords from the second-party sentences from the translated sentence acquisition module 121 and analyze respective usage grammatical structures. Its function is to complete step S05 of the aforementioned method. The definitions of related keywords and the use of grammatical structures are as described above and will not be repeated here.

The lexical grammar modification module 123 can operate to replace a related keyword and a predetermined grammatical structure with the related keyword and use the grammatical structure in order to obtain a modified second-party sentence whose function is to complete the steps of the aforementioned method S06. The definitions and contents of the preset keywords and the preset grammatical structure are as described above, and will not be repeated here.

The issuing module 124 is operable to modify the second-party sentence, accept the operation command from the operation end, and transmit it to a designated end through the network 20 to complete step S07 of the foregoing method. The designated end is, for example, a local blog server 500, and the uploaded second-party sentence can be browsed by others.

The text typesetting module 125 is connected to the lexical modification module 123, and operates to apply a text typesetting format to the modified second-party sentence and return the updated modified second-party sentence to the lexical modification module 123. The definition of the text layout format is as described in the previous embodiment, and will not be repeated here.

The database module 106 is connected to the lexical grammar modification module 123 and the text typesetting module 125 to store, set, and update the default keywords, the default grammatical structure, and the text typesetting format to provide related modules. The database module 106 can divide multiple databases into data structures for preset keywords, preset grammatical structures, and text layout formats, respectively.

The emphasis of the present invention on intelligently learning word editing is mainly the function provided by the word sentence recommendation module 127. The word sentence recommendation module 127 can operate to translate the second party words corresponding to all the translations of a first party word through statistical calculation algorithms or machine learning algorithms in the previous translation data in a specific field, Select the one with the highest frequency of use or dynamically select one of the top few frequencies of use, specify a specific second-party word corresponding to the first-party word, and the second corresponding to all translations of a first-party sentence pattern Fangwen sentence patterns, through statistical calculation algorithms or machine learning algorithms, in the past translation materials in a specific field, select the one with the highest frequency of use or dynamically select one of the highest frequency of use and/or the specified pair A specific second-party sentence pattern of the first-party sentence pattern should be stored and updated in the database module 126. The definition of "first few highs" is as described above; in practice, the function of the word sentence recommendation module 127 is mainly based on machine learning algorithms, preferably the TF-IDF algorithm.

Although the present invention has been disclosed as above in the embodiments, it is not intended to limit the present invention. Anyone with ordinary knowledge in the technical field can make some changes and modifications without departing from the spirit and scope of the present invention. The scope of protection of the present invention shall be subject to the scope defined in the attached patent application.

100: servo host 101: network unit 102: output input unit 103: control unit 104: memory 105: storage unit 110: operating system 120: system 121: translated sentence acquisition module 122: translated sentence analysis module 123: word Lexical Modification Module 124: Publishing Module 125: Text Typesetting Module 126: Database Module 127: Word and Text Recommendation Module 20: Network 201: Desktop 202: Notebook 203: Tablet 204 : Smart phone 400: Foreign blog server 500: Local blog server

FIG. 1 is a flowchart of a method for intelligent learning word editing and multilingual translation according to an embodiment of the present invention; FIG. 2 is a method for intelligent learning word editing and multilingual translation according to another embodiment of the present invention Operation diagram of the system; Figure 3 is a block diagram of the intelligent learning word editing and multi-language translation system; and Figure 4 is an example of translation.

Claims (19)

