Hi!
Let's see... we have a situation here.
The main problem is that the legends and designs are not structured in RDF, they are free text and formatted and in some cases with HTML tags and other texts tags (they are not clean). Therefore, the match between the original text and the text inside Dédalo is set to search with approximate text, it cannot be exact.
But as you indicated, in some situations the match could be totally wrong.
In your example, the RDF says: Altar.
And the search returns the first definition that says Altar, that is, the result of:
Unidentified female deity, towering, draped, standing in front, sacrificing from a patera in her right hand on the Altar, and carrying a box of perfumes in her left hand; on the right, a bareheaded child standing on a globe, raising both hands.
Therefore, it fits and assigns it.
Is possible to change it to perform the exact search, BUT, if we change it to be exact, many of the legends will NOT be found/match, as there are spaces and other characters that prevent them from being found when searched. In other words, the legends/designs with invisible HTML such as <p> or other tags will not match and the records will be duplicated. And yes, the import now creates many duplicates due to the above, but if we change it to exact, the duplicates will grow exponentially.
What is the least bad option?
1.- Let as is. Now it is set to similar, so that the legends/designs match as closely as possible.
2.- Change to be exact, the legends/designs will be more accurate but with many more duplicates (even though they are visually the same).
In other words, do we change the search criteria?
Best
