Skip to content
GitLab
Commit 892a227c authored by Jerome Wuerf's avatar Jerome Wuerf
Browse files

Adopt program to tira requirements

parent 5f0c2cbc
Hide whitespace changes
Inline Side-by-side
python/src/app.py
View file @ 892a227c
from pathlib import Path
from assessment import Assessment
from deduplication import Deduplication
from indexing import Indexing
from retrieval import (Retrieval, MaximalMarginalRelevanceReranking, StructuralDistanceReranking,
ArgumentGraphReranking, WordMoverDistanceReranking, NoReranking)
......@@ -29,10 +27,6 @@ class App:
self._indexing()
elif self.command == SubCommands.RETRIEVAL:
self._retrieval()
elif self.command == SubCommands.ASSESSMENT:
self._assessment()
elif self.command == SubCommands.DEDUPLICATION:
self._deduplication()
def _indexing(self) -> None:
self.logger.info('Reading sentences and embeddings form disk.')
......@@ -44,7 +38,10 @@ class App:
).index_to_es()
def _retrieval(self) -> None:
topics = read_topics(self.config['TOPICS_PATH'], self.config['TOPIC_NRB'])
topics = read_topics(
Path(self.config['INPUT_PATH'],
'topics.xml'),
self.config['TOPIC_NRB'])
self.logger.info('Read topics!')
retrieved_results = None
if self.config['REUSE_UNRANKED']:
......@@ -60,7 +57,6 @@ class App:
self.config['NRB_CONCLUSIONS_PER_TOPIC'],
self.config['NRB_PREMISES_PER_CONCLUSION']
).retrieve()
pickle.dump(retrieved_results, open('/data/baseline_dump.pkl', 'wb'))
# Pattern matching is only available pre Python 3.10 :(
reranker = None
......@@ -84,19 +80,9 @@ class App:
reranker.rerank()
write_terc_file(reranker.generate_trec_rows(),
Path(self.config['TREC_OUTPUT_PATH'],
f'{ self.config["RUN_NAME"]}.trec')
Path(self.config['OUTPUT_PATH'], 'run.txt')
)
def _assessment(self) -> None:
sentences = read_sentences(self.config['SENTENCES_PATH'])
results = read_results(self.config['RESULT_PATH'])
topics = read_topics(self.config['TOPICS_PATH'])
Assessment(results, sentences, topics).cmdloop()
def _deduplication(self):
Deduplication(self.config['ELASTIC_HOST'], self.config['INDICES']).deduplicate()
if __name__ == '__main__':
logger = logging.getLogger(__name__)
......
python/src/assessment/__init__.py deleted 100644 → 0
View file @ 5f0c2cbc
from .assessment import Assessment
python/src/assessment/assessment.py deleted 100644 → 0
View file @ 5f0c2cbc
from cmd import Cmd
from os import listdir
from pathlib import Path
from typing import Tuple
import pandas as pd
class Assessment(Cmd):
"""
TODO
"""
def __init__(self, results: pd.DataFrame, sentences: pd.DataFrame, topics: pd.DataFrame):
super().__init__()
sentences['id'] = sentences['arg_id'] + '_' \
+ sentences['sent_type'] + '_' + sentences['sent_pos']
sentences = sentences.drop(['arg_id', 'sent_type', 'sent_pos'], axis='columns')
self.sentences = sentences
self.results = results
self.topics = topics
self.idx = 0
self._set_current_topic()
def do_1(self, arg):
pass
def do_2(self, arg):
pass
def do_3(self, arg):
pass
def precmd(self, line):
self._print_sentence_pair()
self.idx += 1
return line
def postcmd(self, stop: bool, line: str) -> bool:
self._set_current_topic()
return stop
def preloop(self):
self._print_sentence_pair()
self.idx += 1
def _print_sentence_pair(self) -> None:
result_row = self.results.iloc[self.idx]
conclusion_sent_id, premise_sent_id = result_row['pair'].split(',')
conclusion_sent = self._get_sentence(conclusion_sent_id)
premise_sent = self._get_sentence(premise_sent_id)
print(f'\n\tConclusion: {conclusion_sent}\n\tPremise: {premise_sent}\n')
def _get_sentence(self, sent_id: str) -> str:
return self.sentences[self.sentences['id'] == sent_id].iloc[0]['sent_text']
def _set_current_topic(self) -> None:
self.prompt = f'(Topic: {self.topics.loc[self.results.iloc[self.idx]["qid"]]["topic"]})>'
python/src/deduplication/__init__.py deleted 100644 → 0
View file @ 5f0c2cbc
from .deduplication import Deduplication
python/src/deduplication/deduplication.py deleted 100644 → 0
View file @ 5f0c2cbc
import hashlib
import logging
from elasticsearch import Elasticsearch
from elasticsearch.helpers import streaming_bulk
from tqdm import tqdm
class Deduplication:
"""
TODO
https://www.elastic.co/de/blog/how-to-find-and-remove-duplicate-documents-in-elasticsearch
"""
TOTAL_DUPLICATES = 786383
def __init__(self, elastic_host: str = 'elastic', indices: list = ['conc', 'premise']):
self.logger = logging.getLogger(__name__)
self.es = Elasticsearch(elastic_host)
self.indices = indices
self.keys_to_hash = ['sentence_text']
def deduplicate(self) -> None:
"""
TODO
"""
self.logger.info('Starting deduplication...')
