Patent refusal notice and cited literature batch translation analysis support device

Abstract

A translation analysis support device is provided that reduces the workload of patent applicants or patent attorneys and supports quick and accurate examination responses. [Solution] The patent examination-related document translation and analysis support device translates, reverse-translates, and extracts difference information for patent rejection notices and cited documents written in a foreign language, enabling visualization of consistency and report output. Documents are input in PDF format or other formats, and are translated into the applicant's native language using the DeepL translation API, followed by reverse translation to extract difference information. The difference information results are visualized using background colors, enabling applicants or patent attorneys to respond quickly and accurately to multilingual notices and documents.

Description

[Technical Field]

[0001] This invention relates to a support device for the batch translation of notices of refusal and cited documents received by applicants or patent attorneys during patent examination, which translates, reverse translates, visualizes differences, and outputs documents of the review results in a single transaction, thereby reducing the labor and efficiency of examination work. [Background technology]

[0002] Applicants who file foreign applications often receive office actions and cited documents written in a foreign language. Dealing with these requires not only translation work, but also understanding the meaning of the text and extracting the technical key points.

[0003] However, traditional translation and analysis work is done manually and individually, which can lead to variations in translation accuracy and judgment, and requires a lot of time and cost. Furthermore, when multiple citations in multiple languages are presented at the same time, the work of comparing and examining them becomes even more cumbersome. Summary of the Invention [Problem to be solved by the invention]

[0004] The problem that this invention aims to solve is to provide a translation analysis support device that performs translation and reverse translation processing on multiple foreign language documents (notices of reasons for refusal and cited documents) in one go, extracts difference information, and visualizes the consistency, thereby reducing the workload of applicants or patent attorneys when trying to understand the original text through the translation, and supporting rapid and accurate examination responses. [Means for solving the problem]

[0005] In one embodiment of the present invention, the device comprises the following components: Reading section (1): Inputs the rejection notice or cited references from PDF or other electronic files (Word, text files, etc.) and extracts the original text data. Translation processing unit (2): The extracted original text data is translated into the native language (e.g., Japanese) using an external translation API (e.g., DeepLAPI, a translation API that can set a specialized dictionary function, etc.) to generate translated document data. Reverse translation processor (3): The translated document data is reverse-translated back into the source language to generate reverse-translated document data. Difference information extraction unit (4): Compares the original text data with the reverse-translated document data and extracts data on differences in vocabulary, syntax, and context. Visualization display unit (5): The difference data is formatted as text displayed against a background color and output to the screen of a display device. moreover, Report Generation Unit (6): Automatically generates a Japanese report based on the difference data generated by the Difference Information Extraction Unit and outputs it in Word or PDF format. By linking with a large-scale language model (LLM) (e.g., ChatGPT), it is possible to analyze the examiner's intent while preserving the logical structure and semantic features of the original data, and output AI review comments in Japanese. This invention is not intended for model training to improve translation accuracy, but is specialized in detecting discrepancies based on existing translations and supporting practical responses. [Effects of the Invention]

[0006] This device enables the translation and consistency check of multiple foreign language documents (offices of refusal and cited documents) submitted during patent examination. This reduces the dependency on individual tasks for translation and difference information review, significantly reducing the risk of misjudgments and the workload of applicants and patent attorneys. The output difference information visualization report also standardizes the organization of issues during examination and contributes to improving the quality of written opinions. Furthermore, this device does not include a glossary function, dictionary construction process, or comment generation function using AI. It can be practically configured to utilize existing translation services such as DeepL, enabling immediate implementation. The device is also useful for examiners at English-speaking patent offices when accessing patent examination-related data (offices of refusal, cited documents, opinions, and amendments) from non-English-speaking patent offices. [Brief explanation of the drawings]

[0007] [Figure 1] 1 is a diagram showing the overall configuration of a patent examination related document translation and analysis support device according to an embodiment of the present invention; DETAILED DESCRIPTION OF THE INVENTION

[0008] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. Figure 1 shows the overall configuration of the patent examination related document translation and analysis support system, with each processing block assigned a functional number for easy understanding. Each number is explained below. The reading unit (1) inputs at least one of the document data of the notice of reasons for refusal and the document data of the cited documents as original data in the form of a PDF or other electronic file (Word, text file), and extracts the original text data. The translation processing unit (2) translates the original text data extracted by the reading unit into Japanese using a translation API to obtain translated document data.

