--- title: Patents analysis in GetFocus slug: patents-analysis-in-getfocus icon: 🗃️ docTags: createdAt: 2026-03-03T12:09:09.971Z --- ## Searching for patents Imagine you want to find patents about a specific chemistry or technology. For example, **ammonia-free compositions for hair dye**. To start building a patent landscape: 1. Go to the **Search** page. 2. Select **Technology** as the search type. *(Want to use a different search type? Read *[Search Types in GetFocus](docId\:A76ATaE2TwqaP3J9qM0w5)*)* 3. Enter your query in plain language. For example: **“ammonia-free hair dye composition.”** 4. Run the search. You’ll see a progress window while GetFocus runs the search, then you’ll land on the results page. ![Search page](https://api.archbee.com/api/optimize/ej5ZdwFecA40xVAaeh_s7/2hgW6TLlswMr7AMl2Ct6K-20260304-105649.png "Search by Technology") :::hint{type="info"} 💡 Build your search query using the formula: **\[****Technology****] + \[****Application****] + \[****Context****]** Example: **A********mmonia-free** **hair dye** **for gray coverage****** ::: GetFocus uses **vector-based (semantic) search**, which means you don’t need to write long Boolean strings to get relevant results. *(New to vector search? Read *[What is Vector Search?](docId\:wnOVeI8WGMFERT-aKhlmO)*)* One key difference from Boolean search: with Boolean, broader terms usually return more results. With natural-language (semantic) search, adding specific context helps the system understand your intent, so more detail often leads to *more* relevant results. Adding relevant context is crucial for making AI search perform well. The more clearly you describe what you’re looking for, the better GetFocus can retrieve the right documents. A full sentence works better than a list of keywords. **Good example:** **“*Ammonia-free hair dye using monoethanolamine + 4-amino-2-hydroxytoluene for 100% gray coverage*”. **Why? Well specified, with additional relevant context **What not to do:** ***“Hair dye” ***or ***“Ammonia-free”*****. **Why? Too broad and vague on its own. It doesn’t say *what product*, *which approach*, or *what performance goal* you care about. You’ll pull in a huge range of unrelated patents. *** Now that you’re on the results page, here are a few basics to know: ![](https://api.archbee.com/api/optimize/ej5ZdwFecA40xVAaeh_s7/RDuZGDpa4dl0R62ufh2Mc-20260304-152000.png) - **Search details:** Click the info icon (1) to view your search type and query. - **Patent family count: **The side panel shows how many patent families are in your dataset (2). - **Sorted by relevance: **Patents are ordered from most to least relevant, so your strongest matches appear at the top. You can also see each patent’s relevance score for your search (3). - **Broad first results (recall over precision):** Initial results are intentionally broad, so you may see some noise at first. - **Filter for accuracy:** To sharpen the dataset, apply filters - this is exactly what we’ll cover in the next step. Want to understand how each patent record is shown in GetFocus? See article** Patent card.** *** ## Filtering to find the right patents quickly After you run a search, you’ll get a patent dataset, but it will often include some **noise**. By “noise,” we mean patents that are **somewhat related** (they use similar terms or address a neighboring problem) but **aren’t truly in scope** for what you’re trying to analyze. That’s where **filters **come in. Your toolkit includes many different filters, and you can read more about each one in the article . ![](https://api.archbee.com/api/optimize/ej5ZdwFecA40xVAaeh_s7/T3Y7Us5OG6bjxdXvWOnJf-20260305-090556.png "Filters section") For this Quickstart guide, we’ll keep things simple and focus on the **LLM filter**. It lets you describe what you want in **natural language**, so you can quickly remove noise and keep only the patents that match your topic. 