#!/usr/bin/env python

# ner-carrels.py - given a number of configurations, output named-entities (human values)

# Eric Lease Morgan <emorgan@nd.edu>
# (c) University of Notre Dame; distributed under a GNU Public License

# February 22, 2024 - first cut; a long time coming


# configure
MODEL   = './model/model-best'
LIBRARY = '/shared/reader-patron-library/ericleasemorgan'
PATTERN = '*.txt'
SIZE    = 2207242
# require
from spacy   import load
from rdr     import TXT
from pathlib import Path
from sys     import argv, exit
from os      import stat

# get input
if len( argv ) != 2 : exit( "Usage: " + argv[ 0 ] + " <carrel>" )
carrel = argv[ 1 ]

# initialize
library        = Path( LIBRARY )
corpus         = library/carrel/TXT
nlp            = load( MODEL )
nlp.max_length = SIZE + 1

# process each file in the given corpus
for file in corpus.glob( PATTERN ) :


	# read and model the given file
	with open( file ) as handle : text = handle.read()
	doc = nlp( text )
	
	# output
	for entity in doc.ents : print( '\t'.join( [ file.stem, entity.text.lower() ] ) )

# done
exit()