Spaces:
Running
Running
| import pandas as pd | |
| import streamlit as st | |
| from app_utils import filter_dataframe, calculate_height_to_display | |
| from contants import INFO_CATALOG, CITATION_CATALOG, HOWTO_CATALOG,INFO_BENCHMARK, CITATION_BENCHMARK, HOWTO_BENCHMARK, INFO_MAIN, CITATION_MAIN, HOWTO_TAXONOMY_CAT | |
| from utils import BASE_SUMMARY_METRICS | |
| from utils import load_data_catalog, load_data_taxonomy, load_bench_catalog, load_bench_taxonomy | |
| from utils import datasets_count_and_size, datasets_count_and_size_standard, metadata_coverage, catalog_summary_statistics | |
| from utils import left_align, right_align | |
| import matplotlib.pyplot as plt | |
| import seaborn as sns | |
| st.set_page_config(layout="wide") | |
| # Load PL ASR data survey data | |
| # Cache the dataframe so it's only loaded once | |
| df_data_cat = load_data_catalog() | |
| df_data_tax = load_data_taxonomy() | |
| # Filter out non available datasets | |
| df_data_cat_available = df_data_cat[df_data_cat['Available online'] == 'yes'] | |
| # Available and free | |
| df_data_cat_available_free = df_data_cat[(df_data_cat['Available online'] == 'yes') & (df_data_cat['Price - non-commercial usage'] == 'free')] | |
| # Available and paid | |
| df_data_cat_available_paid = df_data_cat[(df_data_cat['Available online'] == 'yes') & (df_data_cat['Price - non-commercial usage'] != 'free')] | |
| # Load PL ASR benchmarks survey data | |
| df_bench_cat = load_bench_catalog() | |
| df_bench_tax = load_bench_taxonomy() | |
| about, data_cat, data_survey, data_taxonomy, bench_cat, bench_survey, bench_taxonomy = st.tabs(["PL ASR survey", "PL ASR Speech Data **Catalog**", "PL ASR Speech data **Survey**", "ASR Speech Data **Taxonomy**", "PL ASR Benchmarks Catalog", "PL ASR Benchmarks Survey", "ASR Benchmarks Taxonomy"]) | |
| with about: | |
| st.title("About Polish ASR Survey") | |
| st.markdown(INFO_MAIN, unsafe_allow_html=True) | |
| st.header("How to cite this resource?") | |
| st.markdown(CITATION_MAIN, unsafe_allow_html=True) | |
| with data_cat: | |
| st.title("Polish ASR Speech Datasets Catalog") | |
| st.markdown(INFO_CATALOG, unsafe_allow_html=True) | |
| st.header("How to use?") | |
| st.markdown(HOWTO_CATALOG, unsafe_allow_html=True) | |
| # Display catalog contents | |
| st.header("Browse the catalog content") | |
| st.dataframe(filter_dataframe(df_data_cat, "datasets"), hide_index=True, use_container_width=True) | |
| st.header("How to cite this resource?") | |
| st.markdown(CITATION_CATALOG, unsafe_allow_html=True) | |
| with data_survey: | |
| # Display summary statistics | |
| st.title("Polish ASR Speech Datasets Survey") | |
| st.header("Polish ASR speech datasets summary statistics") | |
| df_summary_metrics = catalog_summary_statistics(df_data_cat) | |
| df_basic_stats = df_summary_metrics.loc[BASE_SUMMARY_METRICS[0:5]] | |
| st.dataframe(df_basic_stats, use_container_width=False) | |
| st.header("Speech data available across Polish ASR speech datasets") | |
| df_stats_audio_available = df_summary_metrics.loc[BASE_SUMMARY_METRICS[5:10]] | |
| st.dataframe(df_stats_audio_available, use_container_width=False) | |
| st.header("Transcribed data available across Polish ASR speech datasets") | |
| df_stats_transcribed_available = df_summary_metrics.loc[BASE_SUMMARY_METRICS[10:15]] | |
| st.dataframe(df_stats_transcribed_available, use_container_width=False) | |
| # Display distribution of datasets created per year | |
| st.header("Polish ASR speech datasets created in 1997-2023") | |
| col_groupby = ['Creation year'] | |
| df_datasets_per_year = datasets_count_and_size(df_data_cat, col_groupby, col_sort=col_groupby, col_percent=None, col_sum=['Size audio transcribed [hours]','Audio recordings', 'Speakers'], col_count = ['Dataset ID']) | |
| st.dataframe(df_datasets_per_year, use_container_width=False) | |
| st.header("Institutions contributing Polish ASR speech dataset") | |
| col_groupby = ['Publisher'] | |
