本案係為一種助理服務系統，尤指具有對財務會計資料進行分析及完成問答的助理服務系統。This case is an assistant service system, especially an assistant service system that can analyze financial accounting data and complete question and answer.
中小企業通常沒有專業財務會計人員的編制，兼任人員常因不熟悉記帳技巧而記載用詞不當，導致後續資料整合而產生記錯帳、財務數據不正確以及多繳稅等風險。再者，中小企業的經營者即使得到詳細財報內容，也不見得有能力可以從中得到所需資訊，仍然需要專業的財務專家來對大量資料進行分析與正確地回答問題，但是在現實狀況中，中小企業的員工不容易有財務專家的配置。Small and medium-sized enterprises usually do not have professional financial accountants. Part-time accounting personnel often use inappropriate terms when recording due to unfamiliarity with bookkeeping techniques, which leads to risks such as mis-recording, inaccurate financial data, and overpayment of taxes when the subsequent data is integrated. Furthermore, even if the operators of small and medium-sized enterprises receive detailed financial statements, they may not be able to obtain the required information from them. They still need professional financial experts to analyze large amounts of data and answer questions correctly. However, in reality, it is not easy for small and medium-sized enterprises to have financial experts on staff.
而為能改善上述種種習用手段的缺失以及滿足可快速地對大量財務會計資料進行分析與正確地回答問題的需求，本案係發展出一種助理服務系統，提供給至少一第一用戶與一第二用戶進行信號連結，該系統包含：一服務對話模組，用以分別與該第一用戶與該第二用戶完成信號連結，分別因應該第一用戶所發出之一第一問題與該第二用戶所發出之一第二問題來建立一第一對話串以及一第二對話串；以及一線上資料庫，信號連結至該服務對話模組，其中至少包含一第一用戶專屬資料庫與一第二用戶專屬資料庫，該服務對話模組建立該第一對話串時僅參考該第一用戶專屬資料庫中的內容，該服務對話模組建立該第二對話串時僅參考該第二用戶專屬資料庫中的內容。In order to improve the deficiencies of the above-mentioned various means of practice and to meet the needs of quickly analyzing a large amount of financial accounting data and correctly answering questions, this case is to develop an assistant service system, which is provided to at least one first user and one second user for signal connection. The system includes: a service dialogue module, which is used to complete the signal connection with the first user and the second user respectively, and to respond to a first question issued by the first user and the second user respectively. A second question posed by the user is used to establish a first conversation thread and a second conversation thread; and an online database, which is signal-linked to the service dialogue module, and which at least includes a first user-specific database and a second user-specific database. When the service dialogue module establishes the first conversation thread, it only refers to the content in the first user-specific database, and when the service dialogue module establishes the second conversation thread, it only refers to the content in the second user-specific database.
根據上述構想，本案所述之助理服務系統，其中該第一用戶或該第二用戶係為具有一可辨識用戶身份資料之一個人電腦或是一智慧型手機，該服務對話模組根據該可辨識用戶身份資料而選擇僅使用該第一用戶專屬資料庫或該第二用戶專屬資料庫。According to the above concept, in the assistant service system described in this case, the first user or the second user is a personal computer or a smart phone with an identifiable user identity data, and the service dialogue module chooses to use only the first user's exclusive database or the second user's exclusive database based on the identifiable user identity data.
根據上述構想，本案所述之助理服務系統，其中該第一用戶或該第二用戶係為具有一可辨識用戶身份資料之一個人電腦或是一智慧型手機，該服務對話模組根據該可辨識用戶身份資料而提供對應該第一用戶或該第二用戶之一問答權限，該問答權限中包含有限制該服務對話模組僅能觸及該第一用戶專屬資料庫中或該第二用戶專屬資料庫中的一預設範圍。According to the above concept, in the assistant service system described in this case, the first user or the second user is a personal computer or a smart phone with an identifiable user identity data, and the service dialogue module provides a question and answer permission corresponding to the first user or the second user according to the identifiable user identity data, and the question and answer permission includes limiting the service dialogue module to only access a preset range in the first user's exclusive database or the second user's exclusive database.
根據上述構想，本案所述之助理服務系統，其中該第一用戶專屬資料庫或該第二用戶專屬資料庫中包含有已預先設計成的資料表(tables)或視觀表（Views）的複數個進階財務指標，用以讓該服務對話模組進行資料擷取。According to the above concept, in the assistant service system described in this case, the first user-specific database or the second user-specific database contains a plurality of advanced financial indicators of pre-designed data tables or view tables for the service dialogue module to capture data.
根據上述構想，本案所述之助理服務系統，其中該服務對話模組包含有至少一語言模型，該語言模型用以分別與該第一用戶與該第二用戶完成信號連結，分別因應該第一用戶所發出之該第一問題與該第二用戶所發出之該第二問題來建立該第一對話串以及該第二對話串。According to the above concept, the assistant service system described in this case, wherein the service dialogue module includes at least one language model, and the language model is used to complete the signal connection with the first user and the second user respectively, and establish the first dialogue string and the second dialogue string respectively in response to the first question posed by the first user and the second question posed by the second user.
根據上述構想，本案所述之助理服務系統，其中該服務對話模組中更包含一流程管理器，來將該語言模型工作流程中的不同流程進行分類，並將該語言模型與該第一用戶專屬資料庫與該第二用戶專屬資料庫連接起來。According to the above concept, the assistant service system described in this case further includes a process manager in the service dialogue module to classify different processes in the language model workflow and connect the language model with the first user-specific database and the second user-specific database.
根據上述構想，本案所述之助理服務系統，其中該服務對話模組與該第一用戶或該第二用戶完成連接時，先進行該第一用戶或該第二用戶的身份認證，而在完成該第一用戶或該第二用戶的身份認證後，確認其合法用戶身份甚至是職務身份後，才能進行後續的對話串建立過程。According to the above concept, in the assistant service system described in this case, when the service dialogue module completes the connection with the first user or the second user, the identity of the first user or the second user is first authenticated. After the identity authentication of the first user or the second user is completed and their legal user identity or even job identity is confirmed, the subsequent dialogue string establishment process can be carried out.
