Langsame Hörhilfen für Ausspracheübungen ergänzen

This commit is contained in:
Eugenius Naumenko
2026-10-14 02:13:07 +00:00
parent 16da8cad1d
commit 83944f791e
7 changed files with 261 additions and 32 deletions
@@ -34,10 +34,13 @@ type audioSpeechService struct {
locale string locale string
voice string voice string
text string text string
mode string
word string
} }
func (s *audioSpeechService) Synthesize(_ context.Context, locale, voice, text string) ([]byte, string, error) { func (s *audioSpeechService) SynthesizeExercise(_ context.Context, locale, voice, text, mode, word string) ([]byte, string, error) {
s.locale, s.voice, s.text = locale, voice, text s.locale, s.voice, s.text = locale, voice, text
s.mode, s.word = mode, word
return []byte("mp3-data"), "audio/mpeg", nil return []byte("mp3-data"), "audio/mpeg", nil
} }
@@ -58,11 +61,50 @@ func TestExerciseAudioUsesStoredSentenceAndLanguageVoice(t *testing.T) {
if recorder.Header().Get("Content-Type") != "audio/mpeg" { if recorder.Header().Get("Content-Type") != "audio/mpeg" {
t.Fatalf("content type = %q", recorder.Header().Get("Content-Type")) t.Fatalf("content type = %q", recorder.Header().Get("Content-Type"))
} }
if lookup.requestedUserID != "usr_current" || speech.locale != "en-US" || speech.voice != "en-US-JennyNeural" || speech.text != "Please pass the water." { if lookup.requestedUserID != "usr_current" || speech.locale != "en-US" || speech.voice != "en-US-JennyNeural" || speech.text != "Please pass the water." || speech.mode != "normal" {
t.Fatalf("unexpected synthesis input: user %q, locale %q, voice %q, text %q", lookup.requestedUserID, speech.locale, speech.voice, speech.text) t.Fatalf("unexpected synthesis input: user %q, locale %q, voice %q, text %q", lookup.requestedUserID, speech.locale, speech.voice, speech.text)
} }
} }
func TestExerciseAudioSlowAndFocusModes(t *testing.T) {
for _, test := range []struct {
name string
url string
mode string
word string
}{
{"slow", "/practice/exercises/ex_1/audio?mode=slow", "slow", ""},
{"focus", "/practice/exercises/ex_1/audio?mode=focus&word=water", "focus", "water"},
{"focus punctuation", "/practice/exercises/ex_1/audio?mode=focus&word=water.", "focus", "water"},
} {
t.Run(test.name, func(t *testing.T) {
speech := &audioSpeechService{}
server := &Server{store: &audioExerciseStore{}, speech: speech}
req := httptest.NewRequest(http.MethodGet, test.url, nil)
req.SetPathValue("id", "ex_1")
req = req.WithContext(context.WithValue(req.Context(), userIDKey, "usr_current"))
recorder := httptest.NewRecorder()
server.exerciseAudio(recorder, req)
if recorder.Code != http.StatusOK || speech.mode != test.mode || speech.word != test.word {
t.Fatalf("status = %d, mode = %q, word = %q", recorder.Code, speech.mode, speech.word)
}
})
}
}
func TestExerciseAudioRejectsWordsOutsideSentence(t *testing.T) {
speech := &audioSpeechService{}
server := &Server{store: &audioExerciseStore{}, speech: speech}
req := httptest.NewRequest(http.MethodGet, "/practice/exercises/ex_1/audio?mode=focus&word=unrelated", nil)
req.SetPathValue("id", "ex_1")
req = req.WithContext(context.WithValue(req.Context(), userIDKey, "usr_current"))
recorder := httptest.NewRecorder()
server.exerciseAudio(recorder, req)
if recorder.Code != http.StatusBadRequest || speech.text != "" {
t.Fatalf("status = %d, synthesis text = %q", recorder.Code, speech.text)
}
}
func TestExerciseAudioRejectsInaccessibleExercise(t *testing.T) { func TestExerciseAudioRejectsInaccessibleExercise(t *testing.T) {
lookup := &audioExerciseStore{lookupErr: store.ErrNotFound} lookup := &audioExerciseStore{lookupErr: store.ErrNotFound}
speech := &audioSpeechService{} speech := &audioSpeechService{}
+35 -1
View File
@@ -9,6 +9,7 @@ import (
"net/http" "net/http"
"strings" "strings"
"time" "time"
"unicode"