| df_datasets_per_publisher = datasets_count_and_size(df_data_cat, col_groupby, col_sort='Count Dataset ID', col_percent=None, col_sum=['Size audio transcribed [hours]','Audio recordings', 'Speakers'], col_count = ['Dataset ID']) | |
| st.dataframe(df_datasets_per_publisher, use_container_width=False) | |
| st.header("Repositories hosting Polish ASR speech datasets") | |
| col_groupby = ['Repository'] | |
| df_datasets_per_repo = datasets_count_and_size(df_data_cat, col_groupby, col_sort='Count Dataset ID', col_percent=None, col_sum=['Size audio transcribed [hours]','Audio recordings', 'Speakers'], col_count = ['Dataset ID']) | |
| st.dataframe(df_datasets_per_repo, use_container_width=False) | |
| st.header("Public domain Polish ASR speech datasets") | |
| col_groupby = ['License', "Dataset ID"] | |
| df_datasets_public = datasets_count_and_size(df_data_cat_available_free, col_groupby, col_sort='License', col_percent=None, col_sum=['Size audio transcribed [hours]','Audio recordings', 'Speakers'], col_count = []) | |
| st.dataframe(df_datasets_public, use_container_width=False) | |
| st.header("Commercialy available Polish ASR speech datasets") | |
| col_groupby = ['License', "Dataset ID"] | |
| df_datasets_paid = datasets_count_and_size(df_data_cat_available_paid, col_groupby, col_sort='License', col_percent=None, col_sum=['Size audio transcribed [hours]','Audio recordings', 'Speakers'], col_count = []) | |
| st.dataframe(df_datasets_paid, use_container_width=False) | |
| st.header("Coverage of metadata across Polish ASR speech datasets") | |
| df_meta_all_flat, df_meta_all_pivot = metadata_coverage(df_data_cat, df_data_cat_available_free, df_data_cat_available_paid) | |
| st.dataframe(df_meta_all_pivot, use_container_width=False) | |
| # Display distribution of datasets for various speech types | |
| st.header("Datasets per speech type") | |
| col_groupby = ['Speech type'] | |
| df_datasets_per_speech_type = datasets_count_and_size(df_data_cat, col_groupby, col_sort=col_groupby, col_percent = ['Size audio transcribed [hours]'], col_sum = ['Size audio transcribed [hours]','Audio recordings', 'Speakers'], col_count = ['Dataset ID']) | |
| st.dataframe(df_datasets_per_speech_type, use_container_width=False) | |
| # Display distribution of datasets for various speech types | |
| st.header("Distribution of available speech data per audio device - Public domain datasets") | |
| col_groupby = ['Audio device'] | |
| df_datasets_per_device = datasets_count_and_size(df_data_cat_available_free, col_groupby, col_sort=col_groupby, col_percent = ['Size audio transcribed [hours]'], col_sum = ['Size audio transcribed [hours]','Audio recordings', 'Speakers'], col_count = ['Dataset ID']) | |
| st.dataframe(df_datasets_per_device, use_container_width=False) | |
| # Display distribution of datasets for various speech types | |
| st.header("Distribution of available speech data per audio device - Commercial datasets") | |
| col_groupby = ['Audio device'] | |
| df_datasets_per_device = datasets_count_and_size(df_data_cat_available_paid, col_groupby, col_sort=col_groupby, col_percent = ['Size audio transcribed [hours]'], col_sum = ['Size audio transcribed [hours]','Audio recordings', 'Speakers'], col_count = ['Dataset ID']) | |
| st.dataframe(df_datasets_per_device, use_container_width=False) | |
| with data_taxonomy: | |
| st.title("Polish ASR Speech Data Taxonomy") | |
| st.header("How to use?") | |
| st.markdown(HOWTO_TAXONOMY_CAT, unsafe_allow_html=True) | |
| st.dataframe(df_data_tax, hide_index=True, use_container_width=True) | |
| st.header("How to cite?") | |
| st.markdown(CITATION_CATALOG, unsafe_allow_html=True) | |
| with bench_cat: | |
| st.write("Benchmarks catalog") | |
| # TODO - load and display benchmarks catalog | |
| st.title("Polish ASR Benchmarks Catalog") | |
| # Display catalog contents | |
| st.dataframe(filter_dataframe(df_bench_cat, "benchmarks"), hide_index=True, use_container_width=True) | |
| # Display taxonomy contents |