根據上述構想，本案所述之助理服務系統，其中更包含一資料處理引擎，而該第一用戶專屬資料庫為專屬該第一用戶的一第一會計及企業營運相關資料庫，其中資料內容可以包含銷售客戶地區以及產品庫存，該第二用戶專屬資料庫為專屬該第二用戶的一第二會計及企業營運相關資料庫，而該第一會計及企業營運相關資料庫或該第二會計及企業營運相關資料庫的更新方法係包含下列步驟：匯入一交易資料檔案至該資料處理引擎，該交易資料檔案中包含複數筆資料組合，該複數筆資料組合中之一第一筆資料組合中至少包含有一非會計用語之文字描述與一數值；該資料處理引擎將該交易資料檔案中之該第一筆資料組合中之該非會計用語之文字描述自動轉換成一會計分類用語之描述，進而產生一第一筆處理過資料組合；以及利用該第一筆處理過資料組合來對該第一用戶的該第一會計及企業營運相關資料庫或該第二用戶的該第二會計及企業營運相關資料庫進行內容更新。According to the above concept, the assistant service system described in this case further includes a data processing engine, and the first user-specific database is a first accounting and business operation-related database dedicated to the first user, wherein the data content may include sales customer regions and product inventory, and the second user-specific database is a second accounting and business operation-related database dedicated to the second user, and the updating method of the first accounting and business operation-related database or the second accounting and business operation-related database includes the following steps: importing a transaction data file into the data processing engine, the transaction The data file includes a plurality of data combinations, wherein a first data combination in the plurality of data combinations includes at least a text description of non-accounting terms and a numerical value; the data processing engine automatically converts the text description of the non-accounting terms in the first data combination in the transaction data file into a description of accounting classification terms, thereby generating a first processed data combination; and the first processed data combination is used to update the content of the first accounting and business operation related database of the first user or the second accounting and business operation related database of the second user.
根據上述構想，本案所述之助理服務系統，其中該資料處理引擎為一人工智慧資料處理引擎，該人工智慧資料處理引擎根據一語言模型而將該第一筆資料組合中之該非會計用語之文字描述自動轉換成該會計分類用語之描述，該資料處理引擎所產生之該第一筆處理過資料組合，其係符合一逗號分隔值格式，該會計分類用語之描述被放在一特定欄位中。According to the above concept, the assistant service system described in this case, wherein the data processing engine is an artificial intelligence data processing engine, the artificial intelligence data processing engine automatically converts the text description of the non-accounting term in the first data combination into the description of the accounting classification term according to a language model, the first processed data combination generated by the data processing engine conforms to a comma separated value format, and the description of the accounting classification term is placed in a specific field.
根據上述構想，本案所述之助理服務系統，其中該交易資料檔案中可包含用文字、語音或影像來表達的該非會計用語之文字描述與該數值，該第一筆資料組合中之該數值為一時間數值與一金額數值中之一，而該非會計用語之文字描述為一費用描述文字與一支付形態描述文字中之一，該會計分類用語之描述為一會計代碼，該交易資料檔案先被轉換成一逗號分隔值格式後再匯入該資料處理引擎，該非會計用語之文字描述被編在一特定欄位中，使該資料處理引擎能將放在該特定欄位內的非會計用語之文字描述自動轉換成該會計分類用語之描述。According to the above concept, the assistant service system described in this case, wherein the transaction data file may include the text description of the non-accounting term and the numerical value expressed by text, voice or image, the numerical value in the first data combination is one of a time value and an amount value, and the text description of the non-accounting term is one of an expense description text and a payment form description text, the description of the accounting classification term is an accounting code, the transaction data file is first converted into a comma separated value format and then imported into the data processing engine, the text description of the non-accounting term is compiled in a specific field, so that the data processing engine can automatically convert the text description of the non-accounting term placed in the specific field into the description of the accounting classification term.
為了能對本發明之上述構想有更清楚的理解，下文特舉出多個實施例，並配合對應圖式詳細說明如下。In order to provide a clearer understanding of the above concept of the present invention, a number of embodiments are specifically cited below and described in detail with corresponding drawings.
為了解決上述的問題，本案發明人係發展出如圖1A所示之關於一種會計資料處理方法的步驟流程示意圖，而本案之會計資料處理方法，其可運作在多種資訊系統之上，例如是運作在設於遠端的資料服務伺服器。但為能方便說明，本案係以圖1B所示之關於一種會計資料處理系統的方塊示意圖為例來進行細節描述。圖1B所示之該會計資料處理系統中至少包含有一交易資料檔案100、一用戶10、對應至該用戶10的一會計資料庫13、一資料服務伺服器14以及一資料處理引擎15。而上述之交易資料檔案中可包含用文字、語音或影像來表達的該非會計用語之文字描述與該數值，也就是可以用多種格式所記錄而形成包含有原始交易憑證記錄的資料檔案，例如：交易記錄語音檔、交易記錄資料檔、日記帳、分類帳、傳票影像、試算表或是會計表報。而透過格式轉換程式，便可以將語音或影像轉換成文字來進行後續處理。或是當資料處理引擎15足夠強大，也是可以直接對語音或影像進行解讀。但為說明上的簡潔起見，上述之交易資料檔案是已該用戶所記錄完成之日記帳結果所形成之一日記帳資料檔案為例來進行說明。In order to solve the above problems, the inventor of this case has developed a step flow diagram of an accounting data processing method as shown in FIG1A. The accounting data processing method of this case can be operated on a variety of information systems, for example, on a data service server located at a remote end. However, for the convenience of explanation, this case is described in detail using a block diagram of an accounting data processing system as shown in FIG1B. The accounting data processing system shown in FIG1B at least includes a transaction data file 100, a user 10, an accounting database 13 corresponding to the user 10, a data service server 14, and a data processing engine 15. The transaction data file mentioned above may include the text description and the numerical value of the non-accounting term expressed in text, voice or image, that is, it may be recorded in a variety of formats to form a data file containing the original transaction voucher record, such as: transaction record voice file, transaction record data file, journal, ledger, voucher image, spreadsheet or accounting report. Through the format conversion program, the voice or image can be converted into text for subsequent processing. Or when the data processing engine 15 is powerful enough, the voice or image can also be directly interpreted. However, for the sake of simplicity in the explanation, the transaction data file mentioned above is an example of a journal data file formed by the journal result recorded by the user.
因為以中小企業而言，記錄該日記帳資料檔案的執行者通常是兼職記帳人員或是請款當事人。而該日記帳資料檔案中的複數筆筆資料組合，可能還是由記帳人員或是多個請款當事人所分批輸入，所以極可能產生使用非會計用語之文字描述內容。換句話說，而該複數筆資料組合中之任何一筆資料組合的內容，則可能是包含有至少一非會計用語之文字描述與至少一數值。例如，該筆資料組合中所包含的該數值或多個數值可以包含有一時間數值與一金額數值中之一或其任意組合，而該非會計用語之文字描述則可以是一費用描述文字與一支付形態描述文字中之一。例如，費用描述文字被輸入為”計程車費”、”高鐵票”等口語化文字描述，而支付形態描述文字可以是”付現” 、”轉帳” 等口語化文字描述，而時間數值為20230920(代表2023年9月20日)、金額數值為354(代表金額)。Because for small and medium-sized enterprises, the executors who record the daily book data file are usually part-time bookkeepers or payment requesters. The multiple data combinations in the daily book data file may be entered in batches by bookkeepers or multiple payment requesters, so it is very likely that text descriptions using non-accounting terms will be generated. In other words, the content of any data combination in the multiple data combinations may contain at least one text description in non-accounting terms and at least one value. For example, the value or multiple values contained in the data combination may include one or any combination of a time value and an amount value, and the text description in non-accounting terms may be one of an expense description text and a payment form description text. For example, the expense description text is entered as colloquial text descriptions such as "taxi fare" and "high-speed rail ticket", while the payment form description text can be colloquial text descriptions such as "cash payment" and "transfer", and the time value is 20230920 (representing September 20, 2023) and the amount value is 354 (representing the amount).
