16 results about "Patent search" patented technology

The present invention relates to the technical field of natural language processing, and specifically to a visual patent search method based on a pre-trained language model. In a patent keyword search scenario, the present invention extracts keywords and represents word vectors from patent texts based on the pre-trained language model, reduces the dimensionality of patent features returned by the search, and visually displays the search results in the form of a scatter plot. Compared with search results displayed in a list form, the method is more intuitive and provides richer search information such as the degree of similarity between patents and patent clustering.

To solve the problem in which: causing a language model to learn the feature of a patent journal to be examined requires a large number of similar patent journals including an individually focused patent journal, and thus requires knowledge and effort of a skilled patent technician each time.SOLUTION: A sentence encoder 1 having re-learned a language model 12 by using a large number of patent journals 70 for every section symbol of eight IPC classifications was prepared, claims in a focused patent journal 80 and a patent journal to be examined 90 were divided, for every claim, into segments (82, 92) in a number equal to or less than 1 / 4 of the number of tokens, and the similarity among the combinations of segments was calculated by using the sentence encoder 1 having the same section symbol described at the head of the patent journal to be examined 90. Subsequently, segments with high similarity are displayed and the claims can be checked. Note that the segments displayed in higher ranks are not included in display in lower ranks. When there are a plurality of patent journals to be examined 90, a score based on the similarity is calculated for every patent journal, and segments 92 in the patent journal to be examined 90 were displayed in the order of the scores.SELECTED DRAWING: Figure 16

The invention relates to a construction method of a patent-specific vector model and a patent semantic retrieval method, and the construction method comprises the steps: carrying out the similar patent retrieval class fine tuning of an open-source BGE-M3 vector model, so as to obtain a fine-tuned vector model; performing patent context type fine tuning on the fine-tuned vector model to obtain a patent special vector basic model; performing high-similarity patent retrieval class fine tuning on the patent-specific vector basic model to obtain a patent-specific vector model; and performing data format adaptation type fine tuning on the patent-specific vector model to obtain a final patent-specific vector model. According to the patent semantic retrieval method, similar patent retrieval class fine tuning is performed on the open source BGE-M3 vector model, patent context class fine tuning, high similar patent retrieval class fine tuning and data format adaptation class fine tuning are performed on the fine-tuned vector model to obtain the patent special vector model, and patent semantic retrieval is performed based on the patent special vector model. The technical problem that existing patent information utilization efficiency and depth are low is solved.

1. Name of this design product: Peer search bar for graphical user interface of patent search for electronic device. 2. Application of this design: This design is intended for use in an electronic device. 3. The key design feature of this product is its graphical user interface. 4. The picture or photo that best illustrates the key design points: Design 1 front view. 5. Due to the conventional design of electronic equipment, the rear view, left view, right view, top view, bottom view, and perspective view of Design 1-3 are omitted. 6. Design 1 is designated as the basic design. 7. Uses of the graphical user interface: Overall use: Patent search and interaction; Local use: Searching for peer companies of patent holders or applicants; In the main view of Design 1, check the checkbox before any company name, and click the "Filter" button in the left sidebar to go to the interface change state diagram 1 of Design 1; In the interface change state diagram 1 of Design 1, click the "More" button to go to the interface change state diagram 2 of Design 1; In the interface change state diagram 2 of Design 1, click any company name to jump to the interface change state diagram 3 of Design 1 and view the peer information of the corresponding company. 8. Other situations requiring explanation: This design is a partial design, and the part outside the red transparent area is the part for which protection is sought.

A text generation system that supports understanding of both an inventor and a patent engineer and supports creation of appropriate claims for the invention in patent application. The text generation system is configured to perform processing using a language model, create a key point of the invention from material of the invention, create a claim proposal from the key point of the invention, and create a search formula from the claim proposal. In addition, the text generation system is configured to perform a patent search with use of the search formula.

