Skip to content

Commit

Permalink
Fixed word editing edge cases
Browse files Browse the repository at this point in the history
  • Loading branch information
@beveradb
beveradb committed Jan 27, 2025
1 parent 1951c02 commit b72928a
Show file tree
Hide file tree
Showing 2 changed files with 77 additions and 73 deletions.
111 changes: 53 additions & 58 deletions lyrics-analyzer/src/components/LyricsAnalyzer.tsx
View file
Original file line number Diff line number Diff line change
Expand Up @@ -106,7 +106,7 @@ export default function LyricsAnalyzer({ data: initialData, onFileLoad, apiClien
type: 'gap' as const,
data: {
...info.gap,
position: info.gap.transcription_position,
position: info.gap.transcription_position + (info.wordIndex - info.gap.transcription_position),
word: info.gap.words[info.wordIndex - info.gap.transcription_position]
}
}
Expand All @@ -122,33 +122,31 @@ export default function LyricsAnalyzer({ data: initialData, onFileLoad, apiClien
console.log('Position:', position)
console.log('Updated words:', updatedWords)

// Update manual corrections
setManualCorrections(prev => {
const newCorrections = new Map(prev)
newCorrections.set(position, updatedWords)
return newCorrections
})

// Create a deep clone of the data
const newData = JSON.parse(JSON.stringify(data))

// Find and update the gap sequence
// Find the gap that contains this position
const gapIndex = newData.gap_sequences.findIndex(
(gap: GapSequence) => gap.transcription_position === position
(gap: GapSequence) =>
position >= gap.transcription_position &&
position < gap.transcription_position + gap.words.length
)

if (gapIndex !== -1) {
const originalGap = newData.gap_sequences[gapIndex]
console.log('Found gap at index:', gapIndex)
console.log('Original gap:', {
text: originalGap.text,
words: originalGap.words,
transcription_position: originalGap.transcription_position
const wordIndexInGap = position - originalGap.transcription_position
console.log('Found gap at index:', gapIndex, 'word index in gap:', wordIndexInGap)

// Update manual corrections
setManualCorrections(prev => {
const newCorrections = new Map(prev)
newCorrections.set(position, updatedWords)
return newCorrections
})

// Create a new correction
const newCorrection: WordCorrection = {
original_word: originalGap.text,
original_word: originalGap.words[wordIndexInGap],
corrected_word: updatedWords.join(' '),
segment_index: 0,
original_position: position,
Expand All @@ -161,77 +159,74 @@ export default function LyricsAnalyzer({ data: initialData, onFileLoad, apiClien
reference_positions: {}
}

// Find the corresponding segment first to get timing information
const segmentIndex = newData.corrected_segments.findIndex((segment: LyricsSegment) => {
// Calculate total words before this segment
let totalWords = 0
for (let i = 0; i < newData.corrected_segments.indexOf(segment); i++) {
totalWords += newData.corrected_segments[i].words.length
// Find the corresponding segment by counting words
let currentPosition = 0
let segmentIndex = -1
let wordIndex = -1

for (let i = 0; i < newData.corrected_segments.length; i++) {
const segment = newData.corrected_segments[i]
if (position >= currentPosition && position < currentPosition + segment.words.length) {
segmentIndex = i
wordIndex = position - currentPosition
break
}
currentPosition += segment.words.length
}

// Check if this segment contains our target position
const segmentLength = segment.words.length
return totalWords <= originalGap.transcription_position &&
(totalWords + segmentLength) > originalGap.transcription_position
console.log('Segment search:', {
position,
segmentIndex,
wordIndex,
totalSegments: newData.corrected_segments.length
})

if (segmentIndex !== -1) {
if (segmentIndex !== -1 && wordIndex !== -1) {
const segment = newData.corrected_segments[segmentIndex]
const timingWord = segment.words[wordIndex]

console.log('Found matching segment:', {
text: segment.text,
totalWords: newData.corrected_segments
.slice(0, segmentIndex)
.reduce((sum: number, seg: LyricsSegment) => sum + seg.words.length, 0)
wordCount: segment.words.length,
wordIndex,
word: timingWord?.text
})