[0009] The reverse translation processing unit (3) reverse-translates the translated document data acquired by the translation processing unit back into the language of the original data to acquire reverse-translated document data. The difference information extraction unit (4) compares the original text data extracted by the reading unit with the reverse-translated document data obtained by the reverse-translation processing unit, and extracts difference data on vocabulary, syntax, and context. The visualization display unit (5) formats the difference data obtained by the difference information extraction unit as text displayed in a background color and outputs it to the screen of a display device.The background color data can then be added to the original data input as a PDF or other electronic file (Word, text file) and output as a file as a PDF or other electronic file (Word, text file).

[0010] moreover, The report generation unit (6) automatically generates a Japanese report based on the difference data and outputs it in Word or PDF format. Furthermore, by linking with a large-scale language model (LLM), it is possible to analyze the examiner's intention while preserving the logical structure and semantic features of the original data, and output AI review comments in Japanese.

[0011] Here, the patent examination related document translation and analysis support device The information processing device may further include a selection unit for selecting whether or not to perform generation by the report generation unit using the large-scale language model (LLM). The patent examination-related document translation and analysis support device includes: a confirmation instruction input means for receiving a confirmation instruction input from a user regarding whether or not unpublished data is included in the data of the notice of reasons for refusal or the data of the cited document to be input; If the confirmation instruction input means determines that the data contains undisclosed data, the reading unit may not accept input.

[0012] The translation API (DeepL API) used in this invention is based on a neural network model trained in advance using a large amount of parallel translation data, enabling natural translation based on the source text and context. The large-scale language model (LLM) is a model pre-trained based on a wide range of natural language data, enabling automatic generation of Japanese comments based on an understanding of context, logical structure, and linguistic semantic features. These models are utilized in a format that allows immediate response to user input data without additional training (fine-tuning).

[0013] The device of this invention accepts office action notices and cited documents written in a foreign language in electronic file format (PDF, Word, etc.), and segments the original text data into sentence-by-sentence units using OCR or text extraction processing.

[0014] Each segment is translated into Japanese by the translation processor, and then converted back into the source language by the reverse translation processor. By comparing the source text with the reverse translation results, differences in vocabulary, syntax, and context are extracted as difference information.

[0015] Users of the patent examination-related document translation analysis support device (e.g., patent attorneys and intellectual property officers) can intuitively understand whether mistranslations or ambiguities in translations are affecting the reasons for rejection by checking the displayed original text, translation, reverse-translation, and difference information.

[0016] The present invention does not include a function for automatically classifying the above difference information on the device side, and the output results are provided in monochrome. The applicants can refer to the difference contents and use them for review and evaluation as necessary. Based on the extracted difference information, the system has the function of automatically generating review comments in Japanese for each segment (e.g., "The reason for rejection is that there is no consistency between the applicant's claims and the constituent elements"), and supports the creation of opinions in language similar to that of local patent attorneys.

[0017] Large-scale language models (LLMs) can be used to generate review comments, and users can enable or disable this feature, allowing for flexible operation even in institutions with restrictions on the use of external AI. The difference information obtained by the difference information extraction process may be included as one item in the report output from the report generating unit (6). The report generated by the report generating unit (6) may be displayed in a list format and may be used by applicants and patent attorneys to consider how to handle the examination. Original Name Page Line Differences Notice of reasons for refusal A 10 5 Reasons for refusal → Reasons for refusal Cited Document B 5 17 Translation Analysis Support Device → Translation Aid Device Local attorney's comment C 14 3 "Opinion of local patent attorney" → "Judgment of local lawyer"

[0018] The patent examination-related document translation and analysis support device of this invention uses a translation API to translate rejection notices or cited documents written in a foreign language. However, rather than relying solely on the translation results, it uses an analysis process that combines back-translation, extraction of difference information, and analysis support for multilingual logic using generative AI to avoid misunderstandings of context and legal intent, enabling multifaceted examination responses that are in line with practicality.

