Spaces:
Runtime error
Runtime error
File size: 5,411 Bytes
010bd36 c28d81d 010bd36 1358c21 010bd36 10bdf14 494bd12 010bd36 494bd12 010bd36 d405df4 10bdf14 494bd12 10bdf14 494bd12 10bdf14 494bd12 |
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 |
import pandas as pd
from datasets import load_dataset
import streamlit as st
from clarin_datasets.dataset_to_show import DatasetToShow
class AspectEmoDataset(DatasetToShow):
def __init__(self):
DatasetToShow.__init__(self)
self.dataset_name = "clarin-pl/aspectemo"
self.description = """
AspectEmo Corpus is an extended version of a publicly available PolEmo 2.0
corpus of Polish customer reviews used in many projects on the use of different methods in sentiment
analysis. The AspectEmo corpus consists of four subcorpora, each containing online customer reviews from the
following domains: school, medicine, hotels, and products. All documents are annotated at the aspect level
with six sentiment categories: strong negative (minus_m), weak negative (minus_s), neutral (zero),
weak positive (plus_s), strong positive (plus_m).
Tasks (input, output and metrics)
Aspect-based sentiment analysis (ABSA) is a text analysis method that
categorizes data by aspects and identifies the sentiment assigned to each aspect. It is the sequence tagging
task.
Input ('tokens' column): sequence of tokens
Output ('labels' column): sequence of predicted tokens’ classes ("O" + 6 possible classes: strong negative (
a_minus_m), weak negative (a_minus_s), neutral (a_zero), weak positive (a_plus_s), strong positive (
a_plus_m), ambiguous (a_amb) )
Domain: school, medicine, hotels and products
Measurements:
Example: ['Dużo', 'wymaga', ',', 'ale', 'bardzo', 'uczciwy', 'i', 'przyjazny', 'studentom', '.', 'Warto', 'chodzić',
'na', 'konsultacje', '.', 'Docenia', 'postępy', 'i', 'zaangażowanie', '.', 'Polecam', '.'] → ['O', 'a_plus_s', 'O',
'O', 'O', 'a_plus_m', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'a_zero', 'O', 'a_plus_m', 'O', 'O', 'O', 'O', 'O', 'O']
"""
def load_data(self):
raw_dataset = load_dataset(self.dataset_name)
self.data_dict = {
subset: raw_dataset[subset].to_pandas() for subset in self.subsets
}
def show_dataset(self):
header = st.container()
description = st.container()
dataframe_head = st.container()
class_distribution = st.container()
most_common_tokens = st.container()
with header:
st.title(self.dataset_name)
with description:
st.header("Dataset description")
st.write(self.description)
full_dataframe = pd.concat(self.data_dict.values(), axis="rows")
tokens_all = full_dataframe["tokens"].tolist()
tokens_all = [x for subarray in tokens_all for x in subarray]
labels_all = full_dataframe["labels"].tolist()
labels_all = [x for subarray in labels_all for x in subarray]
with dataframe_head:
df_to_show = full_dataframe.head(10)
st.header("First 10 observations of the dataset")
st.dataframe(df_to_show)
st.text_area(label="LaTeX code", value=df_to_show.style.to_latex())
class_distribution_dict = {}
for subset in self.subsets:
all_labels_from_subset = self.data_dict[subset]["labels"].tolist()
all_labels_from_subset = [
x for subarray in all_labels_from_subset for x in subarray if x != 0
]
all_labels_from_subset = pd.Series(all_labels_from_subset)
class_distribution_dict[subset] = (
all_labels_from_subset.value_counts(normalize=True)
.sort_index()
.reset_index()
.rename({"index": "class", 0: subset}, axis="columns")
)
class_distribution_df = pd.merge(
class_distribution_dict["train"],
class_distribution_dict["test"],
on="class",
)
with class_distribution:
st.header("Class distribution in each subset (without '0')")
st.dataframe(class_distribution_df)
st.text_area(
label="LaTeX code", value=class_distribution_df.style.to_latex()
)
# Most common tokens from selected class (without 0)
full_df_unzipped = pd.DataFrame(
{
"token": tokens_all,
"label": labels_all,
}
)
full_df_unzipped = full_df_unzipped.loc[full_df_unzipped["label"] != 0]
possible_options = sorted(full_df_unzipped["label"].unique())
with most_common_tokens:
st.header("10 most common tokens from selected class (without '0')")
selected_class = st.selectbox(
label="Select class to show", options=possible_options
)
df_to_show = (
full_df_unzipped.loc[full_df_unzipped["label"] == selected_class]
.groupby(["token"])
.count()
.reset_index()
.rename({"label": "no_of_occurrences"}, axis=1)
.sort_values(by="no_of_occurrences", ascending=False)
.reset_index(drop=True)
.head(10)
)
st.dataframe(df_to_show)
st.text_area(label="LaTeX code", value=df_to_show.style.to_latex())
|