Exemplo n.º 1
0
def get_mean_exposures_per_pivot():
    "Returns the number of exposures each pivot received."
    cursor = connection.cursor()
    cursor.execute("""
        SELECT pivot, pivot_type, COUNT(*) as n_exposures
        FROM drill_question
        GROUP BY CONCAT(pivot, "|", pivot_type)
    """)
    data = cursor.fetchall()
    word_c = []
    kanji_c = []
    combined_c = []
    kanji_inc_dist = FreqDist()
    for pivot, pivot_type, n_exposures in data:
        combined_c.append(n_exposures)

        if pivot_type == 'k':
            kanji_c.append(n_exposures)
            kanji_inc_dist.inc(pivot, n_exposures)

        elif pivot_type == 'w':
            word_c.append(n_exposures)
            for kanji in scripts.unique_kanji(pivot):
                kanji_inc_dist.inc(kanji, n_exposures)

        else:
            raise ValueError('unknown pivot type: %s' % pivot_type)

    return [
        ('Words', mean(word_c)),
        ('Kanji', mean(kanji_c)),
        ('Combined', mean(combined_c)),
        ('Kanji combined', mean(kanji_inc_dist.values())),
    ]
Exemplo n.º 2
0
def get_mean_exposures_per_pivot():
    "Returns the number of exposures each pivot received."
    cursor = connection.cursor()
    cursor.execute("""
        SELECT pivot, pivot_type, COUNT(*) as n_exposures
        FROM drill_question
        GROUP BY CONCAT(pivot, "|", pivot_type)
    """)
    data = cursor.fetchall()
    word_c = []
    kanji_c = []
    combined_c = []
    kanji_inc_dist = FreqDist()
    for pivot, pivot_type, n_exposures in data:
        combined_c.append(n_exposures)

        if pivot_type == 'k':
            kanji_c.append(n_exposures)
            kanji_inc_dist.inc(pivot, n_exposures)

        elif pivot_type == 'w':
            word_c.append(n_exposures)
            for kanji in scripts.unique_kanji(pivot):
                kanji_inc_dist.inc(kanji, n_exposures)

        else:
            raise ValueError('unknown pivot type: %s' % pivot_type)

    return [
            ('Words', mean(word_c)),
            ('Kanji', mean(kanji_c)),
            ('Combined', mean(combined_c)),
            ('Kanji combined', mean(kanji_inc_dist.values())),
        ]
Exemplo n.º 3
0
def get_test_size_stats():
    """
    Fetches the distribution of test sizes chosen by users, and the completion
    statistics for each reported size.
    """
    cursor = connection.cursor()
    cursor.execute("""
        SELECT n_items, COUNT(*), AVG(is_finished) FROM (
            SELECT 
                ts.id AS test_set_id,
                COUNT(*) AS n_items,
                (ts.end_time IS NOT NULL) AS is_finished
            FROM drill_testset_questions AS tsq
            INNER JOIN drill_testset AS ts
            ON tsq.testset_id = ts.id
            GROUP BY ts.id
        ) as tmp
        GROUP BY n_items
        ORDER BY n_items ASC
    """)
    data = cursor.fetchall()
    dist = FreqDist()
    completion_rates = {}
    for n_items, n_tests, completion_rate in data:
        dist.inc(n_items, n_tests)
        completion_rates[n_items] = completion_rate
    results = []
    for sample in sorted(dist.samples()):
        results.append((sample, dist.freq(sample), completion_rates[sample]))
    return results
Exemplo n.º 4
0
def get_test_size_stats():
    """
    Fetches the distribution of test sizes chosen by users, and the completion
    statistics for each reported size.
    """
    cursor = connection.cursor()
    cursor.execute("""
        SELECT n_items, COUNT(*), AVG(is_finished) FROM (
            SELECT 
                ts.id AS test_set_id,
                COUNT(*) AS n_items,
                (ts.end_time IS NOT NULL) AS is_finished
            FROM drill_testset_questions AS tsq
            INNER JOIN drill_testset AS ts
            ON tsq.testset_id = ts.id
            GROUP BY ts.id
        ) as tmp
        GROUP BY n_items
        ORDER BY n_items ASC
    """)
    data = cursor.fetchall()
    dist = FreqDist()
    completion_rates = {}
    for n_items, n_tests, completion_rate in data:
        dist.inc(n_items, n_tests)
        completion_rates[n_items] = completion_rate
    results = []
    for sample in sorted(dist.samples()):
        results.append((sample, dist.freq(sample), completion_rates[sample]))
    return results
Exemplo n.º 5
0
def get_accuracy_by_pivot_type():
    cursor = connection.cursor()
    cursor.execute("""
        SELECT
            question.pivot,
            SUM(chosen_option.is_correct) as n_correct,
            COUNT(*) as n_responses
        FROM (
            SELECT mco.question_id, mco.is_correct
            FROM drill_multiplechoiceresponse AS mcr
            INNER JOIN drill_multiplechoiceoption AS mco
            ON mcr.option_id = mco.id
        ) as chosen_option
        INNER JOIN drill_question AS question
        ON chosen_option.question_id = question.id
        WHERE question.pivot_type = "w"
        GROUP BY question.pivot
    """)
    raw_data = cursor.fetchall()
    counts = {
        'Hiragana': FreqDist(),
        'Katakana': FreqDist(),
        'Kanji': FreqDist()
    }
    complex_scripts = set([scripts.Script.Kanji, scripts.Script.Unknown])
    only_katakana = set([scripts.Script.Katakana])
    for word, n_correct, n_responses in raw_data:
        scripts_found = scripts.script_types(word)
        if scripts_found.intersection(complex_scripts):
            dist = counts['Kanji']
        elif scripts_found.intersection(only_katakana):
            dist = counts['Katakana']
        else:
            dist = counts['Hiragana']