[0019] The patent examination-related document translation analysis support device of this invention works in conjunction with a large-scale language model (LLM) equipped with contextual understanding capabilities for the source language (Japanese, English, German, French, Italian, Spanish, Portuguese, Korean, Chinese, Indonesian, Thai, Arabic, etc.), and can analyze and detect the examiner's intention while preserving the meaning, logical structure, and semantic features of the expression in the local language.

[0020] Therefore, for example, if the reason for refusal is in English, German, French, Italian, Spanish or Korean, it is possible to conduct an analysis in accordance with the logical thinking of each language, and if it is in Chinese, it is possible to conduct an analysis in accordance with the Chinese legal expressions, which is a departure from conventional methods that are based simply on reading the Japanese translation together.

[0021] Furthermore, this patent examination-related document translation and analysis support device visualizes the differences between the translated text and the original text through reverse translation processing, and utilizes a large-scale language model to automatically generate "AI review comments" in Japanese that include considerations regarding cited documents and reasons for rejection.

[0022] Furthermore, by inputting the findings and opinions submitted by local patent attorneys into this patent examination-related document translation and analysis support device, the "AI review comments" contain content that can be compared with the original, helping applicants or Japanese patent attorneys make quick decisions regarding the appropriate response. It should be noted that European and American patent offices (e.g., the United States Patent and Trademark Office (USPTO), the European Patent Office (EPO), etc.) are thought to use machine translation (e.g., Google Translate) internally to understand non-English documents such as Japanese, Chinese, and Korean when citing them. However, the details of the translation process, the name of the translation tool, and the original translation text are often not clearly stated in the notice of reasons for refusal. The present invention addresses this lack of transparency in the translation process by enabling applicants to autonomously verify the correspondence between the original text of cited documents and the translation results. This allows applicants to quickly realize the misunderstanding and refute any rejection based on an incorrect technical understanding caused by the translation, allowing them to refute the rejection through an appropriate written argument, etc.

[0023] (Patent examination-related document translation and analysis support device of this invention) This patent examination-related document translation and analysis support device is configured as a computer device equipped with input means, a display device, a storage device, and a processor, and each processing function (PDF reading, translation, back-translation, difference information comparison, visualization, report output, AI comment generation) is realized by a software program executed by the processor. The device is equipped with network communication capabilities, and can be configured to link with the DeepL API used for translation processing and the large-scale language model (LLM) used for review comment generation via an external server and network.

[0024] (An example of the configuration of a patent examination-related document translation and analysis support device) The patent-related document translation analysis support device includes at least the following hardware components: Processor (Central Processing Unit): Performs control calculations to sequentially execute the PDF reading process, translation process, reverse translation process, difference information extraction process, and AI comment generation process. Storage device (memory / storage): This device temporarily or permanently stores input document data, translation / reverse translation data during processing, processing results, generated reports, etc. The storage device is configured as a solid-state drive (SSD) or hard disk drive (HDD). Display device: A monitor that displays the processing results (translated text, difference information results, etc.) and report output content in a form that can be visually confirmed by the user, and has a graphical user interface (GUI). Input means: Includes a keyboard, mouse, or touch panel, and allows the user to select a file, start processing, select comment output, and other instructions. Network interface: Equipped with wired or wireless network communication capabilities for communicating with external translation APIs (e.g., DeepL) and large-scale language models (e.g., ChatGPT).

[0025] This device is equipped with hash value generation, electronic signature, and timestamp functions to verify that the generated translation results, reverse translation results, difference information, and AI review comments have not been altered by the user. This prevents tampering with translation and analysis results and ensures the authenticity and evidentiality of documents submitted by applicants or patent attorneys to examiners. In particular, if an examiner misinterprets a cited document, the output of this device can be used as supporting evidence for technical and linguistic arguments.

