{ "cells": [ { "cell_type": "code", "execution_count": 3, "id": "5ef79f73", "metadata": {}, "outputs": [], "source": [ "from pydub import AudioSegment\n", "import pandas as pd" ] }, { "cell_type": "markdown", "id": "f2ee315b", "metadata": {}, "source": [ "## Multilingual Spoken Word Corpus" ] }, { "cell_type": "code", "execution_count": 4, "id": "d629e053", "metadata": {}, "outputs": [], "source": [ "DATA_DIR = \"/mnt/data-ssd-1/data/spokenwords/mswc_microset/en/\"" ] }, { "cell_type": "code", "execution_count": 13, "id": "08a86ea6", "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", " \n", " " ], "text/plain": [ "" ] }, "execution_count": 13, "metadata": {}, "output_type": "execute_result" } ], "source": [ "AudioSegment.from_file(DATA_DIR + \"mswc_microset/en/clips/dog/common_voice_en_91987.opus\", codec=\"opus\")" ] }, { "cell_type": "code", "execution_count": 28, "id": "caa07c16", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
LINKWORDVALIDSPEAKERGENDER
0backward/common_voice_en_18751858.opusbackwardTrue90a2539a7010c22f8d1379231f2b3e5829a0e645060d82...MALE
1backward/common_voice_en_19346834.opusbackwardTrueda4bd5bc7a79e8f1140a50f2abecb7311ca2aa154f7fc5...MALE
\n", "
" ], "text/plain": [ " LINK WORD VALID \\\n", "0 backward/common_voice_en_18751858.opus backward True \n", "1 backward/common_voice_en_19346834.opus backward True \n", "\n", " SPEAKER GENDER \n", "0 90a2539a7010c22f8d1379231f2b3e5829a0e645060d82... MALE \n", "1 da4bd5bc7a79e8f1140a50f2abecb7311ca2aa154f7fc5... MALE " ] }, "execution_count": 28, "metadata": {}, "output_type": "execute_result" } ], "source": [ "val_df = pd.read_csv(DATA_DIR + \"en_dev.csv\")\n", "val_df.head(2)" ] }, { "cell_type": "markdown", "id": "4c42bf61", "metadata": {}, "source": [ "## People's Speech" ] }, { "cell_type": "code", "execution_count": 5, "id": "01df541b", "metadata": {}, "outputs": [], "source": [ "DATA_DIR = \"/mnt/data-ssd-1/data/peoplespeech/\"" ] }, { "cell_type": "code", "execution_count": 32, "id": "416e009e", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
audio_filepathdurationshard_idtext
0Peters_Twp_DOT__School_Board_-_Policy_Committe...14.89-1are there any other comments but you would don...
1Peters_Twp_DOT__School_Board_-_Policy_Committe...6.24-1i'm wondering what the references to policy ei...
\n", "
" ], "text/plain": [ " audio_filepath duration shard_id \\\n", "0 Peters_Twp_DOT__School_Board_-_Policy_Committe... 14.89 -1 \n", "1 Peters_Twp_DOT__School_Board_-_Policy_Committe... 6.24 -1 \n", "\n", " text \n", "0 are there any other comments but you would don... \n", "1 i'm wondering what the references to policy ei... " ] }, "execution_count": 32, "metadata": {}, "output_type": "execute_result" } ], "source": [ "df = pd.read_json(DATA_DIR + \"flac_train_manifest.jsonl\", lines=True)\n", "df.head(2)" ] }, { "cell_type": "code", "execution_count": null, "id": "fa73a897", "metadata": {}, "outputs": [], "source": [ "# gov_DOT_uscourts_DOT_ca7_DOT_17-3351\n", "# gov_DOT_house_DOT_sgw_DOT_20090304\n", "# snafuinfinityPlanningBoard_102213\n", "# WBA_vs_DOT__Weston_MA_-_January_4_2015\n", "# Storytelling_-_8th_Immigrant_and_Minority_Farmers_Conference\n", "# hrs06SCI2318_100120\n", "# 140722EileenStewartOnIntlDevelopment" ] }, { "cell_type": "code", "execution_count": 80, "id": "bcf21e5a", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "cctv17vt-4_2_18_Winooski_City_Council/4_2_18_Winooski_City_Council_DOT_mp3_00350.flac\n", "\n", "significant harmful full impacts to a new ski residence through potential threats to future funding whereas when you city and state of vermont policymakers and government ploys are reliant on accurate population counts and census data to inform local and state policy\n" ] }, { "data": { "text/html": [ "\n", " \n", " " ], "text/plain": [ "" ] }, "execution_count": 80, "metadata": {}, "output_type": "execute_result" } ], "source": [ "row = df[~df['audio_filepath'].str.split(\"/\").str[0].str.contains(r\"DOT\")].sample(1).iloc[0]\n", "print(row['audio_filepath'])\n", "print()\n", "print(row[\"text\"])\n", "AudioSegment.from_file(DATA_DIR + \"subset/\" + row['audio_filepath'], codec=\"flac\")" ] }, { "cell_type": "code", "execution_count": 46, "id": "dd1194fb", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
audio_document_ididentifiertext_document_idtraining_data
0gov.uscourts.ca9.04-56618.2006-02-16.mp3gov.uscourts.ca9.04-56618gov.uscourts.ca9.04-56618.2006-02-16.asr.srt{'duration_ms': [14610, 14920, 2190, 14800, 14...