The application discloses a kind of appearance patent image retrieval method and system based on customizable semantics.It includes: from appearance patent database, text-patent pair is filtered and obtained, and the hand-drawn sketch corresponding to the appearance patent image of each patent is generated;ResNet method is used to carry out semantic coding to the appearance patent image, and construct semantic feature library;Appearance patent preliminary screening model is constructed and trained;Appearance patent final screening model is constructed and trained;User inputs the keyword text and hand-drawn sketch to be searched, and keyword text is used as input to carry out preliminary screening, and customizable hand-drawn sketch is used as input to carry out final screening, to obtain accurate appearance patent search result.The application is based on the mode of artificial intelligence, mixed use text and image to the mixed mapping of patent, simultaneously realize the efficiency and accuracy of appearance patent search, through customizable semantic hand-drawn sketch search, more effectively reduce the difficulty of image search, improve the ability of search details.

We provide technology that uses AI agents to support patent searches. [Solution] The program causes the computer to perform the following processes: acquiring technical information including technical ideas; acquiring the purpose of the patent search; inputting the technical information and the purpose of the search into a search design AI agent that has been instructed in advance to output search design information that serves as a guideline for the patent search; and acquiring search design information related to the technical information based on the results obtained from the search design AI agent by inputting the technical information and the purpose of the search into the search design AI agent.

The invention provides a patent retrieval total instruction optimization system and method, a program product and a storage medium, and relates to the field of electric digital data processing. In the same hierarchy, the same retrieval units in the same hierarchy structure of the patent retrieval total instruction are focused, and the merging operation is implemented on the basis of the existing logic operators among the associated retrieval units. In the aspect of cross-hierarchy, when the continuous parent retrieval units cover the same retrieval unit and the logic operators applied between the adjacent parent retrieval units are the same, the merging operation is carried out based on the same retrieval unit. The structure is simplified at the same hierarchy, redundant elements are removed, then the hierarchical architecture is further optimized in a cross-hierarchy mode, and the number of unnecessary retrieval units in a complex retrieval formula and a complicated nesting relation are reduced. And it is ensured that the whole patent retrieval process can operate more efficiently and smoothly when coping with complex instructions.

The system according to this embodiment aims to efficiently generate new ideas and verify their novelty. [Solution] The system according to the embodiment comprises a reception unit, an analysis unit, a search unit, a refinement unit, and a proposal unit. The reception unit receives the user's inquiry. The analysis unit analyzes the inquiry received by the reception unit. The search unit searches for existing patents using a patent search API based on the content analyzed by the analysis unit. The refinement unit refines the idea based on the search results obtained by the search unit. The proposal unit proposes the idea refined by the refinement unit to the user.

The application provides a patent search sentence design method and device, a medium and equipment. The method comprises the following steps: identifying a logical operator in a search formula; based on the logical operator, splitting the search formula into at least one search term group; wherein the search term group comprises at least one search term; identifying the category of the search term group; based on the category of the search term group, generating an extended search term of the search term group; and based on the category of the search term group, the search term group and the extended search term, generating a search sentence of the search formula. That is, the search formula is split into search term groups according to the logical operator, the extended search term is generated by identifying the category of the search term group, and a more accurate search sentence is generated in combination with the category of the search term group, the search term group and the extended search term, so as to provide a more expected patent file for the searcher, reduce the search difficulty and improve the search efficiency and effect.

1. Name of this design product: Peer search section of the graphical user interface for patent search of electronic devices (III). 2. Application of this design: This design is intended for use in an electronic device. 3. The key design feature of this product is its graphical user interface. 4. The picture or photo that best illustrates the key design points: Design 1 front view. 5. Due to the conventional design of electronic equipment, the rear view, left view, right view, top view, bottom view, and perspective view of Design 1 to Design 4 are omitted. 6. Design 1 is designated as the basic design. 7. Uses of the graphical user interface: Overall use: Patent search and interaction; Partial use: Searching for peer companies of patent holders or applicants; Clicking on any company name in the search results details in the main view of Design 1 will lead to the interface change state diagram 1 of Design 1, allowing you to search and view the patent information of the corresponding company; Clicking the "View Peers" button in the interface change state diagram 1 of Design 1 will lead to the interface change state diagram 2 of Design 1, allowing you to view the peer information of the corresponding company. 8. Other situations requiring explanation: This design is a partial design, and the part outside the red transparent area is the part for which protection is sought.

