123 lines
7.4 KiB
Python
123 lines
7.4 KiB
Python
|
import os
|
||
|
|
|
||
|
|
# import and build cython
|
||
|
|
import pyximport
|
||
|
|
pyximport.install()
|
||
|
|
|
||
|
|
from main.data.WordnetReader import WordnetReader
|
||
|
|
from main.data.TypedRelationInstances import TypedRelationInstances
|
||
|
|
from main.data.Vocabs import Vocabs
|
||
|
|
from main.data.Split import Split
|
||
|
|
from main.graphs.AdjacencyGraph import AdjacencyGraph
|
||
|
|
from main.features.PathExtractor import PathExtractor
|
||
|
|
from main.experiments.CVSMDriver import CVSMDriver
|
||
|
|
from main.experiments.PRADriver import PRADriver
|
||
|
|
from main.playground.make_data_format import process_paths
|
||
|
|
from main.playground.model2.CompositionalVectorAlgorithm import CompositionalVectorAlgorithm
|
||
|
|
|
||
|
|
# This script is used to run WNRR18 experiments (only our method) with 1:10 postive to negative ratio.
|
||
|
|
|
||
|
|
if __name__ == "__main__":
|
||
|
|
|
||
|
|
# location of Das et al.'s repo
|
||
|
|
CVSM_RUN_DIR = "/home/weiyu/Research/Path_Baselines/CVSM/ChainsofReasoning"
|
||
|
|
# location of Matt's PRA scala repo
|
||
|
|
PRA_RUN_DIR = "/home/weiyu/Research/Path_Baselines/SFE/pra_scala"
|
||
|
|
|
||
|
|
DATASET_NAME = "wn18rr"
|
||
|
|
DATASET_FOLDER = os.path.join("data", DATASET_NAME)
|
||
|
|
PRA_TEMPLATE_DIR = "pra_templates"
|
||
|
|
|
||
|
|
WORDNET_DIR = DATASET_FOLDER
|
||
|
|
DOMAIN_FILENAME = os.path.join(DATASET_FOLDER, "domains.tsv")
|
||
|
|
RANGE_FILENAME = os.path.join(DATASET_FOLDER, "ranges.tsv")
|
||
|
|
EDGES_FILENAME = os.path.join(DATASET_FOLDER, "edges.txt")
|
||
|
|
PRA_DIR = os.path.join(DATASET_FOLDER, "pra")
|
||
|
|
SPLIT_DIR = os.path.join(DATASET_FOLDER, "split")
|
||
|
|
CPR_PATH_DIR = os.path.join(DATASET_FOLDER, "cpr_paths")
|
||
|
|
WORD2VEC_FILENAME = "data/word2vec/GoogleNews-vectors-negative300.bin"
|
||
|
|
ENTITY2VEC_FILENAME = os.path.join(DATASET_FOLDER, "synonym2vec.pkl")
|
||
|
|
|
||
|
|
CVSM_DATA_DIR = os.path.join(DATASET_FOLDER, "cvsm")
|
||
|
|
PRA_PATH_DIR = os.path.join(DATASET_FOLDER, "pra_paths")
|
||
|
|
RELATION_PATH_DIR = os.path.join(DATASET_FOLDER, "relation_paths")
|
||
|
|
PATH_DIR = os.path.join(DATASET_FOLDER, "paths")
|
||
|
|
NEW_PATH_DIR = os.path.join(DATASET_FOLDER, "new_paths")
|
||
|
|
AUGMENT_PATH_DIR = os.path.join(DATASET_FOLDER, "paths_augment")
|
||
|
|
|
||
|
|
CVSM_RET_DIR = os.path.join(DATASET_FOLDER, "cvsm_entity")
|
||
|
|
ENTITY_TYPE2VEC_FILENAME = os.path.join(DATASET_FOLDER, "entity_type2vec.pkl")
|
||
|
|
|
||
|
|
run_step = 1
|
||
|
|
|
||
|
|
# 1. first run main/data/WordnetReader.py to process data and generate necessary files
|
||
|
|
if run_step == 1:
|
||
|
|
wn = WordnetReader(WORDNET_DIR,
|
||
|
|
filter=True,
|
||
|
|
word2vec_filename=WORD2VEC_FILENAME,
|
||
|
|
remove_repetitions=False)
|
||
|
|
wn.read_data()
|
||
|
|
wn.get_entity_types()
|
||
|
|
wn.write_relation_domain_and_ranges()
|
||
|
|
wn.write_edges()
|
||
|
|
|
||
|
|
# 2. use PRA scala code and code here to create train/test/dev split and negative examples
|
||
|
|
if run_step == 2:
|
||
|
|
pra_driver = PRADriver(DATASET_FOLDER, PRA_TEMPLATE_DIR, PRA_RUN_DIR, DATASET_NAME)
|
||
|
|
pra_driver.prepare_split()
|
||
|
|
|
||
|
|
# 3. Extract paths with entities
|
||
|
|
if run_step == 3:
|
||
|
|
typed_relation_instances = TypedRelationInstances()
|
||
|
|
typed_relation_instances.read_domains_and_ranges(DOMAIN_FILENAME, RANGE_FILENAME)
|
||
|
|
typed_relation_instances.construct_from_labeled_edges(EDGES_FILENAME, entity_name_is_typed=False,
|
||
|
|
is_labeled=False)
|
||
|
|
vocabs = Vocabs()
|
||
|
|
vocabs.build_vocabs(typed_relation_instances)
|
||
|
|
split = Split()
|
||
|
|
split.read_splits(SPLIT_DIR, vocabs, entity_name_is_typed=True)
|
||
|
|
graph = AdjacencyGraph()
|
||
|
|
graph.build_graph(typed_relation_instances, vocabs)