"speech-learning-api/internal/config" "speech-learning-api/internal/config"
"speech-learning-api/internal/domain" "speech-learning-api/internal/domain"
@@ -306,12 +307,31 @@ func (s *Server) exerciseAudio(w http.ResponseWriter, r *http.Request) {
writeError(w, http.StatusNotFound, err) writeError(w, http.StatusNotFound, err)
return return
} }
mode := r.URL.Query().Get("mode")
if mode == "" {
mode = "normal"
}
if mode != "normal" && mode != "slow" && mode != "focus" {
writeError(w, http.StatusBadRequest, errors.New("unsupported audio mode"))
return
}
focusWord := r.URL.Query().Get("word")
if mode == "focus" {
focusWord = exerciseFocusWord(exercise.TargetText, focusWord)
if focusWord == "" {
writeError(w, http.StatusBadRequest, errors.New("word must occur in the exercise sentence"))
return
}
} else if focusWord != "" {
writeError(w, http.StatusBadRequest, errors.New("word is only supported in focus mode"))
return
}
language, err := s.store.GetLanguage(r.Context(), exercise.TargetLanguage) language, err := s.store.GetLanguage(r.Context(), exercise.TargetLanguage)
if err != nil { if err != nil {
writeError(w, http.StatusInternalServerError, err) writeError(w, http.StatusInternalServerError, err)
return return
} }
audio, mime, err := s.speech.Synthesize(r.Context(), language.AzureLocale, language.DefaultTTSVoice, exercise.TargetText) audio, mime, err := s.speech.SynthesizeExercise(r.Context(), language.AzureLocale, language.DefaultTTSVoice, exercise.TargetText, mode, focusWord)
if err != nil { if err != nil {
writeError(w, http.StatusBadGateway, err) writeError(w, http.StatusBadGateway, err)
return return
@@ -326,6 +346,20 @@ func (s *Server) exerciseAudio(w http.ResponseWriter, r *http.Request) {
_, _ = w.Write(audio) _, _ = w.Write(audio)
} }
func exerciseFocusWord(sentence, requested string) string {
requested = strings.TrimFunc(strings.TrimSpace(requested), func(r rune) bool { return !unicode.IsLetter(r) && !unicode.IsNumber(r) })
if requested == "" {
return ""
}
for _, token := range strings.Fields(sentence) {
word := strings.TrimFunc(token, func(r rune) bool { return !unicode.IsLetter(r) && !unicode.IsNumber(r) })
if strings.EqualFold(word, requested) {
return word
}
}
return ""
}
func (s *Server) createAttempt(w http.ResponseWriter, r *http.Request) { func (s *Server) createAttempt(w http.ResponseWriter, r *http.Request) {
if err := r.ParseMultipartForm(24 << 20); err != nil { if err := r.ParseMultipartForm(24 << 20); err != nil {
writeError(w, http.StatusBadRequest, err) writeError(w, http.StatusBadRequest, err)
+33 -4
View File
@@ -22,6 +22,7 @@ import (
type SpeechService interface { type SpeechService interface {
AssessPronunciation(ctx context.Context, input PronunciationInput) (domain.PronunciationResult, error) AssessPronunciation(ctx context.Context, input PronunciationInput) (domain.PronunciationResult, error)
Synthesize(ctx context.Context, locale, voice, text string) ([]byte, string, error) Synthesize(ctx context.Context, locale, voice, text string) ([]byte, string, error)
SynthesizeExercise(ctx context.Context, locale, voice, text, mode, focusWord string) ([]byte, string, error)
} }
type PronunciationInput struct { type PronunciationInput struct {
@@ -98,13 +99,41 @@ func (s *AzureSpeechService) Synthesize(ctx context.Context, locale, voice, text
if strings.TrimSpace(text) == "" { if strings.TrimSpace(text) == "" {
return nil, "", errors.New("tts text is required") return nil, "", errors.New("tts text is required")
} }
if s.cfg.AzureMock || s.cfg.AzureSpeechKey == "" || s.cfg.AzureRegion == "" {
return nil, "", nil
}
locale = valueOr(locale, "en-US") locale = valueOr(locale, "en-US")
voice = valueOr(voice, "en-US-JennyNeural") voice = valueOr(voice, "en-US-JennyNeural")