而在該用戶每天所記錄之該日記帳資料檔案(即該交易資料檔案100)被完成後，便可能會有該非會計用語之文字描述參雜其中，於是便可以利用本案所發展出來的會計資料處理方法來進行處理。而如圖1A所示之關於一種會計資料處理方法主要可包含下列步驟：首先，可透過資料服務伺服器14來讀取該用戶10每天所記錄之該日記帳資料檔案(即該交易資料檔案100)，然後再將該交易資料檔案100匯入一資料處理引擎15(步驟11)，例如該交易資料檔案100中包含有複數筆筆資料組合，而該交易資料檔案100得格式可以先被該資料服務伺服器14轉換成逗號分隔值格式(（Comma-Separated Values，CSV）後再送入該資料處理引擎15，而上述之非會計用語之文字描述被則編在一特定欄位中。After the daily journal data file (i.e., the transaction data file 100) recorded by the user is completed, there may be text descriptions of non-accounting terms mixed therein, so the accounting data processing method developed in this case can be used to process it. As shown in FIG. 1A , an accounting data processing method mainly includes the following steps: first, the daily journal data file (i.e., the transaction data file 100) recorded by the user 10 every day can be read through the data service server 14, and then the transaction data file 100 is imported into a data processing engine 15 (step 11). For example, the transaction data file 100 includes a plurality of data combinations, and the format of the transaction data file 100 can be first converted into a comma-separated value format (CSV) by the data service server 14 and then sent to the data processing engine 15, and the above-mentioned non-accounting terminology text description is then compiled in a specific field.
而該資料處理引擎15可以是一大型語言模型的人工智慧資料處理引擎(例如雲端上的ChatGPT或其它類似的而設於近端或遠端的人工智慧引擎)。然後透過輸入指令(prompt)，便可使該資料處理引擎15能將一第一筆資料組合中，放在該特定欄位內的非會計用語之文字描述自動轉換成一會計分類用語之描述，進而產生一第一筆處理過資料組合(步驟12)。而上述之會計分類用語之描述可以是一會計代碼。例如，可將口語化之費用描述文字自動轉換成符合IFRS四碼表或是符合台灣中小企業之「企業會計準則」(簡稱EAS)的費用描述。接著，便可再利用該第一筆處理過資料組合，來對該用戶的該會計資料庫13進行內容更新(步驟13)。The data processing engine 15 can be an artificial intelligence data processing engine with a large language model (such as ChatGPT on the cloud or other similar artificial intelligence engines located at the near end or the far end). Then, by inputting a command (prompt), the data processing engine 15 can automatically convert the text description of non-accounting terms placed in the specific field in a first data combination into a description of accounting classification terms, thereby generating a first processed data combination (step 12). The above-mentioned description of accounting classification terms can be an accounting code. For example, the colloquial expense description text can be automatically converted into an expense description that complies with the IFRS four-code table or the "Enterprise Accounting Standards" (abbreviated as EAS) for small and medium-sized enterprises in Taiwan. Then, the first processed data combination can be used to update the content of the user's accounting database 13 (step 13).
而將口語化之費用描述文字自動轉換成符合會計準則費用描述的例子，可以是像：”計程車費”自動轉換成”6114 旅費”、”付現”自動轉換成” 1111庫存現金”與”轉帳”自動轉換成” 1113銀行存款”。至於上述之人工智慧資料處理引擎，便可以是利用ChatGPT或類似的大型語言生成式模型來完成，並可以對其輸入下列prompt指令：『使用IFRS四碼表或EAS匯入生成式模型、角色扮演為會計師、以CSV格式匯出以下欄位"分錄id, 借方會計代號, 借方科目中文名稱, 借方科目金額, 貸方會計代號, 貸方科目中文名稱, 貸方科目金額"』。於是，人工智慧資料處理引擎便會自動生成多筆處理過資料組合，進而組成一個處理過的日記帳會計檔案，其內容可以是如圖2所示之資料格式示例圖。 如此一來，在所有複數筆筆資料組合皆被處理過後，某一筆處理過資料組合的內容便是"分錄id, 借方會計代號, 借方科目中文名稱, 借方科目金額, 貸方會計代號, 貸方科目中文名稱, 貸方科目金額"，形成內容完全符合會計分類用語，且資料格式符合逗號分隔值格式（Comma-Separated Values，CSV）的日記帳結果。Examples of automatically converting colloquial expense descriptions into expense descriptions that comply with accounting standards include: "taxi fare" is automatically converted into "6114 travel expenses", "cash payment" is automatically converted into "1111 cash on hand", and "transfer" is automatically converted into "1113 bank deposit". As for the above-mentioned artificial intelligence data processing engine, it can be completed using ChatGPT or similar large-scale language generative models, and the following prompt command can be input into it: "Use IFRS four-code table or EAS to import the generative model, role-play as an accountant, and export the following fields in CSV format "entry id, debit accounting code, debit account Chinese name, debit account amount, credit accounting code, credit account Chinese name, credit account amount"". Therefore, the artificial intelligence data processing engine will automatically generate multiple processed data combinations, and then form a processed journal accounting file, the content of which can be the data format example shown in Figure 2. In this way, after all multiple data combinations have been processed, the content of a processed data combination is "entry id, debit accounting code, debit account Chinese name, debit account amount, credit accounting code, credit account Chinese name, credit account amount", forming a journal result that fully complies with accounting classification terms and the data format complies with the comma-separated values (CSV) format.
至於步驟13中，利用該筆或多筆的處理過資料組合來對該用戶的該會計資料庫進行內容更新的細節則可以是：資料服務伺服器14可利用一提取轉換載入程式（extract, transform, load，簡稱ETL）來將包含多筆處理過資料組合的日記帳結果進行處理，對上述處理過資料組合資料進行擷取（extract）、轉換（transform），最後再載入（load）至目的端（例如對應該用戶的一雲端會計資料庫）。但是傳統的提取轉換載入程式（ETL）並不能妥善處理人工智慧資料處理引擎所可能產生的錯誤。As for the details of using the one or more processed data combinations to update the content of the user's accounting database in step 13, the data service server 14 can use an extract, transform, load (ETL) program to process the journal results containing multiple processed data combinations, extract, transform the processed data combination data, and finally load it to the destination (such as a cloud accounting database corresponding to the user). However, the traditional extract, transform, load program (ETL) cannot properly handle the errors that may be generated by the artificial intelligence data processing engine.