1gov.uscourts.ca9.05-55072.2008-05-06.mp3gov.uscourts.ca9.05-55072gov.uscourts.ca9.05-55072.2008-05-06.asr.srt{'duration_ms': [14370, 14920, 14740, 14860, 1...
\n", "
" ], "text/plain": [ " audio_document_id identifier \\\n", "0 gov.uscourts.ca9.04-56618.2006-02-16.mp3 gov.uscourts.ca9.04-56618 \n", "1 gov.uscourts.ca9.05-55072.2008-05-06.mp3 gov.uscourts.ca9.05-55072 \n", "\n", " text_document_id \\\n", "0 gov.uscourts.ca9.04-56618.2006-02-16.asr.srt \n", "1 gov.uscourts.ca9.05-55072.2008-05-06.asr.srt \n", "\n", " training_data \n", "0 {'duration_ms': [14610, 14920, 2190, 14800, 14... \n", "1 {'duration_ms': [14370, 14920, 14740, 14860, 1... " ] }, "execution_count": 46, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# clean and dirty manifest are the same\n", "meta_df = pd.read_json(DATA_DIR + \"manifest_clean.json\", lines=True)\n", "meta_df.head(2)" ] }, { "cell_type": "code", "execution_count": 47, "id": "38ace7ad", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "gov_DOT_uscourts_DOT_ca9_DOT_04-56618/gov_DOT_uscourts_DOT_ca9_DOT_04-56618_DOT_2006-02-16_DOT_mp3_00000.flac\n", "\n", "is there anything to wear it on this one or have your our colloquy with you already developed what you need on this this is the attorney's fees yes or i do want to point out to you simply that\n" ] } ], "source": [ "# print(meta_df.iloc[0]['training_data']['duration_ms'][0])\n", "# print()\n", "print(meta_df.iloc[0]['training_data']['name'][0])\n", "print()\n", "print(meta_df.iloc[0]['training_data']['label'][0])" ] }, { "cell_type": "code", "execution_count": 59, "id": "c006fe9f", "metadata": {}, "outputs": [], "source": [ "import funcy\n", "import string\n", "\n", "unigrams = list(funcy.flatten(meta_df['identifier'].str.split(r\"[\\.\\-\\_]\").values.tolist()))\n", "unigrams = [w.lower().strip() for w in unigrams if len(w.lower().strip()) > 0]\n", "unigrams = [w for w in unigrams if len(set(w) - set(string.ascii_lowercase)) == 0]\n", "vc_s = pd.Series(unigrams).value_counts()" ] }, { "cell_type": "code", "execution_count": 64, "id": "079fcfd9", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "gov 50682\n", "uscourts 49906\n", "illappct 9447\n", "cafc 5130\n", "house 1341\n", "case 1177\n", "cadc 861\n", "ill 854\n", "the 837\n", "cv 684\n", "hbs 591\n", "council 545\n", "cosflv 536\n", "cons 487\n", "of 451\n", "meeting 437\n", "scotus 430\n", "peters 407\n", "board 405\n", "gvtvnews 400\n", "dtype: int64" ] }, "execution_count": 64, "metadata": {}, "output_type": "execute_result" } ], "source": [ "vc_s.head(20)" ] }, { "cell_type": "code", "execution_count": 66, "id": "3c4ad739", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "400" ] }, "execution_count": 66, "metadata": {}, "output_type": "execute_result" } ], "source": [ "vc_s.loc['gvtvnews']" ] }, { "cell_type": "code", "execution_count": 69, "id": "9d8fc08b", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "audio_document_id Gvtvnews-GrassValleyTelevisionUnderwritingAdve...\n", "identifier Gvtvnews-GrassValleyTelevisionUnderwritingAdve...\n", "text_document_id Gvtvnews-GrassValleyTelevisionUnderwritingAdve...\n", "training_data {'duration_ms': [14860, 14620, 14980, 14800, 1...