Searching with Odin

searching for technologies searching for technologies searching the patent family landscape in getfocus works through 'vector search' before you continue with this article, we recommend reading about vector search first what is vector search? docid\ wnovei8wgmfert akhlmo on the platform, you have 4 search categories technology, organization, publication number and similar inventions technology search technology search technology search helps you find all patent families associated with your search term(s) use technology search when you are looking to create a landscape of patents associated with a certain topic this feature allows you to comprehensively identify and analyze all patent families related to your specific area of interest technology search when conducting a technology search, selecting the appropriate search algorithm is crucial to achieving the desired balance between precision and recall below is a guide to help you decide which algorithm to use based on your objectives agentic search this is the default search algorithm and is generally the most effective for generating a highly relevant patent landscape it prioritizes precision, making it well suited for identifying the most pertinent patents however, if the results appear too restrictive and important patents are being overlooked, you can try decreasing similarity filter to widen the search and include more relevant patents—though this may introduce some false positives if the inclusion of false positives becomes problematic, consider applying an llm filter to refine the results and maintain high accuracy within your dataset smart search this is the legacy search algorithm, designed to cast a broad net to maximize coverage and reduce the risk of missing important patents choose this option if you prefer the legacy approach or if the default "agentic search" algorithm does not meet your specific needs exact search use this search algorithm when you require complete control over the search parameters and query structure it is ideal for scenarios where precision targeting is essential and custom query inputs are necessary to meet specific search objectives agentic search agentic search (recommended, default) (recommended, default) agentic search deploys advanced ai assistants powered by large language models (llms), to transform the way patents are explored and analyzed the advanced ai assistants enable a more efficient and intelligent approach to navigating the complex patent landscape agentic search advantages of agentic search advantages of agentic search produces highly relevant patent landscapes automatically optimizes the similarity filter for you supports natural language queries rather than using keywords (e g , “only all iron based electrolyte redox flow batteries – no vanadium, bromine, or any other chemistries”) does a lot of the heavy lifting for you delivers more stable dataset sizes (family counts) compared to "smart search" limitations of agentic search limitations of agentic search may be overly selective compared to "smart search", potentially excluding relevant patents less flexible when broader recall is needed; requires manual adjustment of the similarity filter to widen results slower performance compared to other search algorithms here’s an overview of how agentic search enhances patent discovery 1 exploring the patent landscape 1 exploring the patent landscape an llm agent expands your query by enriching it with important contextual information by expanding your query with important contextual information, the llm agent is able to explore a more complete patent landscape based on the underlying concept(s) rather than based on the presence of specific keywords 2 in depth analysis of patent landscape 2 in depth analysis of patent landscape using the expanded query from step 1, the llm agent performs an in depth analysis by exploring a carefully selected list of patents in detail the process works as follows select candidate patents the llm agent carefully selects candidate patents by moving from most relevant to least relevant in depth analysis of candidate patents for each of the selected candidate patents, the llm agent classifies them to be relevant or irrelevant for your search query based on the title , abstract , and claims 3 learning from in depth analysis of patents 3 learning from in depth analysis of patents using the patents analyzed in detail selected by the llm agent, we learn to reformulate an optimal vector search query the process works as follows optimize query vector using the in depth analysis of the patents we learn to improve the query vector such that it is centered based on all the relevant patents found by the agent optimize similarity using the in depth analysis of the patents we iteratively adjust the similarity filter to find the optimal similarity threshold for your query all of these steps happen automatically as part of one smooth workflow while the search is happening, you can track how many patents are classified as highly relevant , which determines the final dataset returned however, the number of relevant patents may vary depending on the specific case and the nature of the query behind the scenes, the llm checks patents across a wide range of similarity levels by dividing them into small groups it looks at a sample from each group to help decide where the results stop being useful this ensures you get a well balanced and accurate set of relevant patents agentic search finding patents smart search smart search when using smart search to explore a technology domain, the tool leverages generative ai to transform simple keyword inputs into a context rich, four sentence technical description these enriched queries are better suited for vector based search, often yielding more relevant and comprehensive patent families smart search is ideal for quickly exploring a domain when you only have keywords to start with however, if you’ve already crafted a detailed description of the technology, exact search is the better choice example entering the keyword “post combustion carbon capture” prompts the ai to generate a detailed, context enhanced query, as shown below smart search smart search generating enriched query e e xact search exact search preserves your original search terms without any enrichment or modification it uses your precise input to generate results that closely align with your specific wording and intent this approach is especially valuable when you’ve already crafted a clear, well defined description of the technology domain you’re exploring with exact search, you can trust that the results will be directly relevant to your area of focus, without any unintended expansion or narrowing of the search scope exact search organization search organization search organization search allows you to find all patent families associated with one or more organizations it is especially useful when you want to explore the complete patent portfolio of a specific company or compare multiple organizations simultaneously entering multiple organizations to search for multiple organizations at once, use one of the following separators new line semicolon (;) pipe (|) tab for example toshiba corp; mitsubishi; siemens note the following characters will not be treated as separators commas (,) are not treated as separators, as some organization names may contain commas spaces ( ) are also not treated as separators, since many organization names include spaces organization search publication number search publication number search publication number search allows you to find specific patents or patent families using their publication numbers use this feature when you need to retrieve a particular patent or search for multiple publication numbers simultaneously the standard format for publication numbers in getfocus is \[2 digit country code] \[patent number] \[kind code] example us 20180226168 a1 to search for multiple publication numbers at once, separate them using any of the following new line semicolon (;) pipe (|) tab