|
||
|
|
|
||
|
|
path_extractor = PathExtractor(max_length=6, include_entity=True, save_dir=PATH_DIR, include_path_len1=True,
|
||
|
|
max_paths_per_pair=200, multiple_instances_per_pair=False,
|
||
|
|
max_instances_per_pair=None, paths_sample_method="all_lengths")
|
||
|
|
path_extractor.extract_paths(graph, split, vocabs)
|
||
|
|
path_extractor.write_paths(split)
|
||
|
|
|
||
|
|
# 4. Process data for running the model
|
||
|
|
if run_step == 4:
|
||
|
|
# first convert paths to cvsm format
|
||
|
|
cvsm_driver = CVSMDriver(WORDNET_DIR, CVSM_RUN_DIR, dataset="wordnet", include_entity=True, has_entity=True,
|
||
|
|
augment_data=False, include_entity_type=True)
|
||
|
|
cvsm_driver.setup_cvsm_dir()
|
||
|
|
|
||
|
|
# then vectorize cvsm format data
|
||
|
|
process_paths(input_dir=os.path.join(CVSM_RET_DIR, "data/data_input"),
|
||
|
|
output_dir=os.path.join(CVSM_RET_DIR, "data/data_output"),
|
||
|
|
vocab_dir=os.path.join(CVSM_RET_DIR, "data/vocab"),
|
||
|
|
isOnlyRelation=False,
|
||
|
|
getOnlyRelation=False,
|
||
|
|
MAX_POSSIBLE_LENGTH_PATH=8, # the max number of relations in a path + 1
|
||
|
|
NUM_ENTITY_TYPES_SLOTS=15, # the number of types + 1 (the reason we +1 is to create a meaningless type for all entities)
|
||
|
|
pre_padding=True)
|
||
|
|
|
||
|
|
# 5. Run the model
|
||
|
|
# use $tensorboard --logdir runs to see the training progress
|
||
|
|
if run_step == 5:
|
||
|
|
cvsm = CompositionalVectorAlgorithm("wordnet", CVSM_RET_DIR, ENTITY_TYPE2VEC_FILENAME)
|
||
|
|
|
||
|
|
# Not using pretrained word embeddings decreases performance
|
||
|
|
# cvsm = CompositionalVectorAlgorithm("wordnet", CVSM_RET_DIR, None)
|
||
|
|
|
||
|
|
cvsm.train_and_test()
|
||
|
|
|
||
|
|
# Uncomment if need to train only one relation
|
||
|
|
# cvsm.train("/home/weiyu/Research/ChainsOfReasoningWithAbstractEntities/data/wn18rr/cvsm_entity/data/data_output/member_of_domain_region")
|
||
|
|
|
||
|
|
if run_step == 6:
|
||
|
|
cvsm = CompositionalVectorAlgorithm("wordnet", CVSM_RET_DIR, ENTITY_TYPE2VEC_FILENAME,
|
||
|
|
pooling_method="sat", attention_method="sat", early_stopping_metric="map",
|
||
|
|
visualize=True, calculate_path_attn_stats=True, calculate_type_attn_stats=True,
|
||
|
|
best_models={'verb_group': {'val_acc': 1.0, 'val_ap': 1.0, 'epoch': 0, 'test_ap': 1.0, 'test_acc': 1.0}, 'member_meronym': {'val_acc': 0.9431578947368421, 'val_ap': 0.7135667457942702, 'epoch': 19, 'test_ap': 0.6335514032344876, 'test_acc': 0.9408812046848857}, 'hypernym': {'val_acc': 0.989671984536826, 'val_ap': 0.9642082965792328, 'epoch': 16, 'test_ap': 0.9620883932417185, 'test_acc': 0.988660197755088}, 'also_see': {'val_acc': 0.9683306494900698, 'val_ap': 0.9301494111857955, 'epoch': 7, 'test_ap': 0.904053400950515, 'test_acc': 0.9729148753224419}, 'similar_to': {'val_acc': 0.9795918367346939, 'val_ap': 1.0, 'epoch': 3, 'test_ap': 1.0, 'test_acc': 0.9844961240310077}, 'member_of_domain_region': {'val_acc': 0.9590865842055185, 'val_ap': 0.7790840930128, 'epoch': 13, 'test_ap': 0.6968178289261723, 'test_acc': 0.954858454475899}, 'instance_hypernym': {'val_acc': 0.9630209965528047, 'val_ap': 0.8795758247163307, 'epoch': 8, 'test_ap': 0.8778889539424873, 'test_acc': 0.9612403100775194}, 'synset_domain_topic_of': {'val_acc': 0.9447174447174447, 'val_ap': 0.7312231001822086, 'epoch': 13, 'test_ap': 0.7427533669498867, 'test_acc': 0.9436564223798266}, 'derivationally_related_form': {'val_acc': 1.0, 'val_ap': 1.0, 'epoch': 0, 'test_ap': 1.0, 'test_acc': 1.0}, 'has_part': {'val_acc': 0.9431347849559114, 'val_ap': 0.7213513082273592, 'epoch': 13, 'test_ap': 0.6589302705002684, 'test_acc': 0.9376080691642651}, 'member_of_domain_usage': {'val_acc': 0.9627403846153846, 'val_ap': 0.9055411128578176, 'epoch': 6, 'test_ap': 0.8644352979656229, 'test_acc': 0.957487922705314}})
|
||
|
|
cvsm.train_and_test()
|