Vectors are our data but in a language that AI understands. Data can be of a variety of types, such as text, image, and PDF. State-of-the-art vectors are generated by artificial neural networks that are trained on a huge amount of data and tested on different tasks.
Designing such networks and data preparation have proven to be a complicated task. Furthermore, training is quite an expensive process. Relevance AI provides you with a no-code, easily accessible and ready-to-run workflow to vectorize your text and image data.
What are the benefits of vectorizing data?
Neural vectors embody semantics and concepts and enable better similarity comparisons of data. Such quality allows a large category of use cases including search, recommendations, predictions, etc.
Vectorizing is to use the state of the art data representation relying on the latest development in AI and machine learning. Learn more about vectors at What are vectors.
Relevance AI's platform provides you with a no-code workflow to vectorize your "text" and "image" data with a few clicks.
Once you have uploaded your data, select your dataset, click on Vectorize under Workflows and follow the instruction.
Clicking on "Get started" and "Continue" will activate each relevant section.
Keep in mind that you need to specify the data type.
At the step, where you are selecting a model, a list of available vectorizers is shown with a brief description of each model.
Note 1: In general, MPNet and CLIP are good models for processing text and image data types respectively.
Note2 : To vectorize images, your dataset must contain the URLs to the images. Select the field (e.g. image-url), specify the type as image and select your desired image model (e.g. CLIP).
Running the vectorizer will add the corresponding vector field ("descriptionmpnet_vector" in the example on this page) to your dataset that can be seen under the Dataset -> Monitor -> Vectors.
We will learn about the clustering workflow on the next page.
Updated about 2 months ago