[0026] This device can also be equipped with a function to generate a hash value for the output content and add a digital signature and timestamp, making it possible to prove that the user has not manually altered the output translation results or reports. This makes it possible to technically prove to third parties that the translation and analysis results were generated automatically by this device and have not been edited or altered by the user. [Industrial Applicability]

[0027] This device targets patent examination-related documents (notices of reasons for refusal and cited documents) written in foreign languages and supports the efficiency and standardization of examination work for patent applicants and patent attorneys through translation, reverse translation, difference information extraction, visualization, and output of review comments using AI. This reduces the workload and improves the accuracy of judgments in patent practice both domestically and internationally, and has wide industrial applicability.

[0028] This device is an examination support device that reduces inaccuracies in examination decisions caused by translation languages and enables the technical arguments of applicants in non-English-speaking countries to be conveyed objectively and authentically even in examination practices in English-speaking countries, thereby contributing to improving the fairness of international patent examinations. [Explanation of symbols]

[0029] 1 Reading section 2 Translation Processing Section 3. Reverse translation processing section 4 Difference information extraction part 5 Visualization display section 6 Report Generation

Claims

1. An apparatus for supporting translation analysis of patent examination-related documents, comprising: a reading unit (1) that inputs at least one of document data of a notice of reasons for refusal and document data of cited documents as original data in the form of a PDF or other electronic file (Word, text file) and extracts the original text data; a translation processing unit (2) that translates the original text data extracted by the reading unit into Japanese using a translation API and acquires translated document data; a reverse translation processing unit (3) that reverse-translates the translated document data acquired by the translation processing unit into the language of the original data again to acquire reverse-translated document data; a difference information extraction unit (4) that compares the original text data extracted by the reading unit with the reverse-translated document data acquired by the reverse-translation processing unit, and extracts difference information on vocabulary, syntax, and context; a visualization display unit (5) that formats the difference information obtained by the difference information extraction unit as text displayed in a background color and outputs the text to a screen of a display device; A patent examination-related document translation and analysis support device comprising:

2. The reading unit (1) inputs the document data of the notice of reasons for refusal and the document data of the cited documents as original data in the form of PDF or other electronic files (Word, text files), extracts the original text data, a report generating unit (6) that generates a report in Japanese, including at least one of a list of the difference information, a difference information result, a review comment based on the difference information, an analysis result of the notice of reasons for refusal, an analysis result of the cited documents, and a result of creating a response proposal to avoid the reasons for refusal stated in the notice of reasons for refusal, and outputs the report in Word or PDF format; 2. The patent examination related document translation and analysis support device according to claim 1.

3. The report generation unit (6) By utilizing a large-scale language model (LLM), Japanese comments are generated that conform to the logic of the language in which the notice of reasons for refusal was written, enabling an analysis of the examiner's intention in writing the notice of reasons for refusal while preserving the logical structure and semantic features of the language in which the notice of reasons for refusal was written.

3. The patent examination related document translation and analysis support device according to claim 2.

4. a selection means for selecting whether or not to perform generation by the report generation unit using the large-scale language model (LLM), 4. The patent examination related document translation and analysis support device according to claim 3.

5. The patent examination-related document translation and analysis support device includes: a confirmation instruction input means for receiving a confirmation instruction input from a user regarding whether or not unpublished data is included in the data of the notice of reasons for refusal or the data of the cited document to be input; 2. The patent examination related document translation and analysis support device according to claim 1, wherein if the confirmation instruction input means determines that the document contains undisclosed data, the reading unit does not accept input.

6. The patent examination related document translation and analysis support device includes at least one of a hash value generation means, an electronic signature means, and a time stamp assignment means for verifying whether or not each output result, such as a translation result, a reverse translation result, difference information, and a report, has been altered.

6. The patent examination related document translation and analysis support device according to claim 1.

Images