class Main:
def __init__(self):
self.m_preprocessor = Preprocessor()
self.m_similarity = Similarity()
self.m_plt = Plot()
self.m_evaluator = Evaluator()
self.m_file = "sts-train"
self.m_metric = "path"
self.m_ic = "brown"
self.m_metric_w2v = "cosine"
self.m_metric_t = "path"
self.m_thr = 20
self.m_mode = "ontology"
def arg_parse(self):
parser = argparse.ArgumentParser(description="Calculus of similarity",
add_help=True)
parser.add_argument(
"-f",
dest="file",
action="store",
type=str,
choices=["sts-dev", "sts-test", "sts-train"],
help="Select the csv file to calculate similarity.",
)
parser.add_argument(
"-m",
dest="metric",
action="store",
type=str,
choices=["path", "lch", "wup", "res", "jcn", "lin"],
help="Select the metric to calculate similarity.",
)
parser.add_argument(
"-i",
dest="ic",
action="store",
type=str,
choices=["brown", "treebank"],
help="Select the Information Content corpus.",
)
parser.add_argument(
"-mw",
dest="metric_w2v",
action="store",
type=str,
choices=["cosine", "euclidean", "dot"],
help="Select the metric to calculate similarity word2vec.",
)
parser.add_argument(
"-mt",
dest="metric_t",
action="store",
type=str,
choices=["path", "jcn", "euclidean"],
help="Select the metric to calculate textual similarity.",
)
parser.add_argument(
"-t",
dest="thr",
action="store",
type=int,
choices=[20, 30, 40],
help="Select the threshold to calculate textual similarity.",
)
args = parser.parse_args()
if args.file != None:
self.m_file = args.file
if args.metric != None:
self.m_mode = "ontology"
self.m_metric = args.metric
if args.ic != None:
self.m_ic = args.ic
if args.metric_w2v != None:
self.m_mode = "word2vec"
self.m_metric_w2v = args.metric_w2v
if args.metric_t != None:
self.m_mode = "textual"
self.m_metric_t = args.metric_t
if args.thr != None:
self.m_thr = args.thr
self.m_data_frame = self.m_preprocessor.initialize(
"..\\stsbenchmark\\" + self.m_file + ".csv")
def main(self):
df_original = self.m_data_frame.copy()
stats_original = self.data_frame_stats(self.m_data_frame)
print("Printing original stats")
print(stats_original)
df_filtered = self.m_preprocessor.tokenize_and_filter_dataset(
self.m_data_frame)
print("Calcul similarity by " + self.m_mode + " similarity, using " +
self.m_metric_t)
if self.m_mode == "ontology":
self.m_data_frame = self.m_similarity.calculate_ontology_similarity(
self.m_metric, self.m_ic, df_filtered)
elif self.m_mode == "word2vec":
self.m_data_frame = self.m_similarity.calculate_word2vec_similarity(
self.m_metric_w2v, df_filtered)
elif self.m_mode == "textual":
idf = self.m_preprocessor.idf_from_dataset(self.m_data_frame)
self.m_data_frame = self.m_similarity.calculate_textual_similarity(
self.m_thr, self.m_metric_t, df_filtered, idf)
df_syntetic = self.m_data_frame.copy()
stats_syntetic = self.data_frame_stats(self.m_data_frame)
stats_total = stats_original.join(stats_syntetic,
lsuffix="_o",
rsuffix="_s")
print("Printing all stats")
print(stats_total)
#self.m_plt.plot_scatter(stats_total)
self.m_evaluator.evaluate(df_original, df_syntetic)
def plot_base_data(self):
for row in self.m_data_frame.itertuples():
self.m_plt.prepare_data_to_plot(row.set, row.similarity)
self.m_plt.sub_plot_bar(self.m_file)
def plot_pos_tagged_data(self):
df = self.m_preprocessor.tag_dataset(self.m_data_frame)
for row in df.itertuples():
self.m_plt.prepare_tags_to_plot(row.set, row.sentence_a,
row.sentence_b)
self.m_plt.sub_plot_bar(self.m_file)
def calculate_sentences_length(self):
df = self.m_preprocessor.tag_dataset(self.m_data_frame)
lengths = {}
for row in df.itertuples():
if lengths.get(row.set, []) == []:
lengths[row.set] = [len(row.sentence_a), len(row.sentence_b)]
continue
lengths[row.set].append(len(row.sentence_a))
lengths[row.set].append(len(row.sentence_b))
for key in lengths.keys():
lengths[key] = mean(lengths[key])
print(lengths)
def data_frame_stats(self, df):
return df.groupby(["set"])["similarity"].agg(["mean", "std", "median"])