publication number search similar inventions search similar inventions search similar invention search allows you to find patents that are closely related to a known invention by using its publication number this feature is ideal for discovering similar innovations and exploring adjacent technologies it is particularly useful for uncovering additional patents related to the same innovation mapping out the broader technology landscape around a known patent example enter the publication number (us 5603908 a) of a patent related to post combustion carbon capture to retrieve similar patents in that technology area based on the details in the patent similar inventions search recent searches & search history recent searches & search history at the bottom of the screen, you'll find the “recent searches” section, which displays your three most recent searches this allows you to quickly revisit previous searches with a single click if a search has been saved, it will be marked with a “saved” badge for easy identification recent searches next to the “recent searches” section, you’ll find the “show search history” button this provides a complete overview of your search history, with the ability to filter through them using keywords to delete a specific search, click delete in the context menu of the relevant entry note deleting a “saved” search will also remove the associated dataset from its respective project delete search history how to make best use of the search? how to make best use of the search? use the right search type for the right purpose technology search technology search use technology search when you need to analyze the patent landscape related to a specific technology or topic this search type is particularly useful when exploring the latest developments, key application areas in a field or other technical information—for example, understanding the innovation landscape in post combustion carbon capture in this case, you need to first select the technology tab, select the appropriate algorithm type and search for "post combustion carbon capture" or "metal organic frameworks (mofs)" to get the patent landscape for these technologies organization search organization search use organization search when you need to analyze the patent portfolio of one or more organizations this search type is especially helpful for identifying recent inventions and uncovering key technological strategic initiatives—for example, exploring toshiba corp’s latest patents or analyzing the past five years of innovation strategy for toshiba corp and mitsubishi in this case, you need to first select the organization tab and search for "toshiba corp" to landscape patents from toshiba or "toshiba corp; mitsubishi" to get the patent landscape of multiple companies publication number search publication number search use publication number search when you already have one or more publication numbers to analyze this search type is ideal for examining specific patents in detail—for example, understanding what is claimed in kr 101189075 b1 or identifying the application areas of us 20190291042 a1 in this case, you need to first select the publication number tab and search for "kr 101189075 b1" to get the specific patent or "kr 101189075 b1; us 20190291042 a1" to get the multiple patents based on the publication number provided similar inventions search similar inventions search use similar invention search when you want to investigate patents that are semantically similar to a known publication number this search type is particularly useful for identifying related innovations—for example, discovering which patents are similar to kr 101189075 b1 in this case, you need to first select the similar invention tab and search for "kr 101189075 b1" to get the relevant patents that are similar to provided publication number search results search results after completing any type of search, all patents can be found in the families tab instructions on how to use the overview tab can be found in charts docid\ p fayhi5vffd061iilsyz on the right side of the bar at the top of the page, you can find the “new this month" and “sort results” buttons new this month new this month to see patent families which have had new activity in the last month, use the button “new this month” it will show both entirely new inventions that were published in the past month, as well as patent families with newly published members in the past month click it to apply to your search landscape, click it again to revert the selection, and go back to the original state sort results sort results the patent families in the list can be sorted by relevance or publication date by default, your search landscape is sorted from the patent families that are most similar to your search query, to the least similar there are other sorting options available if you click 'sort results', you have 4 options relevance descending (default), relevance ascending , published date ascending , and published date descending relevance ascending will sort your results from least to most similar to your search query this sorting option can be useful to determine from which similarity % patent families start becoming irrelevant (e g which similarity% would be a good cut off point for your dataset to still be relevant) published date ascending will sort from the oldest published patent families to the youngest published date descenging will sort from the latest published date to the oldest simply click your preferred option and your search results will be sorted in the preferred way applying filters applying filters after creating a search landscape, you have various options to filter your dataset vector search gets you in the right direction, but depending on what exactly you are looking for, it might be necessary to filter your landscape further on the right hand side, you will find multiple filters read how to use them in how to filter search results? docid\ wi3kku8jft0fpwd7xnrnz please note that you must always click 'filter' after applying filters to make sure your search results are updatet the 'clear filters' button will clear all currently applied filters the 'filter' button will apply all currently selected filters navigation navigation you can click 'next' to navigate to the next page of results click 'previous' to navigate to the previous page of results the '<<' button will transport you back to the first page of results, and the '>>' button will transport you to the last page of results you can also enter a custom page in the free form number field to navigate to a custom page of your choosing simply enter the number and hit enter saving the results saving the results once you are satisfied with your search results (possibly after applying filters), you can save them to save your search results, click the save button located in the upper right corner above the filters section click 'save' to open the saving interface in the saving interface, you can either select a folder you have previously created or create a new folder with the '+ new folder' button if you create a new folder, you must give it a name and click the tick mark to create the folder once the folder is created, you can click it to store your dataset in this folder once a dataset is saved into a folder, it will appear in the projects tab with all your other saved folders and datasets as soon as a dataset is saved, new relevant patent publications will automatically accumulate in your dataset as long as they fall within the scope of your applied filters learn about the features for managing saved projects managing projects docid\ pd8s05tsffmildaxgc6n3 e e xporting search results xporting search results search results can be exported with the "export" button in the top right corner of the screen when you click export, a csv file will be created with your search results which you can then download to your computer while the export is being generated, do not leave the page if you navigate to another page while the export is being generated, the export will be stopped