File size: 5,527 Bytes
486585a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 |
from collections import defaultdict
import fasttext
import pandas as pd
from sklearn.metrics import classification_report
from tqdm import tqdm; tqdm.pandas()
#!pip install tabulate
import io
from pathlib import Path
import numpy as np
import pandas as pd
import requests
from sklearn.metrics import accuracy_score
from sklearn.metrics import classification_report
from sklearn.metrics import precision_recall_fscore_support
names = pd.read_csv(
io.StringIO(requests.get("https://iso639-3.sil.org/sites/iso639-3/files/downloads/iso-639-3.tab").text
), sep="\t").set_index("Id").rename(
columns={"Ref_Name": "name"}
)[["name"]].to_dict()["name"]
tato_names = pd.read_html(
"https://tatoeba.org/en/stats/sentences_by_language"
)[0].rename(
columns={"Unnamed: 2": "code", "Language": "name"}
).set_index("code")[["name"]].to_dict()["name"]
names.update(tato_names)
# langs = pd.read_csv("train.csv").lang.unique().tolist()
# langs_df = pd.DataFrame({"ISO-639-3": langs}).sort_values("ISO-639-3")
# langs_df["Language"] = langs_df["ISO-639-3"].apply(names.__getitem__)
# langs_df = langs_df.set_index("ISO-639-3")
def pandas_classification_report(y_true, y_pred, labels=None):
metrics_summary = precision_recall_fscore_support(
y_true=y_true,
y_pred=y_pred,
labels=labels)
weighted_avg = list(precision_recall_fscore_support(
y_true=y_true,
y_pred=y_pred,
labels=labels,
average='weighted'))
macro_avg = list(precision_recall_fscore_support(
y_true=y_true,
y_pred=y_pred,
labels=labels,
average='macro'))
accuracy = [np.nan, np.nan, accuracy_score(y_true=y_true, y_pred=y_pred), np.nan]
metrics_sum_index = ['precision', 'recall', 'f1-score', 'support']
class_report_df = pd.DataFrame(
list(metrics_summary),
index=metrics_sum_index,
columns=labels)
support = class_report_df.loc['support']
total = support.sum()
weighted_avg[-1] = total
macro_avg[-1] = total
accuracy[-1] = total
class_report_df['accuracy'] = accuracy
class_report_df['weighted avg'] = weighted_avg
class_report_df['macro avg'] = macro_avg
report = class_report_df.T
report["support"] = report["support"].astype(int)
return report
scores_text = ""
for model_name in ("nordic-lid.bin", "nordic-lid_all.bin"):
print(
f"""
------------
{model_name}
------------
""")
model = fasttext.load_model(model_name)
train = pd.read_csv("train.csv")
ddict = defaultdict(lambda: "---")
for k in train.lang.unique().tolist():
ddict[k] = k
train["nordic-lid"] = train.progress_apply(lambda row: ddict[model.predict(row["text"].replace("\n", " "))[0][0][-3:]], axis=1)
print("TRAIN")
print(model.test("train.txt"))
print(classification_report(train["lang"], train["nordic-lid"], digits=4))
val = pd.read_csv("validation.csv")
val["nordic-lid"] = val.progress_apply(lambda row: ddict[model.predict(row["text"].replace("\n", " "))[0][0][-3:]], axis=1)
print("VALIDATION")
print(model.test("validation.txt"))
print(classification_report(val["lang"], val["nordic-lid"], digits=4))
test = pd.read_csv("test.csv")
test["nordic-lid"] = test.progress_apply(lambda row: ddict[model.predict(row["text"].replace("\n", " "))[0][0][-3:]], axis=1)
print("TEST")
print(model.test("test.txt"))
print(classification_report(test["lang"], test["nordic-lid"], digits=4))
if "_all" in model_name:
train = pd.read_csv("train_all.csv")
ddict = defaultdict(lambda: "---")
for k in train.lang.unique().tolist():
ddict[k] = k
train["nordic-lid"] = train.progress_apply(lambda row: ddict[model.predict(row["text"].replace("\n", " "))[0][0][-3:]], axis=1)
print("TRAIN ALL")
print(model.test("train_all.txt"))
print(classification_report(train["lang"], train["nordic-lid"], digits=4))
val = pd.read_csv("validation_all.csv")
val["nordic-lid"] = val.progress_apply(lambda row: ddict[model.predict(row["text"].replace("\n", " "))[0][0][-3:]], axis=1)
print("VALIDATION ALL")
print(model.test("validation_all.txt"))
print(classification_report(val["lang"], val["nordic-lid"], digits=4))
test = pd.read_csv("test_all.csv")
test["nordic-lid"] = test.progress_apply(lambda row: ddict[model.predict(row["text"].replace("\n", " "))[0][0][-3:]], axis=1)
print("TEST ALL")
print(model.test("test_all.txt"))
print(classification_report(test["lang"], test["nordic-lid"], digits=4))
langs = pd.read_csv("train_all.csv").lang.unique().tolist()
else:
langs = pd.read_csv("train.csv").lang.unique().tolist()
langs_df = pd.DataFrame({"ISO-639-3": langs}).sort_values("ISO-639-3")
langs_df["Language"] = langs_df["ISO-639-3"].apply(names.__getitem__)
langs_df = langs_df.set_index("ISO-639-3")
report_df = pandas_classification_report(test["nordic-lid"], test["lang"], sorted(langs))
scores = report_df.join(langs_df)
scores.columns = map(str.title, scores.columns)
scores.index.name = "ISO-639-3"
scores = scores[["Language"] + [col.title() for col in scores.columns if col != "Language"]]
scores_text += f"## {model_name}\n\n{scores.reset_index().to_markdown(index=False, floatfmt='.4f')}\n\n"
print()
print(scores_text)
Path("./scores.md").write_text(scores_text)
|