A method for intelligently learning word editing and translation between multiple languages, including the steps of: a) obtaining at least one first-party sentence from a remote server via a network; b) encoding the at least one first-party sentence through at least APIs (Application Programming Interfaces) provided by the three translation platforms are sent to the at least three translation platforms for translation; c) the at least three translation platforms are used to obtain the translation of the first party sentence The encoding of the second second sentence to a servo host; d) The encoding of the second party sentences by the servo host into corresponding at least 3 sentences of second party sentences; e) The Select related keywords from these second-party sentences and analyze their respective use grammatical structures; f) replace a related keyword and use grammatical structure with a preset keyword and a preset grammatical structure in the servo host, So as to obtain a modified second-party sentence; and g) transmit the modified second-party sentence to a designated end. As described in item 1 of the patent application scope, the method of intelligent learning word editing and translation between multiple languages further includes a step f1) after step f): a text layout format in the servo host is applied to the amendment In the second sentence. The method of intelligent learning vocabulary editing and multi-language mutual translation as described in item 2 of the patent application scope, wherein the text layout format is sentence length limitation, sentence breaking method, punctuation method, designated use font, designated insertion of non-text symbols , Or a combination of any two or more of the foregoing. The method for compiling intelligent learning words and multi-language mutual translation as described in item 1 of the patent application scope, wherein the associated keyword is to translate all the homonym second-party words corresponding to a first-party word. The method of intelligent learning word editing and multi-language mutual translation as described in item 1 of the patent application scope, wherein the grammatical structure is formed by at least one of the following: sentence translation rules, noun singular and plural translation rules, noun yin positive translation rules Translation rules with articles. The method of intelligent learning word editing and multi-language mutual translation as described in item 1 of the patent application scope, wherein the preset keyword is a second party word corresponding to all translations of a first party word, after statistics Operational algorithm or machine learning algorithm in the past in a specific field of translation data, select the highest frequency of use or dynamically select one of the top N frequency of use, or a designated one corresponding to the first party word Specific second-party words. The method of intelligent learning word editing and multi-language mutual translation as described in item 1 of the patent scope, wherein the preset grammatical structure is a second-party sentence pattern corresponding to all translations of a first-party sentence pattern, Through statistical calculation algorithms or machine learning algorithms, in the previous translation data in a specific field, select the one with the highest frequency of use or dynamically select one of the top N of the frequency of use, or specify the first sentence pattern corresponding to the first party. A specific second-party sentence pattern. For example, the method of intelligent learning vocabulary editing and multi-language translation as described in item 6 or item 7 of the patent scope, where N is 2, 3, 4 or 5. The method of intelligent learning word editing and multi-language mutual translation as described in item 6 or item 7 of the patent application scope, wherein the machine learning algorithm is a TF-IDF algorithm. An intelligent learning word editing and translation and multi-language mutual translation system, installed in a server host, the server host is connected to an operating terminal through the network, including: a translation sentence acquisition module, accepting the operating instructions of the operating terminal, to Obtain at least one first-party sentence from a remote server via the network, and send the code of the at least one first-party sentence to the at least three translation platforms for translation through APIs provided by at least three translation platforms, Obtaining, by the at least three translation platforms, the codes of a second-party sentence after translation of the first-party sentence, and converting the codes of the second-party sentences into at least 3 second-party sentences; A translated sentence analysis module, which operates to select related keywords from the second-party sentences from the translated sentence acquisition module and analyze the respective usage grammatical structure; the one-word lexical modification module, which operates to convert A preset keyword and a preset grammatical structure to replace the associated keyword and use the grammatical structure in order to obtain a modified second-party sentence; and a publishing module that operates to modify the second-party sentence and accept the operation The operation command of the terminal is sent to a designated terminal through the network. As described in item 10 of the patent application scope, the system of intelligent learning word editing and translation between multiple languages further includes a text typesetting module, which is connected to the word grammar correction module, and operates to apply a text typesetting format to the In the correction of the second-party sentence, the updated modified second-party sentence is returned to the lexical correction module. The system of intelligent learning vocabulary editing and multi-language mutual translation as described in item 11 of the patent application scope, in which the text layout format is sentence length limitation, sentence breaking method, punctuation method, designated use font, designated insertion of non-text symbols , Or a combination of any two or more of the foregoing. The system for editing and editing intelligent learning words and translating multiple languages as described in item 11 of the scope of patent application further includes a database module, the database module and the lexical grammar modification module and the text typesetting module The connection is used to store, set and update the preset keyword, the preset grammatical structure and the text layout format to provide related modules. The system of intelligent learning word editing and multi-language mutual translation as described in item 13 of the patent scope further includes a one-word sentence sentence recommendation module, which operates to correspond to all translations of a first-party word to a second party Words, through statistical calculation algorithms or machine learning algorithms, in the past in a particular field of translation data, select the most frequently used or dynamically select one of the top N of the most frequently used frequency, and the first word corresponding to the specified A specific second-party word of a word, a second-party sentence pattern corresponding to all translations of a first-party sentence pattern, and translation data in a specific field in the past through a statistical operation algorithm or a machine learning algorithm Select the one with the highest frequency of use or dynamically select one of the N highest before the frequency of use and/or specify a specific second-party sentence pattern corresponding to the first-party sentence pattern, and store or update the operation results in In the database module. The system of intelligent learning word editing and multi-language mutual translation as described in item 14 of the patent application scope, wherein the translated sentence acquisition module, the translated sentence analysis module, the word lexical modification module, the release module , The text typesetting module, the database module and the word and sentence recommendation module are software installed in the server host, external boards installed in the server host, or part of the software part installed in the server host It is an external board installed in the servo host. The system of intelligent learning words editing and multi-language translation as described in item 14 of the patent application scope, where N is 2, 3, 4 or 5. The system of intelligent learning word editing and multi-language mutual translation as described in item 14 of the patent application scope, wherein the machine learning algorithm is a TF-IDF algorithm. The intelligent learning word editing and multilingual translation system as described in item 10 of the patent application scope, wherein the related keyword is to translate all the homonym second-party words corresponding to a first-party word. The system of intelligent learning word editing and multi-language mutual translation as described in item 10 of the patent application scope, wherein the used grammatical structure is formed by at least one of the following: sentence translation rules, noun singular and plural translation rules, noun yin positive translation rules Translation rules with articles.

Images