add test module

2026-08-16 15:39:52 +08:00
parent d0310620fc
commit 6e4d93cea6
46 changed files with 3880 additions and 206 deletions
--- a/sentence_api/assessment.py
+++ b/sentence_api/assessment.py
@@ -0,0 +1,107 @@
+import uuid
+from pathlib import Path
+from typing import Optional
+
+from .audio_metrics import analyze_audio
+from .models import AssessmentResult, SentenceBoundaryDocument, TextSubstitution
+from .repository import VideoRepository
+from .scoring import SCORING_VERSION, score_attempt
+from .transcription import Transcriber
+
+
+class AssessmentService:
+    def __init__(
+        self,
+        repository: VideoRepository,
+        transcriber: Transcriber,
+        pass_score: float,
+    ):
+        self.repository = repository
+        self.transcriber = transcriber
+        self.pass_score = pass_score
+
+    def assess(
+        self,
+        *,
+        document: SentenceBoundaryDocument,
+        sentence_index: int,
+        audio_path: Path,
+        language: Optional[str] = None,
+        retained_audio_filename: Optional[str] = None,
+    ) -> AssessmentResult:
+        if not self.transcriber.available:
+            raise RuntimeError("MOSS transcription is not configured on this server.")
+        sentence = next(
+            (item for item in document.sentences if item.index == sentence_index),
+            None,
+        )
+        if sentence is None:
+            raise LookupError("Sentence index was not found.")
+        if not sentence.text or not sentence.text.strip():
+            raise ValueError("The sentence has no reviewed reference text yet.")
+
+        student_metrics = analyze_audio(audio_path)
+        transcript = self.transcriber.transcribe(
+            audio_path,
+            language or sentence.language,
+        )
+        recognized_text = transcript.text.strip()
+        if not recognized_text:
+            raise ValueError("MOSS did not recognize any speech in the recording.")
+
+        reference_duration_ms = sentence.end_ms - sentence.start_ms
+        reference_speech_duration_ms = (
+            sentence.reference_speech_duration_ms or reference_duration_ms
+        )
+        score = score_attempt(
+            reference_text=sentence.text,
+            recognized_text=recognized_text,
+            reference_speech_duration_ms=reference_speech_duration_ms,
+            student_metrics=student_metrics,
+        )
+        attempt_id = uuid.uuid4().hex
+        result = AssessmentResult(
+            attempt_id=attempt_id,
+            scoring_version=SCORING_VERSION,
+            overall_score=score.overall_score,
+            passed=score.overall_score >= self.pass_score,
+            pass_score=self.pass_score,
+            content_score=score.content_score,
+            completeness_score=score.completeness_score,
+            fluency_score=score.fluency_score,
+            pronunciation_score=None,
+            prosody_score=None,
+            duration_score=score.duration_score,
+            pause_score=score.pause_score,
+            speech_rate_score=score.speech_rate_score,
+            reference_text=sentence.text,
+            recognized_text=recognized_text,
+            reference_duration_ms=reference_duration_ms,
+            reference_speech_duration_ms=reference_speech_duration_ms,
+            student_recording_duration_ms=student_metrics.recording_duration_ms,
+            student_speech_duration_ms=student_metrics.speech_duration_ms,
+            duration_ratio=score.duration_ratio,
+            missing_tokens=score.missing_tokens,
+            extra_tokens=score.extra_tokens,
+            substitutions=[
+                TextSubstitution(expected=expected, actual=actual)
+                for expected, actual in score.substitutions
+            ],
+            feedback=score.feedback,
+            details={
+                "score_kind": "朗读内容与流畅度匹配分",
+                "content_weight": "0.80",
+                "fluency_weight": "0.20",
+                "duration_weight_within_fluency": "0.35",
+                "phoneme_scoring": "not_enabled",
+            },
+        )
+        if self.repository.get_video(document.video_hash) is not None:
+            self.repository.record_attempt(
+                attempt_id=attempt_id,
+                video_hash=document.video_hash,
+                sentence_index=sentence_index,
+                result=result.model_dump(mode="json"),
+                audio_filename=retained_audio_filename,
+            )
+        return result