with tqdm(total=Deduplication.TOTAL_DUPLICATES) as progress_bar:
for ok, action in streaming_bulk(client=self.es,
actions=self.gen_delete_actions()):
progress_bar.update()
self.logger.info('Finished deduplication!')
def gen_delete_actions(self):
for index in self.indices:
for id_list in self.get_duplicates(index).values():
# Skip the fist element, delete all following with same hash
for id in id_list[1:]:
yield {"_op_type": "delete", "_index": index, "_id": id}
def get_duplicates(self, index: str) -> dict:
"""
TODO
"""
duplicates = {}
data = self.es.search(index=index, scroll='1m', body={"size": 10000, "query": {
"match_all": {}}, "fields": self.keys_to_hash, "_source": False})
sid = data['_scroll_id']
scroll_size = len(data['hits']['hits'])
while scroll_size > 0:
self._add_duplicates_of_batch(data['hits']['hits'], duplicates)
data = self.es.scroll(scroll_id=sid, scroll='2m')
sid = data['_scroll_id']
scroll_size = len(data['hits']['hits'])
return duplicates
def _add_duplicates_of_batch(self, hits, duplicates: dict) -> dict:
"""
TODO
"""
for item in hits:
key_values = ''.join([item['fields'][k][0] for k in self.keys_to_hash])
_id = item["_id"]
hashval = hashlib.md5(key_values.encode('utf-8')).digest()
duplicates.setdefault(hashval, []).append(_id)
return duplicates
python/src/utils/commands.py
View file @ 892a227c
......@@ -4,8 +4,6 @@ from enum import Enum
class SubCommands(Enum):
INDEXING = 'indexing'
RETRIEVAL = 'retrieval'
ASSESSMENT = 'assessment'
DEDUPLICATION = 'deduplication'
class RerankingOptions(Enum):
......
python/src/utils/configuration.py
View file @ 892a227c
......@@ -16,8 +16,8 @@ class Configuration():
'EMBEDDINGS_PATH',
'CREATE_INDEX',
'ELASTIC_HOST'],
SubCommands.RETRIEVAL: ['TOPICS_PATH',
'TREC_OUTPUT_PATH',
SubCommands.RETRIEVAL: ['INPUT_PATH',
'OUTPUT_PATH',
'RUN_NAME',
'TOPIC_NRB',
'NRB_CONCLUSIONS_PER_TOPIC',
......@@ -27,11 +27,6 @@ class Configuration():
'REUSE_UNRANKED',
'LAMBDA_CONCLUSIONS',
'LAMBDA_PREMISES'],
SubCommands.ASSESSMENT: ['TOPICS_PATH',
'SENTENCES_PATH',
'RESULT_PATH'],
SubCommands.DEDUPLICATION: ['ELASTIC_HOST',
'INDICES']
}
def __init__(self, args: Namespace):
......@@ -54,8 +49,8 @@ class Configuration():
args.elastic_host]
elif self.command == SubCommands.RETRIEVAL:
args_list = [Path(args.topics_path),
Path(args.trec_output_path),
args_list = [Path(args.input_path),
Path(args.output_path),
args.run_name,
args.topic_nrb,
args.nrb_conclusions_per_topic,
......@@ -67,14 +62,7 @@ class Configuration():
args.lambda_premises
]
elif self.command == SubCommands.ASSESSMENT:
args_list = [Path(args.topics_path),
Path(args.sentences_path),
args.result_path]
elif self.command == SubCommands.DEDUPLICATION:
args_list = [args.elastic_host, args.indices]
config = dict(zip(self.keys[self.command], args_list))
self.logger.info(config)
return config
python/src/utils/parse_cli_args.py
View file @ 892a227c
......@@ -11,27 +11,18 @@ class Text:
TODO
"""
description = '$$$$ Graph based Argument Mining on Sentence Embeddings $$$$'
sub = 'Lorem Ipsum dolor' # TODO
indexing = 'This sub command creates a semantic index with elastic search for an initial set' \
'retrieval.'