ssml := fmt.Sprintf(`<speak version="1.0" xml:lang="%s"><voice name="%s">%s</voice></speak>`, locale, voice, xmlEscape(text)) ssml := fmt.Sprintf(`<speak version="1.0" xml:lang="%s"><voice name="%s">%s</voice></speak>`, locale, voice, xmlEscape(text))
return s.synthesizeSSML(ctx, ssml)
}
func (s *AzureSpeechService) SynthesizeExercise(ctx context.Context, locale, voice, text, mode, focusWord string) ([]byte, string, error) {
if strings.TrimSpace(text) == "" {
return nil, "", errors.New("tts text is required")
}
if mode == "normal" {
return s.Synthesize(ctx, locale, voice, text)
}
locale = valueOr(locale, "en-US")
voice = valueOr(voice, "en-US-JennyNeural")
var content string
switch mode {
case "slow":
content = fmt.Sprintf(`<prosody rate="0.8">%s</prosody>`, xmlEscape(text))
case "focus":
if strings.TrimSpace(focusWord) == "" {
return nil, "", errors.New("focus word is required")
}
content = fmt.Sprintf(`<prosody rate="0.7">%s</prosody><break time="400ms"/><prosody rate="0.8">%s</prosody>`, xmlEscape(focusWord), xmlEscape(text))
default:
return nil, "", errors.New("unsupported audio mode")
}
ssml := fmt.Sprintf(`<speak version="1.0" xml:lang="%s"><voice name="%s">%s</voice></speak>`, locale, voice, content)
return s.synthesizeSSML(ctx, ssml)
}
func (s *AzureSpeechService) synthesizeSSML(ctx context.Context, ssml string) ([]byte, string, error) {
if s.cfg.AzureMock || s.cfg.AzureSpeechKey == "" || s.cfg.AzureRegion == "" {
return nil, "", nil
}
endpoint := fmt.Sprintf("https://%s.tts.speech.microsoft.com/cognitiveservices/v1", s.cfg.AzureRegion) endpoint := fmt.Sprintf("https://%s.tts.speech.microsoft.com/cognitiveservices/v1", s.cfg.AzureRegion)
req, err := http.NewRequestWithContext(ctx, http.MethodPost, endpoint, strings.NewReader(ssml)) req, err := http.NewRequestWithContext(ctx, http.MethodPost, endpoint, strings.NewReader(ssml))
+67 -1
View File
@@ -1,6 +1,72 @@
package services package services
import "testing" import (
"context"
"encoding/xml"
"io"
"net/http"
"strings"
"testing"
"speech-learning-api/internal/config"
)
type speechRoundTripFunc func(*http.Request) (*http.Response, error)
func (fn speechRoundTripFunc) RoundTrip(req *http.Request) (*http.Response, error) {
return fn(req)
}
func TestSynthesizeExerciseSendsValidSSML(t *testing.T) {
for _, test := range []struct {
mode string
word string
rates []string
}{
{"slow", "", []string{"0.8"}},
{"focus", "chips", []string{"0.7", "0.8"}},
} {
t.Run(test.mode, func(t *testing.T) {
var body string
service := &AzureSpeechService{
cfg: config.Config{AzureSpeechKey: "test-key", AzureRegion: "test-region"},
client: &http.Client{Transport: speechRoundTripFunc(func(req *http.Request) (*http.Response, error) {
payload, err := io.ReadAll(req.Body)
if err != nil {
t.Fatal(err)
}
body = string(payload)
return &http.Response{StatusCode: http.StatusOK, Body: io.NopCloser(strings.NewReader("mp3")), Header: make(http.Header)}, nil
})},
}
_, _, err := service.SynthesizeExercise(context.Background(), "en-US", "en-US-JennyNeural", "Fish & chips are tasty.", test.mode, test.word)
if err != nil {
t.Fatal(err)
}
decoder := xml.NewDecoder(strings.NewReader(body))
for {
_, err := decoder.Token()
if err == io.EOF {
break
}
if err != nil {
t.Fatalf("invalid SSML: %v; body %q", err, body)
}
}
if !strings.Contains(body, "Fish &amp; chips are tasty.") {
t.Fatalf("sentence was not XML-escaped: %q", body)
}
for _, rate := range test.rates {
if !strings.Contains(body, `rate="`+rate+`"`) {
t.Fatalf("missing rate %s: %q", rate, body)
}
}
if test.mode == "focus" && (!strings.Contains(body, `<break time="400ms"/>`) || !strings.Contains(body, `>chips</prosody>`)) {
t.Fatalf("focus clip lacks the word or pause: %q", body)
}
})
}
}
func TestMockPronunciationReturnsWordScores(t *testing.T) { func TestMockPronunciationReturnsWordScores(t *testing.T) {
result := mockPronunciation("Hello, my name is Anna.") result := mockPronunciation("Hello, my name is Anna.")