1. Click **Create a new LLM filter**. ![](https://api.archbee.com/api/optimize/ej5ZdwFecA40xVAaeh_s7/yDIt8kD8AsNdL6ZkinLY8-20260305-120327.png "Create LLM Filter button") 2. In the instruction box, describe what you want to keep in plain language. **Example instruction:** “Only include patents about 100% ammonia-free hair dye compositions for grey coverage that also have conditioning or nourishment benefits.” :::Paragraph{indent="1"} Need help writing a good instruction? See [How to write LLM filters that work.](https://docs.getfocus.eu/prompting-like-a-pro#undefined-llm-filter-that-workundefined) ::: 3. Click **Create LLM filter** (or press **Enter**) to start. GetFocus will review each patent family and decide whether it matches your instruction. Anything that doesn’t match will be filtered out. ![](https://api.archbee.com/api/optimize/ej5ZdwFecA40xVAaeh_s7/S1M4cskZ3eZ3Q3fMBmC_2-20260305-160149.png "How LLM Filter works") :::hint{type="info"} 💡**Runs in the background:** LLM filters process about **12 patent families per second** and can keep running while you continue working in GetFocus. When you’re ready, come back and click **Apply filter**. ::: 4. When processing is finished, click **Apply filter** to see the results. ![](https://api.archbee.com/api/optimize/ej5ZdwFecA40xVAaeh_s7/WxTvb8r9ZyGBgHeCkQJlg-20260306-143306.png "Apply LLM Filter") :::hint{type="warning"} ‼️One LLM filter can process up to **10,000 patent families**. If your dataset is larger, GetFocus will process the **10,000 highest-relevance families** first. ::: After you apply the LLM filter, the dataset shrinks because patents that don’t match your criteria are filtered out. ![](https://api.archbee.com/api/optimize/ej5ZdwFecA40xVAaeh_s7/EqS4ZMLLcJU6b2i9H3Ydo-20260306-143654.png) *** ## Extracting the insights Now, getting to the insights. When you’re ready to turn your dataset into answers, GetFocus offers two ways to chat with your patents: - **Chat with Set** → ask questions across a dataset (up to **1,000 families**) - **Chat with Invention** → deep-dive into **one** patent in detail Use **Chat with Set **for *“What’s happening across the landscape?” *and **Chat with Invention **for *"What does this patent specifically say?”* ![](https://api.archbee.com/api/optimize/ej5ZdwFecA40xVAaeh_s7/sREi_N66bT3urLe91bx7R-20260306-153546.png) ### Chat with Set Use **Chat with Set **to ask high-level questions across a group of patents—trends, themes, and portfolio comparisons. **What it reads:** titles, abstracts, and claims (for all families in the set) **Best for:** - spotting trends and how they change over time - comparing portfolios across companies - summarizing what’s in your filtered dataset **How it works:** Type your question, and your answer will be generated below. ![](https://api.archbee.com/api/optimize/ej5ZdwFecA40xVAaeh_s7/DM_3JgY-Iiw2JZn6UJS8d-20260311-132312.png "Chat with Set example") :::hint{type="info"} If your dataset contains **more than 1,000 families**, Chat with Set will use a **subset of 1,000** families to generate an answer. 💡 If you want full control over *which* families are included, filter your dataset down to **≤ 1,000** before asking your question. You can also apply common filters directly in Chat with Set (e.g., **publication year**, **status**, **patent office**, **ultimate owner**). ::: ### Chat with Invention Use **Chat with Invention** when you want detailed Q\&A on a specific patent. **What it reads:** title, abstract, claims, and the **full description** (including images, if available) **Best for:** - understanding a single invention thoroughly - extracting implementation details, embodiments, or claim scope - answering precise technical questions Chat with Invention includes references and cites sources at the bottom of its answer. ![](https://api.archbee.com/api/optimize/ej5ZdwFecA40xVAaeh_s7/0KK7wr5r3PPUQZS7w1GcD-20260306-163544.png "Chat with Invention example") :::hint{type="info"} Chat with Invention supports two modes: - **Non-reasoning mode:** best for quick, straightforward questions (fast responses) - **Reasoning mode:** best for complex questions (takes more time, higher quality reasoning) ::: *** ## What's next? You've built your dataset and have the tools to analyze it. To go advanced, explore the articles and interactive walkthroughs below: ::::LinkArray{contentSource="CUSTOM" itemsPerRow="2"} :::LinkArrayItem ⚖️ **Patentability/Prior Art Search** Identify novelty of an idea and find prior art most closely related to the concept [](https://getfocus.storylane.io/share/jiahhkfjenia) ::: :::LinkArrayItem{headerType="NONE"} ⚖️ **Invalidity search** Identify patents that destroy the novelty of a patent you want to invalidate [](https://getfocus.storylane.io/share/f7dm0kkv6utc) ::: ::::