[Search] Relevance
The revelence methodology have to be improved to help users find the datasets they needs.
The test conducted revealed the following issues : When searching for a single term (ex: Tree) :
- Some datasets containing "tree" in the abstract are shown after datasets that contains "tree" only in his data.
- Some datasets are shown first despite the fact that they contain less occurences of the word "tree" in their data than datasets that are shown after.
When searching for a combinaison of term (ex : Tree Garibaldi) :
- Datasets showing results in their data are classified as "more relevant" than datasets showing results in their abstract ex : "Arbres remarquables -> Last page
How I think it should work : Most relevant datasets when searching a single element (ex: Tree) :
- The word is in within the title
- The word is within the abstract
- The word is within the data
Most relevant datasets when searching multiple terms (ex : Tree Garibaldi) :
- Both elements are in the title
- Both elements are either in the title, either in the abstract
- One of the element is in the title or the abstract, the other element is within data
- Only one element exist and he is within the title or abstract
- Both elements are within data
- Only one element exist and he is within the data