def add_qrel_entry(qrel_dict, qid, did, grade):
qrel_key = (qid, did)
if qrel_key in qrel_dict:
prev_grade = qrel_dict[qrel_key].rel_grade
if prev_grade != grade:
raise Exception(
f'Repeating inconsistent QREL values for query {qid} and document {did}, got grades: ',
grade, prev_grade)
qrel_dict[qrel_key] = QrelEntry(query_id=qid, doc_id=did, rel_grade=grade)
continue
qid_orig, query, did, _ = fields
qid = ORCAS_QID_PREF + qid_orig
query_lemmas, query_unlemm = nlp.procText(query)
if query_lemmas == '':
continue
if query_lemmas in ignoreQueries:
print(f"Ignoring query, which is found in specified query files. Raw query: '{query}' lemmatized query '{query_lemmas}'")
query_toks = query_lemmas.split()
if len(query_toks) >= minQueryTokQty:
qrelList.append(QrelEntry(queryId=qid, docId=did, relGrade=1))
# Entries are sorted by the query ID
if prevQid != qid:
doc = {DOCID_FIELD: qid,
TEXT_FIELD_NAME: query_lemmas,
TEXT_UNLEMM_FIELD_NAME: query_unlemm,
TEXT_RAW_FIELD_NAME: query.lower()}
addRetokenizedField(doc, TEXT_RAW_FIELD_NAME, TEXT_BERT_TOKENIZED_NAME, bertTokenizer)
docStr = json.dumps(doc) + '\n'
outFileQueries.write(docStr)
prevQid = qid
if ln % REPORT_QTY == 0:
type=str,
help='input run file')
parser.add_argument('--out_qrels',
metavar='output QREL file',
required=True,
type=str,
help='output QREL file')
parser.add_argument('--top_k',
metavar='top k',
default=1,
type=int,
help='top k entries to use as psedo relevant labels')
parser.add_argument('--grade',
metavar='grade',
default=1,
type=int,
help='a grade for the relevance item')
args = parser.parse_args()
inp_run = read_run_dict(args.input_run)
qrels = []
for qid, run_dict in inp_run.items():
for did, score in get_sorted_scores_from_score_dict(
run_dict)[0:args.top_k]:
qrels.append(QrelEntry(query_id=qid, doc_id=did, rel_grade=args.grade))
write_qrels(qrels, args.out_qrels)
required=True,
type=str,
help='input run file')
parser.add_argument('--out_qrels',
metavar='output QREL file',
required=True,
type=str,
help='output QREL file')
parser.add_argument('--top_k',
metavar='top k',
required=True,
type=int,
help='top k entries to use as psedo relevant labels')
parser.add_argument('--grade',
metavar='grade',
default=1,
type=int,
help='a grade for the relevance item')
args = parser.parse_args()
inp_run = readRunDict(args.input_run)
qrels = []
for qid, run_dict in inp_run.items():
for did, score in getSorteScoresFromScoreDict(run_dict)[0:args.top_k]:
qrels.append(QrelEntry(queryId=qid, docId=did, relGrade=args.grade))
writeQrels(qrels, args.out_qrels)
qid_orig, query_orig, did, _ = fields
qid = ORCAS_QID_PREF + qid_orig
query_lemmas, query_unlemm = nlp.proc_text(query_orig)
if query_lemmas == '':
continue
if query_lemmas in ignore_queries:
print(
f"Ignoring query, which is found in specified query files. Raw query: '{query_orig}' lemmatized query '{query_lemmas}'"
)
query_toks = query_lemmas.split()
if len(query_toks) >= min_query_tok_qty:
qrel_list.append(QrelEntry(query_id=qid, doc_id=did, rel_grade=1))
# Entries are sorted by the query ID
if prev_qid != qid:
doc = {
DOCID_FIELD: qid,
TEXT_FIELD_NAME: query_lemmas,
TEXT_UNLEMM_FIELD_NAME: query_unlemm,
TEXT_RAW_FIELD_NAME: query_orig
}
add_retokenized_field(doc, TEXT_RAW_FIELD_NAME,
TEXT_BERT_TOKENIZED_NAME, bert_tokenizer)
doc_str = json.dumps(doc) + '\n'
out_file_queries.write(doc_str)
gen_query_qty += 1