因此，能強化本案效能的提取轉換載入程式被提出，可以由該資料服務伺服器14來執行。其流程步驟示意圖如圖3之所示，其主要包含四個步驟：首先是擷取步驟(步驟31)，其實例可以是『利用搜尋演算法擷取出僅包含完整日記帳結果』，其細節可包含下列步驟：(A)先找"分錄id"欄位，再找到本日記帳應出現之最後一個分錄id號碼。(B)從該分錄id號碼位置找是否存在有”\n”符號，若沒有”\n”符號則將所有擷取。若有則僅擷取到\n之前一個index的內容。Therefore, an extraction conversion loader that can enhance the performance of this case is proposed and can be executed by the data service server 14. The process step diagram is shown in Figure 3, which mainly includes four steps: First, there is the extraction step (step 31), and its example can be "using the search algorithm to extract only the complete diary results", and its details can include the following steps: (A) First find the "entry id" field, and then find the last entry id number that should appear in this diary. (B) Find whether there is a "\n" symbol from the entry id number position. If there is no "\n" symbol, all will be extracted. If there is, only the content of the index before \n will be extracted.
再來是檢查是否出錯的步驟(步驟32)：檢查是否出錯的實例可以是『利用一檢查器程式(checker)確認輸出結果正常可運行，針對常見錯誤進行處理』，若有出錯則回到圖1中的步驟12，讓該資料處理引擎對該原始資料重新再生成一次日記帳檔案。而檢查器程式的步驟細節可包含：檢查各項明顯錯誤，例如分錄id編列錯誤、逗號分隔值格式(CSV)格式出錯、產出資料內容錯誤等。如果這些錯誤檢查都通過，檢查器程式會回傳成功，則進入下一筆資料進行處理。如果有其中一個錯誤檢查不通過，就會讓資料處理引擎再生成一次，直到回傳成功，或是這個迴圈運行超過3次後，就會跳過這筆資料列為錯誤，不會列入日記帳中，並提示用戶或後台程式管理員進錯誤內容的調整。Next is the step of checking for errors (step 32): An example of checking for errors may be "using a checker program (checker) to confirm that the output result is normal and can be run, and to handle common errors". If there is an error, return to step 12 in Figure 1 and let the data processing engine regenerate a journal file for the original data. The step details of the checker program may include: checking for various obvious errors, such as entry id compilation errors, comma separated value format (CSV) format errors, output data content errors, etc. If these error checks are passed, the checker program will return success, and then enter the next data for processing. If one of the error checks fails, the data processing engine will generate it again until the return is successful, or after the loop runs more than 3 times, the data will be skipped and listed as an error, and will not be included in the journal. The user or background programmer will be prompted to adjust the error content.
至於轉換（transform）步驟(步驟33)的實例則可以是『利用演算法將日記帳結果轉換為資料庫格式之資料列』，其可包含下列步驟：(A)將每筆生成式AI產出之日記帳資料轉換為一筆會計科目，其中於每一個日記帳分錄皆為一筆資料列，用以符合該資料庫的欄位。(B)取得基本資料(交易時間、公司ID、原始交易資料ID、交易資料備註)。An example of the transform step (step 33) may be "using an algorithm to transform the journal results into a data row in a database format", which may include the following steps: (A) transforming each journal data generated by the generative AI into an accounting account, where each journal entry is a data row to match the database field. (B) obtaining basic data (transaction time, company ID, original transaction data ID, transaction data notes).
而載入（load）步驟(步驟34)的實例則可以是『填入每一筆日記帳ID、分錄ID、分錄內科目ID』，而在載入後還可以使用下列步驟來處理三種情況：(A)正常情況 : 借方科目數量與貸方科目數量相等，(B) 科目數量不對稱情況：兩方科目數量不相等，(C)錯誤情況：借方或貸方某一方或兩方各自金額、科目數量無法對稱，則重新生成。其中當(B) 科目數量不對稱情況發生時，表示生成科目數量不對稱，例如下列三種情況：（1）貸方或借方科目「名稱」與「金額數量」不一致、（2）貸方或借方科目「名稱」與「四級會計科目代碼」數量不一致以及（3）貸方或借方科目「金額」與「四級會計科目代碼」數量不一致。當步驟12所生成的結果，出現上述(B) 科目數量不對稱情況中的三種情況時，就會進行特別處理。而特別處理可以是將其應出現而未出現的項目內容填上「0000」，讓後端工程師知道這項內容出錯，並顯示到前端來進行手動處理。An example of the load step (step 34) is to "fill in each journal ID, entry ID, and account ID within the entry." After loading, the following steps can be used to handle three situations: (A) Normal situation: the debit account quantity is equal to the credit account quantity, (B) Asymmetric account quantity situation: the account quantities on both sides are not equal, (C) Error situation: the debit or credit amount or both sides are not symmetric, so regenerate. When (B) the asymmetric account quantity occurs, it means that the generated account quantity is asymmetric, such as the following three situations: (1) the "name" and "amount" of the credit or debit account are inconsistent, (2) the "name" and "four-level accounting account code" of the credit or debit account are inconsistent, and (3) the "amount" and "four-level accounting account code" of the credit or debit account are inconsistent. When the result generated in step 12 shows any of the above (B) asymmetric account quantity, special processing will be performed. The special processing can be to fill in "0000" for the item content that should appear but does not appear, so that the back-end engineer knows that the content is wrong and displays it to the front-end for manual processing.
以上步驟再舉一實例來詳細說明：當正常情況時，透過步驟12自動生成：借方科目名稱：交際費、交通費、借方科目四級會計代碼：5012、5013以及借方科目金額：2000、500。但若透過步驟12自動生成的內容為：借方科目名稱：交際費、交通費、借方科目四級會計代碼：5012以及借方科目金額：2000、500。由上考看出自動生成的會計代碼少一個，而這便是一種錯誤情況，因此我們會將其應出現而未出現的項目內容填上「0000」，讓後端工程師知道這項內容出錯，並顯示到前端工程師，讓其知曉並可以進行手動處理。Let's take another example to explain the above steps in detail: Under normal circumstances, the following debit account names are automatically generated through step 12: Social expenses, transportation expenses, debit account level 4 accounting codes: 5012, 5013, and debit account amounts: 2000, 500. However, if the content automatically generated through step 12 is: Debit account names: Social expenses, transportation expenses, debit account level 4 accounting codes: 5012, and debit account amounts: 2000, 500. From the above test, we can see that the automatically generated accounting code is missing one, which is an error. Therefore, we will fill in "0000" for the item that should appear but does not appear, so that the back-end engineer knows that there is an error in this content, and display it to the front-end engineer so that he can know and handle it manually.
