FAQs
> Quick Start. How does Nuclia works?
> Start a subscription.
How does Nuclia works?
Nuclia out-of-the-box indexing process:
Nuclia extracts all the content from your files, that means, in short that:
- Nuclia extracts and indexes the full text of any files. If its a video or an audio it runs an automatic speech to text process. It also OCRs the images in videos.
- Besides the full text, Nuclia indexes every single paragraph from files.
- Once the paragraphs are generated, Nuclia, out-of-the-box, detects the key insights of every paragraph. This process is also called Named Entity Recognition. That means that Nuclia detects automatically, as an example, the name of people, dates, organisations, locations, etc..
- Once all the paragraphs are indexed, all text gets vectorized, providing a unique AI search on top of files. The vectorization is also the base of the generated answers you can perform with Nuclia.
Your first steps using Nuclia, a 3 minutes step-by-step onboarding
1. Upload your first files
How to upload files? Start using Nuclia by uploading some files. The fastes way to do so is:
- Go to “Resource view”
- Click the “Upload” button.
- Your files will start being indexed almost immediately.
- Wait until your files are processed
Discover other ways to upload data to Nuclia
- API (read the API documentation)
- Nuclia Desktop Application
NDA (Nuclia Desktop Application) it an application than can be downloaded in a laptop and provides some out-of-the-box connectors with popular data sources such Drive, SharePoint, Dropbox, One Drive or Youtube.
2. Files: understand how your files are indexed
In the resource view, click to any of the files that are already index.
You will be able see the full text extracted from the file and also visit the original file.
By clicking in the “annotation” option in the bar menu, you can check all the insights detected.
3. Search: understand the basics
In the section “search”, once your initial files are indexed, you can search.
Generative answer: ask Nuclia a question and you it will answer your question based on your own information. Make sure to activate the “Generative” search in the “Widget section”.
Search using natural language: ask Nuclia what do you want to find. If the information exists in any of the paragraphs from your data, it will be displayed in the results. If you click on a paragraph the original file will open and the result highlighted.
Search by keyword: if you want Nuclia to perform as a traditional search engine delivery results based on keywords, make a search query by adding “” in your query. Example: “London bridge”
4. Add Nuclia search to your application
To add Nuclia search to your application you have two options:
- Use the API (API documentation).
- Use the non code widget you can embed into any application. To add a “Nuclia search box” is as easy to configure some parameters and copy & paste the snipet of code generated. Note is very important to understand very good how the widget works. Check it here.
Resource view page
- What is the resource view pages?
- What are insights? -> Add your own insights
- Annotate insights
- What are labels?
- How to train your models
- What is training?
- Label search intent training
- Automatic resource label training
- Paragraphs label training
- Named entites recognition training
- What can you train?
- What is KB (Knowledge box)?
- What is a widget?
The widget is a very fast way to add a Nuclia AI search to any application. It does not require technical skills an its no-code. The widget generates a “search box” where you can perform questions and searches.
The widget can be enriched with some functionalities:
- Suggestions: suggest results while typing a search query
- Suggest labels: when label search intent training has been done, it displays lagesl suggested based on the query
- Filter: displays filters in the search bar
- Relations: Displays information about related insights (entities) in the search results
- Navigate to links: When using Nuclia to search web pages, clicking on a result will open the original page rather than rendering in the viewer.
- Permalinks: add extra parameters in URL allowing direct opening of a resource or search results.
- Generative: add generative answers when a search is performed.