elastic_host = 'The hostname of the server that runs elastic search.'
create = 'If flag is present two new indices are created, one for conclusions, one for ' \
'premises. If there is already an existing index it will be overridden.'
indexing = 'Sub command to create a semantic index with elastic search.'
elastic_host = 'The hostname of the server/docker container that runs elastic search.'
create = 'If flag is present two new indices are created, overriding existing ones.'
sentences_path = 'The file path to the csv file containing the sentences.'
embeddings_path = 'The file path to the embeddings of the argument units.' \
'Overrides the EMBEDDINGS_PKL_PATH environment variable.'
retrieval = 'This sub command is intended for a run on the TIRA evaluation system.'
topic_nrb = 'Lorem ipsum' # TODO
embeddings_path = 'The file path to the embeddings of the argument units.'
retrieval = 'Retrive sentence pairs from the index.'
topic_nrb = 'Restrict the current indexing and/or reranking to a given topic number.'
nrb_conclusions = 'The number of conclusions that should be retrieved from the index per topic.'
nrb_premises = 'The number of premises that should be retrieved from the index per conclusion.'
run_name = 'The run name that will be included in the last column of the trec output file'
topics_path = 'The file path to the xml file containing the topics.' \
'Overrides the TOPICS_PATH environment variable.'
trec_output_path = 'The file path to the output file.' \
'Overrides the TREC_OUTPUT_PATH environment variable.'
assessment = 'Starts the interactive assesment of a result file.'
result_path = 'The path to a result file for an interactive assessment.'
deduplication = 'Deduplicates the indices given.'
run_name = 'The run name that will be included in the last column of the trec file.'
output_path = 'The file path to the directory containing the output files.'
input_path = 'The file path to the directory containing the input files.'
def parse_cli_args() -> argparse.Namespace:
......@@ -39,7 +30,7 @@ def parse_cli_args() -> argparse.Namespace:
TODO
"""
parser = argparse.ArgumentParser(description=Text.description)
subparsers = parser.add_subparsers(required=True, dest='subparser_name', help=Text.sub)
subparsers = parser.add_subparsers(required=True, dest='subparser_name')
# Indexing
......@@ -95,23 +86,7 @@ def parse_cli_args() -> argparse.Namespace:
required=False,
default=0.5)
parser_retrieval.add_argument('run_name', type=str, help=Text.run_name)
parser_retrieval.add_argument('topics_path', type=str, help=Text.topics_path)
parser_retrieval.add_argument('trec_output_path', type=str, help=Text.trec_output_path)
# Assessment
parser_assessment = subparsers.add_parser(SubCommands.ASSESSMENT.value, help=Text.assessment)
parser_assessment.add_argument('sentences_path',
type=str,
help=Text.sentences_path)
parser_assessment.add_argument('result_path', type=str, help=Text.result_path)
parser_assessment.add_argument('topics_path', type=str, help=Text.topics_path)
# Deduplication
parser_deduplicate = subparsers.add_parser(SubCommands.DEDUPLICATION.value,
help=Text.deduplication)
parser_deduplicate.add_argument('--elastic-host', help=Text.elastic_host)
parser_deduplicate.add_argument("--indices", nargs="+")
parser_retrieval.add_argument('input_path', type=str, help=Text.input_path)
parser_retrieval.add_argument('output_path', type=str, help=Text.output_path)
return parser.parse_args()
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment