{% extends "app/base_generic.html" %}
{% load static %}

{% block title %}<title>TTS Datasets for African languages</title>{% endblock %}
{% block content %}
    <div class="page-content">
        <p>The AfricanVoices corpus is a speech corpus containing datasets of aligned sentences and audio for
            11 languages. We have uploaded data for {{ num_languages }} different languages in this website so far.
            We obtain the datasets in three ways:
        <ul>
            <li>Create and record</li>
            <li> Align audio books from sources like <a href="https://open.bible/resources/"
                                                        target="_blank">Open.Bible </a>, <a
                    href="https://github.com/getalp/ALFFA_PUBLIC" target="_blank">ALLFA project</a>, <a
                    href="http://www.llsti.org/" target="_blank"> LLSTI </a></li>
            <li> Obtain single speaker audio-text pairs from sources like <a href="https://commonvoice.mozilla.org/"
                                                                             target="_blank"> Mozilla CommonVoice</a>
            </li>
        </ul>

        <section id="sect-datasets">
            <h2 class="h2-primary">Datasets</h2>
            <div class="table-responsive">

                <table class="table table-striped ">
                    <tr>
                        <th scope="col"> Data_id</th>
                        <th scope="col"> Lang code</th>
                        <th scope="col"> Language</th>
                        <th scope="col"> Source</th>
                        <th scope="col"> Speaker</th>
                        <th scope="col"> No. of sentences</th>
                        <th scope="col"> Hrs</th>
                        <th scope="col"> MCD*</th>
                        <th scope="col"> Quality</th>

                        {#                    <th> rfs #utt</th>#}
                        {#                    <th> rfs #mcd </th>#}
                        <th> Download</th>
                    </tr>
                    {% for dataset in datasets %}
                        <tr>
                            <td>{{ dataset.data_id }}</td>
                            <td>{{ dataset.lang.lang_code_639_2 }}</td>
                            {% if not dataset.lang.lang_wikipedia_url %}
                                <td>{{ dataset.lang.lang_name }}  </td>
                            {% else %}
                                <td>
                                    <a href="{% url 'language' dataset.lang.lang_code_639_2 %}"
                                       target="_blank">{{ dataset.lang.lang_name }}</a>
                                </td>
                            {% endif %}
                            <td>{{ dataset.source }}</td>
                            <td> {{ dataset.speaker_gender }}</td>
                            <td>{{ dataset.pass1_utt }}</td>
                            <td>{{ dataset.duration|floatformat:2 }}</td>
                            <td>{{ dataset.pass1_mcd|floatformat:2 }}</td>
                            {% if dataset.pass1_mcd <= 6 %}
                                <td> Good</td>
                            {% elif dataset.pass1_mcd <= 7 %}
                                <td> Okay</td>
                            {% elif dataset.pass1_mcd <= 8 %}
                                <td> Bad</td>
                            {% else %}
                                <td>}Something is wrong</td>
                            {% endif %}
                            {% if not dataset.data_location %}
                                <td> **Unavailable</td>
                            {% else %}
                                <td><a href="{% static 'app/' %}{{ dataset.data_location }}" download target=_blank>Available</a>
                                </td>
                            {% endif %}
                        </tr>
                    {% endfor %}
                </table>
            </div>

        </section>

    </div>

    * "MCD is a distortion measure, comparing synthesized examples with originals. Smaller is better. For TTS, less than
    5 is probably good, less than 6 is probably fine, greater than 6 is possibly bad (but still statistically useful),
    greater than 7 probably indicates something is wrong. For alignment, MCDs seem to be about 1 larger than for TTS
    (TTS only uses the best examples, and uses a much more complex prediction model)." <a href="http://festvox.org/cmu_wilderness/">Alan W Black</a>
    <br>
    For data from Open.Bible and Bible.is, the MCD reported is for alignment while for the rest the MCD reported is for TTS.

    <br>
    <br>

    ** Suba data is unavailable because the license doesn't allow us to redistribute. You can download it from the
    <a href="https://www.faithcomesbyhearing.com/audio-bible-resources/mp3-downloads" target=_blank>Faith Comes by
        Hearing </a> website for your personal use.
    <br>
    <br>
{% endblock %}