54 lines
1.5 KiB
Python
54 lines
1.5 KiB
Python
"""Bewertung und Sortierung von Suchtreffern.
|
|
|
|
Der Score ist eine Aehnlichkeitsbewertung (0-100) und ausdruecklich keine
|
|
statistische Wahrscheinlichkeit.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
from typing import TYPE_CHECKING
|
|
|
|
if TYPE_CHECKING: # pragma: no cover
|
|
from .engine import SearchHit
|
|
|
|
|
|
def clamp_score(score: float) -> float:
|
|
return max(0.0, min(100.0, float(score)))
|
|
|
|
|
|
def dedupe_overlapping(hits: list["SearchHit"], gap_tokens: int = 3) -> list["SearchHit"]:
|
|
"""Entfernt Treffer, die denselben Textabschnitt mehrfach abdecken."""
|
|
if not hits:
|
|
return []
|
|
|
|
ordered = sorted(hits, key=lambda hit: (-hit.score, hit.start_token))
|
|
kept: list["SearchHit"] = []
|
|
for hit in ordered:
|
|
overlapping = False
|
|
for existing in kept:
|
|
if existing.audio_id != hit.audio_id:
|
|
continue
|
|
same_region = (
|
|
abs(existing.start_token - hit.start_token) <= gap_tokens
|
|
or (hit.start_token < existing.end_token and existing.start_token < hit.end_token)
|
|
)
|
|
if same_region:
|
|
overlapping = True
|
|
break
|
|
if not overlapping:
|
|
kept.append(hit)
|
|
return kept
|
|
|
|
|
|
def rank_hits(hits: list["SearchHit"], limit: int) -> list["SearchHit"]:
|
|
"""Sortiert Treffer nach Relevanz und begrenzt die Anzahl."""
|
|
ordered = sorted(
|
|
hits,
|
|
key=lambda hit: (
|
|
-hit.score,
|
|
hit.filename.lower(),
|
|
hit.start_time,
|
|
),
|
|
)
|
|
return ordered[: max(0, int(limit))]
|