另外，本案還可以外加”折舊處理”的步驟，其細節可以是：當取得固定資產買賣紀錄表中關於本月賣出、買入之交易資料，以及上個月狀態為持有之交易資料。使用者便可依照行政院固定資產編號建議表輸入該物品之固定資產建議表，本案系統便可利用製作好的對照表映射至對應的IFRS會計四級科目或EAS科目，自動填入會計科目名稱與會計代碼。其實例可以有下列步驟：In addition, this case can also add the step of "depreciation processing", the details of which can be: when obtaining the transaction data of the fixed asset sales record table about the sale and purchase of this month, and the transaction data of the status of holding last month. The user can enter the fixed asset recommendation table of the item according to the Executive Yuan fixed asset number recommendation table, and the system of this case can use the prepared comparison table to map to the corresponding IFRS accounting level 4 account or EAS account, and automatically fill in the accounting account name and accounting code. The example can have the following steps:
步驟1 - 針對買入狀態之資料列進行處理：將使用者輸入該物品之原價值、殘值、折舊年限、折舊法存入資料庫。該物品狀態改為持有，累積折舊為0、淨值等於原價值。判斷買入時間：若為本月15日以前買入，則需要進行一次折舊處理；若15日以後買入，則不需進行折舊處理。買入所需折舊處理：確認折舊期限未到期；利用使用者之折舊法進行折舊計算，更新到本月該資料之累計折舊、淨值；將該月產生折舊費用與該物品對應的累計折舊輸入至日記帳。Step 1 - Process the data row of the purchase status: store the original value, residual value, depreciation period, and depreciation method entered by the user into the database. The status of the item is changed to holding, the accumulated depreciation is 0, and the net value is equal to the original value. Determine the purchase time: if it is purchased before the 15th of this month, a depreciation process is required; if it is purchased after the 15th, no depreciation process is required. Depreciation processing required for purchase: confirm that the depreciation period has not expired; use the user's depreciation method to calculate depreciation, and update the accumulated depreciation and net value of the data to this month; enter the depreciation expense generated in this month and the accumulated depreciation corresponding to the item into the daily book.
步驟2 - 針對賣出狀態之資料列進行處理：尋找該物品最新持有狀態內，是否有與該賣出物品的資料，若有則繼續，若無則跳出會計處理。判斷賣出時間：若為本月15日以後賣出，則需要進行一次折舊處理；若15日以前賣出，則不須進行本月折舊處理。賣出所需折舊處理：確認折舊期限未到期；利用使用者之折舊法進行折舊計算，更新到本月該資料之累計折舊、淨值。將該月產生折舊費用與該物品對應的累計折舊輸入至日記帳。取出該物品最新持有之資料列，取出該累計折舊、殘值、淨值等等，進行賣出處理，並列入日記帳中。將該持有資料列排除，不保留至下個月持有狀態列。Step 2 - Process the data column of the sold status: Find whether there is data related to the sold item in the latest holding status of the item. If there is, continue; if not, exit the accounting process. Determine the time of sale: If it is sold after the 15th of this month, a depreciation process is required; if it is sold before the 15th, no depreciation process is required for this month. Depreciation process required for sale: Confirm that the depreciation period has not expired; use the user's depreciation method to calculate depreciation, and update the accumulated depreciation and net value of the data to this month. Enter the depreciation expense generated this month and the accumulated depreciation corresponding to the item into the daily journal. Take out the latest data column of the item, take out the accumulated depreciation, residual value, net value, etc., perform the sale process, and include it in the daily journal. Exclude the holding data row and do not keep it in the holding status row for the next month.
步驟3 - 針對持有狀態之資料列進行折舊處理：檢查折舊期限是否到期，無則進行折舊處理；繼承至下個月的固定資產紀錄表，並將折舊結果認列至日記帳。而經過上述三個步驟後，便可以準確地自動完成折舊的計算與資料更新。Step 3 - Depreciation processing for the data in the holding status: check whether the depreciation period has expired, if not, depreciation processing; inherit to the fixed asset record sheet of the next month, and recognize the depreciation results in the journal. After the above three steps, the depreciation calculation and data update can be completed accurately and automatically.
另外，為能將日記帳自動轉換成損益表，本案還提供下列方法步驟：利用IFRS編碼規則或EAS編碼規則，自日記帳取會計代碼第一碼為5~8者進行損益表科目認列。為能將日記帳自動轉換成資產負債表，本案還提供下列方法步驟：取出上個月份之資產負債表作為底進行更動；利用IFRS編碼規則或EAS編碼規則，自日記帳取會計代碼第一碼為1~3者進行資產負債表科目認列；取出本期折舊結果；進行認列。In addition, in order to automatically convert the journal into the income statement, this case also provides the following method steps: using the IFRS coding rules or EAS coding rules, take the first digit of the accounting code from the journal to 5-8 to recognize the income statement account. In order to automatically convert the journal into the balance sheet, this case also provides the following method steps: take the balance sheet of the previous month as the base for modification; use the IFRS coding rules or EAS coding rules to take the first digit of the accounting code from the journal to 1-3 to recognize the balance sheet account; take out the depreciation result of the current period; and recognize it.
再者，本案還能將損益表與資產負債表，利用間接法生成現金流量表，其步驟包含：製作現金流量表與損益表、資產負債表之科目轉換表；為現金流量表會計科目標上自製編號，為9開頭；列出現金流量表所需IFRS四級會計科目開頭為1~8所需會計科目代碼與名稱以及其於現金流量表之計算方式(預設為資產為減、負債與權益為加)；於損益表、資產負債表抓取所需科目金額；按照現金流量表編號，依照預設計算方式進行科目聚合。Furthermore, this case can also generate a cash flow statement from the income statement and balance sheet using the indirect method. The steps include: preparing the account conversion table of the cash flow statement, the income statement, and the balance sheet; marking the accounting accounts of the cash flow statement with self-made numbers starting with 9; listing the required IFRS level 4 accounting accounts starting with 1 to 8 required for the cash flow statement, and the required accounting account codes and names and their calculation methods in the cash flow statement (the default is assets as a reduction and liabilities and equity as a plus); capturing the required account amounts in the income statement and balance sheet; and aggregating the accounts according to the cash flow statement number and the preset calculation method.
至於”前處理”的細節則可是：轉換為資料庫格式或轉換為輸出PDF格式下列步驟：轉換為資料庫格式：1.填入現金流量表基本資料(公司ID、本表開始認列時間、本表結束認列時間)；2.填入現金流量表代碼、科目名稱、金額。轉換為輸出PDF格式：按照正式現金流量樣式，將特定科目進行縮排處理。As for the details of "pre-processing", it can be: Convert to database format or convert to output PDF format. The following steps: Convert to database format: 1. Fill in the basic data of the cash flow statement (company ID, start recognition time of this statement, end recognition time of this statement); 2. Fill in the cash flow statement code, account name, and amount. Convert to output PDF format: Indent specific accounts according to the formal cash flow pattern.