        dist.inc(True, int(n_correct))
        dist.inc(False, int(n_responses - n_correct))

    keys = ('Hiragana', 'Katakana', 'Kanji')

    data = [(key, counts[key].freq(True)) for key in keys]

    average = FreqDist()
    for key in keys:
        average.inc(True, counts[key][True])
        average.inc(False, counts[key][False])

    data.append(('Average', average.freq(True)))

    return data
Exemplo n.º 6
0
def get_accuracy_by_pivot_type():
    cursor = connection.cursor()
    cursor.execute("""
        SELECT
            question.pivot,
            SUM(chosen_option.is_correct) as n_correct,
            COUNT(*) as n_responses
        FROM (
            SELECT mco.question_id, mco.is_correct
            FROM drill_multiplechoiceresponse AS mcr
            INNER JOIN drill_multiplechoiceoption AS mco
            ON mcr.option_id = mco.id
        ) as chosen_option
        INNER JOIN drill_question AS question
        ON chosen_option.question_id = question.id
        WHERE question.pivot_type = "w"
        GROUP BY question.pivot
    """)
    raw_data = cursor.fetchall()
    counts = {'Hiragana': FreqDist(), 'Katakana': FreqDist(), 'Kanji':
        FreqDist()}
    complex_scripts = set([scripts.Script.Kanji, scripts.Script.Unknown])
    only_katakana = set([scripts.Script.Katakana])
    for word, n_correct, n_responses in raw_data:
        scripts_found = scripts.script_types(word)
        if scripts_found.intersection(complex_scripts):
            dist = counts['Kanji']
        elif scripts_found.intersection(only_katakana):
            dist = counts['Katakana']
        else:
            dist = counts['Hiragana']

        dist.inc(True, int(n_correct))
        dist.inc(False, int(n_responses - n_correct))

    keys = ('Hiragana', 'Katakana', 'Kanji')

    data = [(key, counts[key].freq(True)) for key in keys]

    average = FreqDist()
    for key in keys:
        average.inc(True, counts[key][True])
        average.inc(False, counts[key][False])

    data.append(('Average', average.freq(True)))

    return data
Exemplo n.º 7
0
def get_pivot_response_stats(pivot_id, pivot_type):
    """
    Given a particular pivot, generate a distribution of observed erroneous 
    responses for each type of plugin.
    """
    cursor = connection.cursor()

    cursor.execute("""
        SELECT plugin_option.plugin_id, plugin_option.value
        FROM drill_multiplechoiceresponse AS mcr
        INNER JOIN (
            SELECT pivot_qn.plugin_id, mco.id AS option_id, mco.value
            FROM (
                SELECT id, question_plugin_id AS plugin_id
                FROM drill_question
                WHERE pivot_type = "%(pivot_type)s"
                    AND pivot_id = %(pivot_id)d
            ) AS pivot_qn
            INNER JOIN drill_multiplechoiceoption AS mco
            ON mco.question_id = pivot_qn.id
        ) AS plugin_option
        ON plugin_option.option_id = mcr.option_id
    """ % {
        'pivot_type': pivot_type,
        'pivot_id': pivot_id
    })
    rows = cursor.fetchall()
    dist_map = {}
    plugin_ids_used = set(plugin_id for (plugin_id, error_value) in rows)
    for plugin_id in plugin_ids_used:
        dist_map[plugin_id] = FreqDist()

