I'm building an application that searches for queries in OCR data. My documents are numerous and have many pages.
I'm using Wmd Similarity to query my data with gensim
instance = WmdSimilarity(text_data_preproc, model, num_best=10)
Given that I'm working with a large dataset of OCR data from documents, what should be fed into this function to get better results among the following options?
- The text each page only (one page at time)
- The text of the full document
- The text of all documents togheter
I'm using GloVe as my model.