而以上述會計資料處理方法與系統所建立完成之該會計資料庫13，還可以利用下列技術手段所描述的助理服務系統來提供進一步的服務。如圖4之所示，本案所發展出來的助理服務系統4可提供給複數個用戶來使用，而該助理服務系統4至少包含有一服務對話模組41以及一線上資料庫42，該線上資料庫42的一實例便可以是上述會計資料庫13，當然也可以是其他方式建置完成的各類資料庫，例如及企業營運相關資料庫(包含銷售客戶地區產品庫存等)。其中該服務對話模組41用以分別與複數個用戶完成信號連結(例如透過網際網路)。The accounting database 13 established by the above accounting data processing method and system can also provide further services by using the assistant service system described by the following technical means. As shown in FIG. 4 , the assistant service system 4 developed in this case can be provided to a plurality of users, and the assistant service system 4 at least includes a service dialogue module 41 and an online database 42. An example of the online database 42 can be the above accounting database 13, and of course it can also be various databases established by other means, such as enterprise operation-related databases (including sales customer area product inventory, etc.). The service dialogue module 41 is used to complete signal connections with a plurality of users respectively (for example, through the Internet).
為能簡要說明，本例的該服務對話模組41係分別與該第一用戶51與該第二用戶52完成信號連結(例如透過網際網路)，而且該服務對話模組41分別因應該第一用戶51所發出之一第一問題與該第二用戶52所發出之一第二問題來建立一第一對話串以及一第二對話串。至於信號連結至該服務對話模組41之該線上資料庫42，其中至少包含一第一用戶專屬資料庫421與一第二用戶專屬資料庫422，該服務對話模組41建立該第一對話串時僅參考該第一用戶專屬資料庫421中的內容，該服務對話模組41建立該第二對話串時僅參考該第二用戶專屬資料庫422中的內容。而該第一用戶專屬資料庫421的實例可以是為專屬該第一用戶的一第一會計及企業營運相關資料庫，其中資料內容可以包含銷售客戶地區以及產品庫存等等，該第二用戶專屬資料庫422的實例可以是為專屬該第二用戶的一第二會計及企業營運相關資料庫。而上述第一會計及企業營運相關資料庫與第二會計及企業營運相關資料庫皆可以由圖1A所示之會計資料處理方法來建置完成。如此一來，第一用戶與第二用戶便可以透過與服務對話模組41間的問答，快速地得到想要的財務會計資訊，例如應收帳款數目、現金與銀行存款數目、產品庫存明細表以及企業營運相關資訊(例如本月營業毛利是多少, 本月營業費用是多少錢, 本月銷售金額最高的顧客是誰, 本月哪個商品賣最好等資訊)。For the sake of brief description, the service dialogue module 41 of this example completes signal connection with the first user 51 and the second user 52 respectively (e.g., via the Internet), and the service dialogue module 41 establishes a first dialogue thread and a second dialogue thread respectively in response to a first question issued by the first user 51 and a second question issued by the second user 52. As for the online database 42 signal-linked to the service dialogue module 41, it at least includes a first user-specific database 421 and a second user-specific database 422. When the service dialogue module 41 establishes the first dialogue thread, it only refers to the content in the first user-specific database 421, and when the service dialogue module 41 establishes the second dialogue thread, it only refers to the content in the second user-specific database 422. The example of the first user-specific database 421 may be a first accounting and business operation-related database dedicated to the first user, wherein the data content may include sales customer regions and product inventory, etc. The example of the second user-specific database 422 may be a second accounting and business operation-related database dedicated to the second user. The first accounting and business operation-related database and the second accounting and business operation-related database may be constructed by the accounting data processing method shown in FIG1A. In this way, the first user and the second user can quickly obtain the desired financial accounting information, such as accounts receivable, cash and bank deposits, product inventory details, and business operation-related information (such as this month's operating gross profit, this month's operating expenses, who is the customer with the highest sales amount this month, which product is the best-selling this month, etc.) through questions and answers with the service dialogue module 41.
詳言之，線上資料庫42可由多個用戶專屬資料庫所組合而成，而每個用戶專屬資料庫皆對應至一特定用戶，也就是該特定用戶才能調用該用戶專屬資料庫中的資料，因此該服務對話模組41與該用戶完成連接時，便可先進行該用戶的身份認證，例如利用身份註冊時所建立的用戶編號與密碼，或是利用公司派發的內含有安全私鑰的實體安全權杖（Security token），來與該服務對話模組41進行公私鑰身份認證。而在完成該用戶的身份認證後，確認其合法用戶身份甚至是職務身份(例如是經理人或是一般職員等)後，才能進行後續的對話串建立過程，該服務對話模組41才能去調用該用戶專屬資料庫中的專屬資料來產生對話串內容。如此一來，該服務對話模組41建立該第一對話串時僅參考該第一用戶專屬資料庫421中的內容，該服務對話模組41建立該第二對話串時僅參考該第二用戶專屬資料庫422中的內容，用以避免機密資料不當外流。In detail, the online database 42 can be composed of a plurality of user-specific databases, and each user-specific database corresponds to a specific user, that is, only the specific user can call the data in the user-specific database. Therefore, when the service dialogue module 41 completes the connection with the user, the user's identity authentication can be performed first, for example, using the user number and password established during identity registration, or using a physical security token containing a security private key issued by the company, to perform public-private key identity authentication with the service dialogue module 41. After the user's identity authentication is completed and his legal user identity or even job identity (such as a manager or general staff member) is confirmed, the subsequent dialog string establishment process can be carried out, and the service dialog module 41 can call the exclusive data in the user's exclusive database to generate the dialog string content. In this way, when the service dialog module 41 establishes the first dialog string, it only refers to the content in the first user's exclusive database 421, and when the service dialog module 41 establishes the second dialog string, it only refers to the content in the second user's exclusive database 422, so as to avoid improper leakage of confidential information.