\n", "Name: 922, dtype: object" ] }, "execution_count": 69, "metadata": {}, "output_type": "execute_result" } ], "source": [ "import re\n", "row = meta_df[meta_df['audio_document_id'].str.contains(r\"gvtvnews\", flags=re.IGNORECASE)].iloc[0]\n", "row" ] }, { "cell_type": "code", "execution_count": 71, "id": "731aa37b", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "{'duration_ms': [14860, 14620, 14980, 14800, 14920, 14560],\n", " 'label': ['focuses on newest technologies to reach the public i pods cell phones podcasts and television it might surprise you that our program for advertising is not',\n", " 'expensive we believe that our success is based on your success so we provide you with proven options that you can streamline to your budget grass valley television produces g v t v news we care is locally on',\n", " 'two programs which combined there are over two hundred times a month on channel eleven and unlimited time johnny unitas underwriting for six months contract is hundred dollars per month which breaks down to fifty cents per management also this shows that if you support your local media',\n", " \"as discovery lifetime fox news m t v and many many more if you underwrite g v t v news and we create the overlay ads for comcast channels in the grass valley area for free that's right it's free and that's a value of a dunder to sixteen\",\n", " \"speak for the future way to reach your market for a limited time we offer to shoot your underwriter mention and eric for one week on our news for free that's right if you act right now we will schedule a time convenient for you and provide\",\n", " 'you with visual proof that we can help you no obligation if you decide not to underwrite our news you still receive a free week of exposure we have put together some samples of some of our work with this d v d and you can see that'],\n", " 'name': ['Gvtvnews-GrassValleyTelevisionUnderwritingAdvertisingInfo682-3/Gvtvnews-GrassValleyTelevisionUnderwritingAdvertisingInfo500_DOT_mp3_00000.flac',\n", " 'Gvtvnews-GrassValleyTelevisionUnderwritingAdvertisingInfo682-3/Gvtvnews-GrassValleyTelevisionUnderwritingAdvertisingInfo500_DOT_mp3_00001.flac',\n", " 'Gvtvnews-GrassValleyTelevisionUnderwritingAdvertisingInfo682-3/Gvtvnews-GrassValleyTelevisionUnderwritingAdvertisingInfo500_DOT_mp3_00002.flac',\n", " 'Gvtvnews-GrassValleyTelevisionUnderwritingAdvertisingInfo682-3/Gvtvnews-GrassValleyTelevisionUnderwritingAdvertisingInfo500_DOT_mp3_00003.flac',\n", " 'Gvtvnews-GrassValleyTelevisionUnderwritingAdvertisingInfo682-3/Gvtvnews-GrassValleyTelevisionUnderwritingAdvertisingInfo500_DOT_mp3_00004.flac',\n", " 'Gvtvnews-GrassValleyTelevisionUnderwritingAdvertisingInfo682-3/Gvtvnews-GrassValleyTelevisionUnderwritingAdvertisingInfo500_DOT_mp3_00005.flac']}" ] }, "execution_count": 71, "metadata": {}, "output_type": "execute_result" } ], "source": [ "row['training_data']" ] }, { "cell_type": "code", "execution_count": null, "id": "a888d19d", "metadata": {}, "outputs": [], "source": [] }, { "cell_type": "code", "execution_count": null, "id": "fa4c68e0", "metadata": {}, "outputs": [], "source": [] } ], "metadata": { "kernelspec": { "display_name": "Python 3 (ipykernel)", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.8.10" } }, "nbformat": 4, "nbformat_minor": 5 }