Natural Language Search (NLS) is part of our AI Search Experience. It enables you to search for images in your portal using simple, everyday language. With NLS, you can quickly find image assets—even those without tags—while benefiting from support for over 100 languages. This makes locating the right asset faster and easier.
For example, instead of searching for tags like “person” or “headphones,” you can type a natural query like “a person listening to music.” NLS then retrieves relevant assets based on the semantic meaning of the query and the image content itself.
Unlike conventional searches that depend solely on metadata, NLS uses AI to interpret search keywords and the visual content of assets, providing intuitive and precise results.
By leveraging advanced AI technologies such as semantic search and vector embeddings, NLS bridges the gap between how users describe and retrieve assets.
How to Enable AI Search Experience
Reach out to your Customer Success Contact to learn how to enable the AI Search Experience in your portal and any associated costs.
How to Search using Natural Language Search (NLS)
- Navigate to your Portal > Assets tab.
- NLS only shows in the search bar when in the Assets tab.
- Click the search bar.
- A new banner will appear.
- Click the
to turn natural language search on. Once activated, the toggle will remain on, even if you navigate to other pages or close the window, until you manually turn it off.
- Type in the phrase to describe the image you are looking for.
- See here for best practices and examples.
- The asset overview will display results matching your phrase.
- (Optional) You can combine search methods and use NLS alongside filters or keyword-based searches to further refine your results.
Best Practices for Using Natural Language Search
Use Natural Descriptions
Describe what you’re looking for in simple, everyday language as if explaining it to someone else. This approach aligns with how semantic search processes input to understand intent.
-
- A woman sitting in a car
- A person wearing headphones
- A person wearing a blue shirt
Be Precise and Clear
Frame your search query with precise, descriptive details, avoiding ambiguity or overly complex terms. Think of how you’d explain it to a child for simplicity.
-
- Instead of saying, “An individual in an abstract setting,” use “A person standing in a garden.”
Avoid Statistically Rare or Ambiguous Terms
Certain uncommon words, slang, or vague phrases might not be recognized by the system effectively. Replace these with commonly used terms.
-
- Ambiguous: “Sunrise event in autumn”
- Clear: “Sunrise over a forest.”
Avoid Time and Location Expressions
Natural language systems may only be able to interpret references to specific times, dates, or locations if the metadata includes such details.
-
- Instead of: “A picture taken in New York in 2020.”
- Use “A cityscape with skyscrapers” and refine it using filters for location and date.
Leverage Metadata for Precision
If natural language search doesn’t return the desired results, refine your search by using filters date, location, file type, or specific keywords
-
- .Search for "mountain landscape" and filter by "file type: photo" and "color: blue."
Combine Approaches
Switch between semantic and traditional keyword or metadata searches to maximize results. For instance, start with a natural description and refine the search using more structured methods.
-
- “A group of people on a beach”
- Refine with a metaproperty filter, i.e., location, campaign, or date.
Supported Languages
Natural Language Search supports over 100 languages, with the following as the most common. If the language you're interested in is not included in the list, please contact your Customer Success Contact to verify its support status.
FAQ
Can NLS find images embedded in PDFs or other file types?
No, NLS works exclusively with image assets.
Can I save my NLS search results?
No, you cannot save your search.
What languages are supported in NLS?
NLS supports over 100 + languages, including the languages listed here.
How many search results does it produce?
NLS can provide a max—of 48 results, which the AI model ranks.
Can I apply a filter first and then search using NLS?
Yes, you can, and both the search and the filter will be applied to your results.
Does it support conversational language?
No, it currently does not support conversational languages.
Does it work with synonyms?
Yes, it does.
How does NLS search?
NLS uses high-quality image embeddings and training data to understand image content and search queries semantically. It does not use metadata and may struggle with highly specific searches where metadata excels. Review what works best for your specific query.
Can I apply a ‘search by image’ or a ‘similarity search’ and then search using NLS?
You cannot combine these; the latest search action will override the former.