再舉實例來進行說明，如圖5之所示，上述服務對話模組41中可包含一語言模型411，可分別與圖4中的該第一用戶51與該第二用戶52完成信號連結，便可分別因應該第一用戶51所發出之第一問題與該第二用戶52所發出之第二問題來建立該第一對話串以及該第二對話串參考用戶專屬資料庫來分別與相對應的用戶產生對話串。而對話串內容的實例可以如下所列：用戶問題：本月現金水位是多少？系統回答：本月公司的現金水位是X元。用戶問題：本月銷售額實際值是多少？系統回答：本月銷售額實際值是Ｙ元。另外，用戶問題的例子還可以是：本月銷售額目標是多少？本月毛利跟上個月相差多少%?本月營業收入是多少？本月營業淨額是多少？本月營業成本是多少？本月營業毛利是多少？本月應付多少錢？本季應付多少錢本？月營業費用是多少錢？本月銷售最高的平台是甚麼？本月銷售最高的地區是哪裡？本月銷售額離目標還差多少？本月現金流入多少？本月哪個商品賣最好？本月銷售金額最高的顧客？是誰本月產品營業成本是多少？等等各類提問，但不限於上述問題，而本案系統則可以根據專屬資料庫內的內容，清楚回答上述用戶的提問，用以即時滿足用戶的需求。To illustrate with another example, as shown in FIG5 , the service dialogue module 41 may include a language model 411, which may respectively complete signal connection with the first user 51 and the second user 52 in FIG4 , so that the first dialogue string may be established in response to the first question raised by the first user 51 and the second question raised by the second user 52, and the second dialogue string may refer to the user-specific database to generate dialogue strings with the corresponding users. Examples of dialogue string content may be listed as follows: User question: What is the cash level this month? System answer: The company's cash level this month is X yuan. User question: What is the actual sales value this month? System answer: The actual sales value this month is Y yuan. In addition, examples of user questions may also be: What is the sales target this month? What is the difference in gross profit this month from last month? What is the operating income this month? What is the net operating amount this month? What is the operating cost this month? What is the operating gross profit this month? How much is due this month? How much is due this quarter? How much is the monthly operating expense? Which platform has the highest sales this month? Which region has the highest sales this month? How far is the sales amount from the target this month? How much cash inflow this month? Which product sells the best this month? Who is the customer with the highest sales amount this month? What is the operating cost of the product this month? And so on, but not limited to the above questions. The system in this case can clearly answer the above questions of users based on the content in the exclusive database to meet the needs of users immediately.
語言模型411可以使用大型語言模型(Large Language Model ，以下簡稱LLM)作為底層系統，用來進行訓練或串接。例如可以使用OpenAI 所推出的ChatGPT 3.5 turbo-16k，或是使用結構化查詢語言伺服器代理程式(SQL Server Agent)來進行訓練與串接。並在訓練內容中可以加入自行開發的優化LLM模型或其他技術模組包來進行調整，用以達成額外的使用需求，例如限定該用戶僅能問答屬於用戶自己的數據。而如圖5之所示，服務對話模組41中除了語言模型411之外，還包含一流程管理器412，例如是由 OpenAI 發布的LangChain，流程管理器412主要是提供了一種系統化的方法來將語言模型411工作流程中的不同流程進行分類，並可將其他語言模型也整合進來。流程管理器412的功能還可將語言模型與其他數據源(例如上述的複數個資料庫)連接起來，並允許語言模型與其環境進行互動。The language model 411 can use a large language model (LLM) as the underlying system for training or concatenation. For example, ChatGPT 3.5 turbo-16k launched by OpenAI can be used, or SQL Server Agent can be used for training and concatenation. In addition, the training content can be adjusted by adding a self-developed optimized LLM model or other technical module packages to meet additional usage requirements, such as limiting the user to only ask questions about the user's own data. As shown in FIG5 , in addition to the language model 411, the service dialogue module 41 also includes a process manager 412, such as LangChain released by OpenAI. The process manager 412 mainly provides a systematic method to classify different processes in the language model 411 workflow, and can also integrate other language models. The function of the process manager 412 can also connect the language model with other data sources (such as the above-mentioned multiple databases) and allow the language model to interact with its environment.
因此，本案可以對語言模型411進行訓練以及對流程管理器412進行設定，用以讓本案所完成之助理服務系統具備更多功能，訓練以及設定的內容如下：(1)為能限定該用戶僅能問答屬於自己公司的數據，在服務對話模組41中增加用戶身份辨識的功能。(2)同一公司內的用戶還可以依照其職務身份來決定用戶的權限，進而來限定對話串中問答內容的範圍，例如可以依照職務身份來限制langchain可觸及的資料庫種類或範圍。(3)優化使用者可問答題目深度：將不同進階財務指標先行設計成資料表(tables)或視觀表（Views），讓Langchain能進行資料擷取，用以降低語言模型411的任務複雜度。(4)為能優化Langchain對本案建立資料庫的熟悉程度，可以將本案資料庫中關於資料表(tables)或視觀表（Views）的各個欄位與資料內容概要輸入進Langchain中。(5)為能解決Langchain預設模組包與關聯式資料庫管理系統(PostgreSQL)架構不相容問題，可以將結構化查詢語言(SQL)語法利用Python程式語言進行調整，用以成功與資料庫進行互動，並可以解決Langchain預設模組包之其他原生限制，例如解決無法取用資料庫完整資料的問題，以及無法回答時可以避免胡亂回答的問題。承上所述，本案所發展出來之技術手段，可以讓中小企業的兼任人員在不需特別熟悉正式用詞的情況下，也可以完成記帳任務，透過本案所發展的技術手段，以及避免財務數據不正確的問題。另外，本案還提供自動除錯、折舊與產生現金流量表等功能，還可透過本案所發展的技術手段，可以提供不同公司用戶以遠端線上問答的方式直接獲取其相對應資料庫的重要資訊，但機密資料不會不當外流。雖然本發明以實施案例揭露如上，但並非用以限定本發明。本發明所屬技術領域中具有通常知識者，在不脫離本發明之技術精神和範圍內，當可作各種之更動與潤飾。因此，本發明之保護範圍當視後附之申請專利範圍請求項所界定者為準。Therefore, the present invention can train the language model 411 and configure the process manager 412 to enable the assistant service system completed in the present invention to have more functions. The contents of the training and configuration are as follows: (1) In order to limit the user to only ask questions about data belonging to his own company, a user identity recognition function is added to the service dialogue module 41. (2) Users in the same company can also determine the user's authority according to their job identity, thereby limiting the scope of the question and answer content in the dialogue thread. For example, the type or scope of the database that the langchain can access can be limited according to the job identity. (3) Optimize the depth of user-answerable questions: Different advanced financial indicators are designed as tables or views in advance so that Langchain can perform data extraction to reduce the task complexity of language model 411. (4) In order to optimize Langchain's familiarity with the database establishment of this case, the fields and data content summaries of the tables or views in the database of this case can be entered into Langchain. (5) In order to solve the incompatibility problem between the Langchain default module package and the relational database management system (PostgreSQL) architecture, the structured query language (SQL) syntax can be adjusted using the Python programming language to successfully interact with the database, and other native limitations of the Langchain default module package can be solved, such as solving the problem of not being able to access complete database data and avoiding random answers when unable to answer. As mentioned above, the technical means developed in this case can allow part-time employees of small and medium-sized enterprises to complete accounting tasks without being particularly familiar with formal terms, and avoid the problem of incorrect financial data through the technical means developed in this case. In addition, this case also provides functions such as automatic debugging, depreciation, and generation of cash flow statements. Through the technical means developed in this case, users of different companies can directly obtain important information from their corresponding databases in the form of remote online question and answer, but confidential information will not be leaked improperly. Although the present invention is disclosed as above by the implementation case, it is not used to limit the present invention. Those with common knowledge in the technical field to which the present invention belongs can make various changes and embellishments without departing from the technical spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be defined by the attached patent application scope claim.
