clarin_datasets/cst_wikinews_dataset.py

import pandas as pd
from datasets import load_dataset
import streamlit as st

from clarin_datasets.dataset_to_show import DatasetToShow


class CSTWikinewsDataset(DatasetToShow):
    def __init__(self):
        DatasetToShow.__init__(self)
        self.dataset_name = "clarin-pl/cst-wikinews"
        self.description = f"""
        Dataset link: https://huggingface.co/datasets/{self.dataset_name}
        """

    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()
        dataframe_head = st.container()
        class_distribution = st.container()

        with header:
            st.title(self.dataset_name)

        with dataframe_head:
            st.header("First 10 observations of the chosen subset")
            subset_to_show = st.selectbox(
                label="Select subset to see", options=self.subsets
            )
            df_to_show = self.data_dict[subset_to_show].head(10)
            st.dataframe(df_to_show)
            st.text_area(label="LaTeX code", value=df_to_show.style.to_latex())

        class_distribution_df = pd.merge(
            pd.DataFrame(
                self.data_dict["train"]["label"]
                .value_counts(normalize=True)
                .reset_index(drop=False)
                .rename({"index": "class"}, axis="columns")
            ),
            pd.DataFrame(
                self.data_dict["test"]["label"]
                .value_counts(normalize=True)
                .reset_index(drop=False)
                .rename({"index": "class"}, axis="columns")
            ),
            on="class",
        ).rename({"label_x": "train", "label_y": "test"}, axis="columns")

        with class_distribution:
            st.dataframe(class_distribution_df)