+8 -2
View File
@@ -16,8 +16,14 @@ export function fetchProgress() {
return request<ProgressSummary>('/progress'); return request<ProgressSummary>('/progress');
} }
export function fetchExerciseAudio(exerciseId: string, signal?: AbortSignal) { export type ExerciseAudioMode = 'normal' | 'slow' | 'focus';
return requestBlob(`/practice/exercises/${encodeURIComponent(exerciseId)}/audio`, { signal });
export function fetchExerciseAudio(exerciseId: string, mode: ExerciseAudioMode = 'normal', word = '', signal?: AbortSignal) {
const params = new URLSearchParams();
if (mode !== 'normal') params.set('mode', mode);
if (mode === 'focus') params.set('word', word);
const query = params.size ? `?${params.toString()}` : '';
return requestBlob(`/practice/exercises/${encodeURIComponent(exerciseId)}/audio${query}`, { signal });
} }
export async function fetchNextLesson(fresh = false, recentExerciseIds: string[] = []) { export async function fetchNextLesson(fresh = false, recentExerciseIds: string[] = []) {
+71 -21
View File
@@ -1,8 +1,8 @@
import { RotateCcw, Send, Volume2 } from 'lucide-react'; import { RotateCcw, Send, Volume2 } from 'lucide-react';
import { useEffect, useRef, useState } from 'react'; import { useEffect, useRef, useState } from 'react';
import { fetchExerciseAudio, fetchNextLesson } from '../api/lessons'; import { fetchExerciseAudio, fetchNextLesson, type ExerciseAudioMode } from '../api/lessons';
import { submitAttempt } from '../api/practice'; import { submitAttempt } from '../api/practice';
import type { AttemptResponse, Exercise, Lesson } from '../api/types'; import type { AttemptResponse, Exercise, Lesson, WordScore } from '../api/types';
import { AudioRecorder } from '../components/AudioRecorder'; import { AudioRecorder } from '../components/AudioRecorder';
import { PhonemeAnalysis } from '../components/PhonemeAnalysis'; import { PhonemeAnalysis } from '../components/PhonemeAnalysis';
import { PronunciationText } from '../components/PronunciationText'; import { PronunciationText } from '../components/PronunciationText';
@@ -21,12 +21,12 @@ export function PracticePage({ initialSelection }: Props) {
const [loading, setLoading] = useState(false); const [loading, setLoading] = useState(false);
const [loadingNext, setLoadingNext] = useState(false); const [loadingNext, setLoadingNext] = useState(false);
const [recording, setRecording] = useState(false); const [recording, setRecording] = useState(false);
const [audioLoading, setAudioLoading] = useState(false); const [audioLoadingKey, setAudioLoadingKey] = useState<string | null>(null);
const [audioError, setAudioError] = useState(''); const [audioError, setAudioError] = useState('');
const [error, setError] = useState(''); const [error, setError] = useState('');
const [notice, setNotice] = useState(''); const [notice, setNotice] = useState('');
const submittingRef = useRef(false); const submittingRef = useRef(false);
const referenceAudioRef = useRef<{ exerciseId: string; url: string; player: HTMLAudioElement } | null>(null); const referenceAudioRef = useRef<{ exerciseId: string; urls: Map<string, string>; player: HTMLAudioElement | null } | null>(null);
const audioRequestRef = useRef<AbortController | null>(null); const audioRequestRef = useRef<AbortController | null>(null);
const audioGenerationRef = useRef(0); const audioGenerationRef = useRef(0);
const recentExerciseIds = useRef<string[]>(initialSelection?.exercise.id ? [initialSelection.exercise.id] : []); const recentExerciseIds = useRef<string[]>(initialSelection?.exercise.id ? [initialSelection.exercise.id] : []);
@@ -49,35 +49,41 @@ export function PracticePage({ initialSelection }: Props) {
audioGenerationRef.current += 1; audioGenerationRef.current += 1;
audioRequestRef.current?.abort(); audioRequestRef.current?.abort();
audioRequestRef.current = null; audioRequestRef.current = null;
referenceAudioRef.current?.player.pause(); referenceAudioRef.current?.player?.pause();