100:交易資料檔案 10:用戶 13:會計資料庫 14:資料服務伺服器 15:資料處理引擎 4:助理服務系統 41:服務對話模組 42:線上資料庫 51:第一用戶 52:第二用戶 421:第一用戶專屬資料庫 422:第二用戶專屬資料庫 411:語言模型 412:流程管理器100: transaction data file10: user13: accounting database14: data service server15: data processing engine4: assistant service system41: service dialogue module42: online database51: first user52: second user421: first user exclusive database422: second user exclusive database411: language model412: process manager
圖1A ，其係本案所發展出來關於一種會計資料處理方法的步驟流程示意圖。FIG. 1A is a schematic diagram of the steps of an accounting data processing method developed in this case.
圖1B，其係本案所發展出來關於一種會計資料處理系統的方塊示意圖。FIG. 1B is a block diagram of an accounting data processing system developed in this case.
圖2，其係本案中關於該處理過資料組合的資料格式示例圖Figure 2 is an example of the data format of the processed data combination in this case.
圖3，其係本案提出可強化效能的提取轉換載入程式流程步驟示意圖。FIG. 3 is a schematic diagram of the extraction, conversion and loading process steps for enhancing performance proposed in this application.
圖4，其係本案提出的助理服務系統的功能方塊示意圖。FIG. 4 is a functional block diagram of the assistant service system proposed in this case.
圖5，其係本案助理服務系統中服務對話模組的內部功能方塊示意圖。FIG. 5 is a schematic diagram of the internal functional blocks of the service dialogue module in the assistant service system of this case.
4:助理服務系統4:Assistant service system
41:服務對話模組41: Service dialogue module
42:線上資料庫42: Online database
51:第一用戶51: First user
52:第二用戶52: Second user
421:第一用戶專屬資料庫421: The first user-specific database
422:第二用戶專屬資料庫422: Second user-specific database
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| TW113127509ATWI863884B (en) | 2023-12-21 | 2023-12-21 | Service assistant system |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| TW113127509ATWI863884B (en) | 2023-12-21 | 2023-12-21 | Service assistant system |
| Publication Number | Publication Date |
|---|---|
| TWI863884Btrue TWI863884B (en) | 2024-11-21 |
| TW202526792A TW202526792A (en) | 2025-07-01 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| TW113127509ATWI863884B (en) | 2023-12-21 | 2023-12-21 | Service assistant system |
| Country | Link |
|---|---|
| TW (1) | TWI863884B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110730953A (en)* | 2017-10-03 | 2020-01-24 | 谷歌有限责任公司 | Customizing interactive dialog applications based on creator-provided content |
| CN109661675B (en)* | 2016-07-08 | 2020-05-05 | 艾赛普公司 | Using semantic processing for customer support |
| TWM600888U (en)* | 2020-06-29 | 2020-09-01 | 臺灣新光商業銀行股份有限公司 | Intelligent business assistant system |
| CN114072832A (en)* | 2019-06-28 | 2022-02-18 | 脸谱科技有限责任公司 | Memory-based conversational reasoning and question-answering for assistant systems |
| CN115053244A (en)* | 2019-11-27 | 2022-09-13 | 亚马逊技术有限公司 | System and method for analyzing customer contact |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109661675B (en)* | 2016-07-08 | 2020-05-05 | 艾赛普公司 | Using semantic processing for customer support |
| CN111555954A (en)* | 2016-07-08 | 2020-08-18 | 艾赛普公司 | Automatically responding to a user's request |
| CN110730953A (en)* | 2017-10-03 | 2020-01-24 | 谷歌有限责任公司 | Customizing interactive dialog applications based on creator-provided content |
| CN114072832A (en)* | 2019-06-28 | 2022-02-18 | 脸谱科技有限责任公司 | Memory-based conversational reasoning and question-answering for assistant systems |
| CN115053244A (en)* | 2019-11-27 | 2022-09-13 | 亚马逊技术有限公司 | System and method for analyzing customer contact |
| TWM600888U (en)* | 2020-06-29 | 2020-09-01 | 臺灣新光商業銀行股份有限公司 | Intelligent business assistant system |
| Publication number | Publication date |
|---|---|
| TW202526792A (en) | 2025-07-01 |
| Publication | Publication Date | Title |
|---|---|---|
| AU2023200333A1 (en) | Systems and methods for identifying and explaining schema errors in the computerized preparation of a payroll tax form | |
| US8689096B2 (en) | Computer-based method, software module and computer program product for processing information in transaction-tax related applications | |
| CN106504079A (en) | A kind of composite type financial management method and its management platform | |
| US20050043998A1 (en) | Computer- based transaction tax processing system service module, method and computer program product for providing transaction tax services | |
| CN108734563B (en) | Method for automatically generating intelligent accounting document | |
| AU2016380691A1 (en) | Systems and methods for identifying and explaining errors in the preparation of a payroll tax form using error graphs | |
| US20030105687A1 (en) | Methods, data record, software interface, data warehouse module and software application for exchanging transaction- tax-related data | |
| US8620777B2 (en) | Methods, software modules and software application for logging transaction-tax-related transactions | |
| CN107924538A (en) | financial information reporting system | |
| US20050131805A1 (en) | Software interface, method and computer program product product for linking a business application to a component of a computer-based transaction tax processing system | |
| CN110427254A (en) | Task processing method, device, equipment and computer readable storage medium | |
| Rahmayanti et al. | Digital accounting for small to medium enterprises using mobile applications | |
| CN118644352A (en) | A financial and tax consulting service method and system based on blockchain | |
| CN118628263A (en) | Financial audit methods, devices, equipment, media and products | |
| Singireddy | AI-Enhanced Tax Preparation and Filing: Automating Complex Regulatory Compliance | |
| BE1026870B1 (en) | SYSTEM AND METHOD FOR AUTOMATIC VERIFICATION OF EXPENSE NOTE | |
| TWI863884B (en) | Service assistant system | |
| JP4373642B2 (en) | Supplier requirements system, supplier trend display control method, and program | |
| TWI883591B (en) | Accounting data processing method and system | |
| CN117974285A (en) | Accounting method, system, computer device and readable storage medium | |
| TW202526793A (en) | Service assistant system | |
| US12243082B1 (en) | Invoice processing platform | |
| Ma et al. | RPA Financial Robot Boosts the Digital and Intelligent Transformation of Enterprise Finance | |
| Tsaneva | Challenges of GDPR compliance in consumer financing companies | |
| JP4024267B2 (en) | Supplier guidelines system |