    for plugin_id, error_value in rows:
        dist_map[plugin_id].inc(error_value)

    plugin_map = drill_models.QuestionPlugin.objects.in_bulk(dist_map.keys())

    results = [(plugin_map[plugin_id].name, dist) \
            for (plugin_id, dist) in dist_map.iteritems()]
    combined_dist = FreqDist()
    for name, dist in results:
        combined_dist.inc(name, dist.N())
    results[0:0] = [('By plugin type', combined_dist)]

    return results
Exemplo n.º 8
0
def get_pivot_response_stats(pivot_id, pivot_type):
    """
    Given a particular pivot, generate a distribution of observed erroneous 
    responses for each type of plugin.
    """
    cursor = connection.cursor()
    
    cursor.execute("""
        SELECT plugin_option.plugin_id, plugin_option.value
        FROM drill_multiplechoiceresponse AS mcr
        INNER JOIN (
            SELECT pivot_qn.plugin_id, mco.id AS option_id, mco.value
            FROM (
                SELECT id, question_plugin_id AS plugin_id
                FROM drill_question
                WHERE pivot_type = "%(pivot_type)s"
                    AND pivot_id = %(pivot_id)d
            ) AS pivot_qn
            INNER JOIN drill_multiplechoiceoption AS mco
            ON mco.question_id = pivot_qn.id
        ) AS plugin_option
        ON plugin_option.option_id = mcr.option_id
    """ % {'pivot_type': pivot_type, 'pivot_id': pivot_id})
    rows = cursor.fetchall()
    dist_map = {}
    plugin_ids_used = set(plugin_id for (plugin_id, error_value) in rows)
    for plugin_id in plugin_ids_used:
        dist_map[plugin_id] = FreqDist()
    
    for plugin_id, error_value in rows:
        dist_map[plugin_id].inc(error_value)
    
    plugin_map = drill_models.QuestionPlugin.objects.in_bulk(dist_map.keys())
    
    results = [(plugin_map[plugin_id].name, dist) \
            for (plugin_id, dist) in dist_map.iteritems()]
    combined_dist = FreqDist()
    for name, dist in results:
        combined_dist.inc(name, dist.N())
    results[0:0] = [('By plugin type', combined_dist)]
    
    return results
Exemplo n.º 9
0
def get_language_data(name):
    "Fetches information about user first and second languages."

    fields_needed = ['user_id']
    if name in ['first', 'combined']:
        fields_needed.append('first_language')
    elif name in ['second', 'combined']:
        fields_needed.append('second_languages')
    else:
        assert name == 'combined'

    valid_users = set(u.id for u in User.objects.all()
                      if u.testset_set.exclude(end_time=None).count() > 0)

    profiles = UserProfile.objects.filter(user__id__in=valid_users).values(
        *fields_needed)

    dist = FreqDist()
    for profile in profiles:
        if 'first_language' in profile:
            lang = profile['first_language'].title()
            dist.inc(lang)
        if 'second_languages' in profile:
            if profile['second_languages']:
                for lang in profile['second_languages'].split(','):
                    lang = lang.strip().title()
                    if lang == 'Japanese':
                        continue
                    dist.inc(lang)
            elif name != 'lang_combined':
                dist.inc('None')

    return dist
Exemplo n.º 10
0
def get_language_data(name):
    "Fetches information about user first and second languages."

    fields_needed = ['user_id']
    if name in ['first', 'combined']:
        fields_needed.append('first_language')
    elif name in ['second', 'combined']:
        fields_needed.append('second_languages')
    else:
        assert name == 'combined'

    valid_users = set(u.id for u in User.objects.all() if u.testset_set.exclude(
            end_time=None).count() > 0)
        
    profiles = UserProfile.objects.filter(user__id__in=valid_users).values(
            *fields_needed)
    
    dist = FreqDist()
    for profile in profiles:
        if 'first_language' in profile:
            lang = profile['first_language'].title()
            dist.inc(lang)
        if 'second_languages' in profile:
            if profile['second_languages']:
                for lang in profile['second_languages'].split(','):
                    lang = lang.strip().title()
                    if lang == 'Japanese':
                        continue
                    dist.inc(lang)
            elif name != 'lang_combined':
                dist.inc('None')

    return dist