Yes, there is a demo site available where you can see the features of semantic search. Here are some examples:
- Similar Image Search: You can search for images similar to photos or illustrations stored on devices like smartphones. Semantic search represents the meaning of images in vectors, enabling image-to-image search.
- Text-based Image Search: You can search for images using naturally expressed text, such as "a cat on a car" or "people playing musical instruments". The images being searched do not have any text labels. This type of search is possible because both the query text and the target images are processed into semantic vectors.
- Idiom Search: Traditional dictionary searches only allow searching by matching keywords in headings or descriptions, which does not satisfy the need to find idioms or phrases like four-character idioms that convey a desired expression. However, with semantic search, you can vectorize the meaning of descriptions, allowing you to search using natural text. For example, searching "learn from past experiences" instantly yields the idiom "温故知新" (meaning to learn new things by studying the past). The demo also includes a checkbox for traditional keyword searches, so you can compare the differences.
The demo site offers various other demonstrations of semantic search as well.
