Whats the effect of both long encoding and chunked encoding of same document with ColbertV2.0??

#10
by DracoDev - opened

If we have document containing multiple sections what would be the effect of encoding the document as a while (assuming its under fits) and also encoding logical chunks of the code based on sentence splitting of subsections of the document? Would this increase or decrease accuracy when encoding with colbertv2.0? I imagine it would affect the return ranking of the main document depending on the encoded query. Would this practice tend to increase of decrease the accuracy of searches?

DracoDev changed discussion title from Whats the effect of both long encoding and chunked encoded of same document in bge-m3? to Whats the effect of both long encoding and chunked encoding of same document in bge-m3?
DracoDev changed discussion title from Whats the effect of both long encoding and chunked encoding of same document in bge-m3? to Whats the effect of both long encoding and chunked encoding of same document with bge-m3?
DracoDev changed discussion title from Whats the effect of both long encoding and chunked encoding of same document with bge-m3? to Whats the effect of both long encoding and chunked encoding of same document with ColbertV2.0??

Sign up or log in to comment