1. Name of this design product: Patent Visualization Recommendation Graphical User Interface for Electronic Devices (Zhi Chuang Tu Xin Patent Visualization Recommendation Platform). 2. Purpose of this design product: for use in an electronic device. 3. The key design point of this design product lies in the interface content of the graphical user interface. 4. The image or photograph that best illustrates the design's key features: the front view. 5. The hardware design of this design product is a conventional design, omitting the rear view, left view, right view, top view, bottom view, and stereoscopic view. 6. Purpose of the graphical user interface: This web interface is used for the operation and display of the Zhichuangtu new project website. Users can perform various operations such as patent search and knowledge graph query on the website platform. 7. Other situations that require explanation are explained in the following diagrams: The main view is the main interface for users to enter the program. In this interface, users can interact with the intelligent robot to ask and answer questions. The knowledge graph is matched according to the content of the questions and answers, and is displayed as interface change state diagram 1; click the upper right corner of the question and answer box to enter interface change state diagram 2; in interface change state diagram 2, users can interact with the knowledge graph; click the green entity node in the screen to enter interface change state diagram 3, in which the specific content of the entity can be viewed; click "Media Resources" on the right to enter interface change state diagram 4, in which media resources related to the patent can be viewed.

The application provides a patent retrieval method and device and a storage medium. The most suitable document set satisfying a retrieval parameter combination condition can be quickly found by browsing a label of the retrieval parameter combination, and then corresponding patent documents are retrieved, so that the retrieval time is greatly reduced and the retrieval efficiency is improved. The method comprises the following steps: providing a retrieval interface; determining a retrieval condition in the retrieval interface according to a condition determination operation instruction of a user; determining a retrieval parameter corresponding to the retrieval condition, wherein the retrieval parameter comprises at least one of a keyword, an IPC and a CPC; determining P retrieval parameter combinations corresponding to the retrieval parameter, wherein P is an integer greater than or equal to 1; and performing retrieval on a target data set based on the P retrieval parameter combinations to obtain a patent document set hit by each of the P retrieval parameter combinations, wherein the target data set corresponds to the retrieval condition.

The invention discloses a patent retrieval method based on semantic understanding, relates to the technical field of patent document retrieval, and solves the technical problem that a patent retrieval method in the prior art is difficult to capture deep semantic association behind a query intention, so that the accuracy and efficiency of a user in a patent retrieval process are not high. According to the method, the text content and the structure information in the image can be accurately extracted from the original text through the document analyzer and the graphic information detection model; and the structure information is optimized and sorted by adopting a sorting tree algorithm, so that high-quality integration of formatted patent texts is ensured. Key information is extracted based on large model analysis of historical query records, a corresponding embedded vector is generated in combination with a semantic model, and the embedded vector and a corresponding formatted text are stored in a vector database. When a user submits a query, the query vector is generated through the semantic model, and the matched result set is quickly retrieved in the vector database, so that the retrieval efficiency and accuracy are improved.

The invention discloses an artificial fish reef patent intelligent classification retrieval method and system based on multi-modal fusion and cross-modal Hash, and the method comprises the steps: obtaining artificial fish reef patent metadata, analyzing and splitting the artificial fish reef patent metadata into multi-modal data, inputting the multi-modal data into a multi-tower network structure, obtaining fusion features, and generating corresponding Hash codes; performing classification marking on the artificial fish reef patent metadata according to a preset patent data classification label system, and performing association storage on the artificial fish reef patent metadata and the corresponding hash codes; hash codes of retrieval conditions are obtained on the basis of a multi-tower network structure, initial retrieval results are obtained on the basis of the Hash codes of the retrieval conditions, the initial retrieval results are screened through multivariate similarity measurement, and intelligent classification retrieval of the intelligent artificial fish reef patent is completed. By constructing an artificial fish reef modeling hierarchical classification label system and integrating functional modules such as multi-modal feature extraction, data integration and efficient retrieval are realized, and patent retrieval efficiency and precision are improved.

A patent search support method according to the present invention is for supporting an infringement search to determine whether or not technology being searched infringes an existing patent, and involves a computer executing: a first accepting step (S5) for accepting search target text information written in the form of a patent claim relating to the technology being searched; a second accepting step (S5) for accepting comparison target text information, which comprises the claims of an existing patent; a determination step (S6 to S7) for determining the possibility that the technology being searched falls within the technical scope of the invention set forth in the existing patent claims by using a generative AI to compare the search target text information with the comparison target text information; and an output step (S9) for outputting the determination result obtained by the determination step.