def extract_eidos_text(docnames):
texts = {}
# Extract the evidence text for all events into a dict
for docname in docnames:
key = docname
texts[key] = []
print(docname)
fname = 'docs/%s.txt' % docname
with open(fname, 'r') as fh:
print('Reading %s' % fname)
txt = fh.read()
json_fname = 'eidos/%s.txt.jsonld' % docname
ep = eidos.process_json_ld_file(json_fname)
for stmt in ep.statements:
for ev in stmt.evidence:
txt = ev.text
txt = txt.replace('\n', ' ')
texts[key].append(txt)
# Now clean up all the texts to remove redundancies
for key, sentences in texts.items():
cleaned_sentences = copy.copy(sentences)
for s1, s2 in itertools.combinations(sentences, 2):
if s1 in s2 and s1 in cleaned_sentences:
cleaned_sentences.remove(s1)
elif s2 in s1 and s2 in cleaned_sentences:
cleaned_sentences.remove(s2)
texts[key] = cleaned_sentences
return texts
def text_to_stmts(text):
"""Run Eidos reading on a given text and return INDRA Statements."""
# We use some caching here so that sentences we have already read
# are not re-read.
fname = text.replace(' ', '_').replace(',', '_') + '.jsonld'
if os.path.exists(fname):
ep = eidos.process_json_ld_file(fname)
else:
ep = eidos.process_text(text)
shutil.move('eidos_output.json', fname)
return ep.statements
def read_eidos(docnames):
stmts = []
for docname in docnames:
fname = os.path.join('docs', '%s.txt' % docname)
jsonname = os.path.join('eidos', '%s.txt.jsonld' % docname)
if os.path.exists(jsonname):
ep = eidos.process_json_ld_file(jsonname)
else:
with open(fname, 'r') as fh:
print('Reading %s' % docname)
txt = fh.read()
ep = eidos.process_text(txt,
save_json=jsonname,
out_format='json_ld')
print('%d stmts from %s' % (len(ep.statements), docname))
# Set the PMID on these statements so that we can get the document ID
# during assembly
for stmt in ep.statements:
stmt.evidence[0].pmid = docname
stmts += ep.statements
return stmts
def test_process_json_ld_file():
ep = eidos.process_json_ld_file(test_jsonld)
assert len(ep.statements) == 1
assert 'UN' in ep.statements[0].subj.db_refs
assert 'UN' in ep.statements[0].obj.db_refs