if (referenceAudioRef.current) URL.revokeObjectURL(referenceAudioRef.current.url); referenceAudioRef.current?.urls.forEach((url) => URL.revokeObjectURL(url));
referenceAudioRef.current = null; referenceAudioRef.current = null;
}, [exercise?.id]); }, [exercise?.id]);
async function playReference() { async function playReference(mode: ExerciseAudioMode, word = '') {
if (!exercise || audioLoading || audioRequestRef.current || recording) return; if (!exercise || audioLoadingKey || audioRequestRef.current || recording) return;
setAudioError(''); setAudioError('');
const exerciseId = exercise.id; const exerciseId = exercise.id;
const audioKey = `${mode}:${word.toLocaleLowerCase()}`;
const generation = audioGenerationRef.current; const generation = audioGenerationRef.current;
let controller: AbortController | null = null; let controller: AbortController | null = null;
try { try {
if (!referenceAudioRef.current || referenceAudioRef.current.exerciseId !== exerciseId) { if (!referenceAudioRef.current || referenceAudioRef.current.exerciseId !== exerciseId) {
setAudioLoading(true); referenceAudioRef.current = { exerciseId, urls: new Map(), player: null };
}
let url = referenceAudioRef.current.urls.get(audioKey);
if (!url) {
setAudioLoadingKey(audioKey);
controller = new AbortController(); controller = new AbortController();
audioRequestRef.current = controller; audioRequestRef.current = controller;
const blob = await fetchExerciseAudio(exerciseId, controller.signal); const blob = await fetchExerciseAudio(exerciseId, mode, word, controller.signal);
if (controller.signal.aborted || audioGenerationRef.current !== generation) return; if (controller.signal.aborted || audioGenerationRef.current !== generation) return;
const url = URL.createObjectURL(blob); url = URL.createObjectURL(blob);
referenceAudioRef.current = { exerciseId, url, player: new Audio(url) }; referenceAudioRef.current.urls.set(audioKey, url);
} }
const player = referenceAudioRef.current.player; referenceAudioRef.current.player?.pause();
player.currentTime = 0; const player = new Audio(url);
referenceAudioRef.current.player = player;
await player.play(); await player.play();
} catch { } catch {
if (audioGenerationRef.current !== generation || controller?.signal.aborted) return; if (audioGenerationRef.current !== generation || controller?.signal.aborted) return;
setAudioError('Satz konnte nicht abgespielt werden. Bitte erneut versuchen.'); setAudioError('Satz konnte nicht abgespielt werden. Bitte erneut versuchen.');
} finally { } finally {
if (audioGenerationRef.current === generation) setAudioLoading(false); if (audioGenerationRef.current === generation) setAudioLoadingKey(null);
if (audioRequestRef.current === controller) audioRequestRef.current = null; if (audioRequestRef.current === controller) audioRequestRef.current = null;
} }
} }
@@ -87,8 +93,8 @@ export function PracticePage({ initialSelection }: Props) {
audioGenerationRef.current += 1; audioGenerationRef.current += 1;
audioRequestRef.current?.abort(); audioRequestRef.current?.abort();
audioRequestRef.current = null; audioRequestRef.current = null;
referenceAudioRef.current?.player.pause(); referenceAudioRef.current?.player?.pause();
setAudioLoading(false); setAudioLoadingKey(null);
setLoadingNext(true); setLoadingNext(true);
setError(''); setError('');
setNotice(''); setNotice('');
@@ -137,6 +143,8 @@ export function PracticePage({ initialSelection }: Props) {
); );
} }
const focusWords = result ? practiceFocusWords(exercise.target_text, result.attempt.word_scores) : [];
return ( return (
<div className="view-stack"> <div className="view-stack">
<header className="view-header"> <header className="view-header">
@@ -155,9 +163,26 @@ export function PracticePage({ initialSelection }: Props) {
<section className="panel practice-panel"> <section className="panel practice-panel">
<p className="muted">{exercise.prompt_text}</p> <p className="muted">{exercise.prompt_text}</p>
<PronunciationText targetText={exercise.target_text} words={result?.attempt.word_scores} /> <PronunciationText targetText={exercise.target_text} words={result?.attempt.word_scores} />
<button className="ghost-button" type="button" disabled={loadingNext || audioLoading || recording} onClick={() => void playReference()}> <div className="action-row">
<Volume2 size={18} /> {audioLoading ? 'Satz wird geladen...' : 'Satz anhören'} <button className="ghost-button" type="button" disabled={loadingNext || audioLoadingKey !== null || recording} onClick={() => void playReference('normal')}>
<Volume2 size={18} /> {audioLoadingKey === 'normal:' ? 'Satz wird geladen...' : 'Satz anhören'}
</button> </button>
<button className="ghost-button" type="button" disabled={loadingNext || audioLoadingKey !== null || recording} onClick={() => void playReference('slow')}>
<Volume2 size={18} /> {audioLoadingKey === 'slow:' ? 'Satz wird geladen...' : 'Langsam anhören'}
</button>
</div>
{focusWords.length > 0 && (
<div className="recording-preview">
<p className="muted">Diese Wörter kannst du gezielt anhören:</p>
<div className="action-row">
{focusWords.map((word) => (
<button className="ghost-button" type="button" key={word} disabled={loadingNext || audioLoadingKey !== null || recording} onClick={() => void playReference('focus', word)}>
<Volume2 size={18} /> {audioLoadingKey === `focus:${word.toLocaleLowerCase()}` ? 'Wort wird geladen...' : `${word} langsam üben`}
</button>
))}
</div>
</div>
)}
{audioError && <p className="form-error">{audioError}</p>} {audioError && <p className="form-error">{audioError}</p>}
<div className="action-row"> <div className="action-row">
<AudioRecorder <AudioRecorder
@@ -167,8 +192,8 @@ export function PracticePage({ initialSelection }: Props) {
audioGenerationRef.current += 1; audioGenerationRef.current += 1;
audioRequestRef.current?.abort(); audioRequestRef.current?.abort();
audioRequestRef.current = null; audioRequestRef.current = null;
referenceAudioRef.current?.player.pause(); referenceAudioRef.current?.player?.pause();
setAudioLoading(false); setAudioLoadingKey(null);
setAudio(null); setAudio(null);
setResult(null); setResult(null);
setError(''); setError('');
@@ -215,3 +240,28 @@ export function PracticePage({ initialSelection }: Props) {
</div> </div>
); );
} }
function practiceFocusWords(sentence: string, scores: WordScore[]): string[] {
const sentenceWords = sentence.split(/\s+/).map(cleanWord);
const seen = new Set<string>();
return scores
.map((score, index) => ({
word: sentenceWords.find((word) => word.toLocaleLowerCase() === cleanWord(score.word).toLocaleLowerCase()) ?? '',
score: Math.min(score.accuracy, ...(score.phoneme_scores ?? []).map((phoneme) => phoneme.accuracy)),
index
}))
.filter((item) => item.word && item.score < 75)
.sort((a, b) => a.score - b.score || a.index - b.index)
.filter((item) => {
const key = item.word.toLocaleLowerCase();
if (seen.has(key)) return false;
seen.add(key);
return true;
})
.slice(0, 3)
.map((item) => item.word);
}
function cleanWord(value: string): string {
return value.replace(/^[^\p{L}\p{N}]+|[^\p{L}\p{N}]+$/gu, '');
}
+2
View File
@@ -32,6 +32,8 @@ Authorization: Bearer <token>
- `GET /practice/exercises/{id}/audio` (Bearer-Token erforderlich; liefert den Satz der zugänglichen Übung als `audio/mpeg`) - `GET /practice/exercises/{id}/audio` (Bearer-Token erforderlich; liefert den Satz der zugänglichen Übung als `audio/mpeg`)
- `POST /practice/attempts` - `POST /practice/attempts`
`GET /practice/exercises/{id}/audio` unterstützt `mode=normal` (Standard), `mode=slow` und `mode=focus&word=water`. Im Fokusmodus spricht Azure zuerst das Wort, dann nach einer Pause den Satz langsam. Das Wort muss im gespeicherten Übungssatz vorkommen.
`POST /practice/attempts` expects multipart form data: `POST /practice/attempts` expects multipart form data:
- `exercise_id` - `exercise_id`