// Calculate the word index within the segment
let wordsBefore = 0
for (let i = 0; i < segmentIndex; i++) {
wordsBefore += newData.corrected_segments[i].words.length
if (!timingWord) {
console.error('Could not find timing word in segment')
console.groupEnd()
return
}
const wordIndex = originalGap.transcription_position - wordsBefore
const timingWord = segment.words[wordIndex]

console.log('Found word in segment:', {
index: wordIndex,
word: timingWord.text,
timing: {
start: timingWord.start_time,
end: timingWord.end_time
}
})

// Update gap sequence with timing from segment
// Update gap sequence
const newWords = [...originalGap.words]
newWords[wordIndexInGap] = updatedWords[0]
newData.gap_sequences[gapIndex] = {
...originalGap,
words: updatedWords.map(word => ({
text: word,
start_time: timingWord.start_time,
end_time: timingWord.end_time
})),
text: updatedWords.join(' '),
words: newWords,
text: newWords.join(' '),
corrections: originalGap.corrections
.filter((c: WordCorrection) => c.source !== 'manual')
.concat([newCorrection])
}

// Now update the segment
const newWords = [...segment.words]
newWords[wordIndex] = {
// Update segment
const newSegmentWords = [...segment.words]
newSegmentWords[wordIndex] = {
...timingWord,
text: updatedWords[0],
confidence: 1.0
}

newData.corrected_segments[segmentIndex] = {
...segment,
words: newWords,
text: newWords.map(word => word.text).join(' ')
words: newSegmentWords,
text: newSegmentWords.map(word => word.text).join(' ')
}

console.log('Updated both gap and segment')
} else {
console.log('No matching segment found')
console.error('Could not find matching segment for position:', position)
}
}

Expand Down
39 changes: 24 additions & 15 deletions lyrics_transcriber/core/controller.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -291,40 +291,49 @@ def correct_lyrics(self) -> None:
# Add human review step
if self.output_config.enable_review:
from ..review import start_review_server
import re
import json
from copy import deepcopy

self.logger.info("Starting human review process")

# Get auto-corrected data as JSON string
auto_corrected_json = json.dumps(self.results.transcription_corrected.to_dict(), indent=4).splitlines()
def normalize_data(data_dict):
"""Normalize numeric values in the data structure before JSON conversion."""
if isinstance(data_dict, dict):
return {k: normalize_data(v) for k, v in data_dict.items()}
elif isinstance(data_dict, list):
return [normalize_data(item) for item in data_dict]
elif isinstance(data_dict, float):
# Convert whole number floats to integers
if data_dict.is_integer():
return int(data_dict)
return data_dict
return data_dict

# Normalize and convert auto-corrected data
auto_data = normalize_data(deepcopy(self.results.transcription_corrected.to_dict()))
auto_corrected_json = json.dumps(auto_data, indent=4).splitlines()

# Pass through review server
reviewed_data = start_review_server(self.results.transcription_corrected)

# Get reviewed data as JSON string
human_corrected_json = json.dumps(reviewed_data.to_dict(), indent=4).splitlines()
# Normalize and convert reviewed data
human_data = normalize_data(deepcopy(reviewed_data.to_dict()))
human_corrected_json = json.dumps(human_data, indent=4).splitlines()

self.logger.info("Human review completed")

# Compare the strings, normalizing numbers first
def normalize_numbers(line: str) -> str:
# Convert "x.0" to "x" in the line
return re.sub(r"(\d+)\.0([,\s}])", r"\1\2", line)

# Normalize numbers in both strings before comparing
auto_corrected_lines = [normalize_numbers(line) for line in auto_corrected_json]
human_corrected_lines = [normalize_numbers(line) for line in human_corrected_json]

# Compare the normalized JSON strings
diff = list(
difflib.unified_diff(auto_corrected_lines, human_corrected_lines, fromfile="auto-corrected", tofile="human-corrected")
difflib.unified_diff(auto_corrected_json, human_corrected_json, fromfile="auto-corrected", tofile="human-corrected")
)

if diff:
self.logger.warning("Changes made by human review:")
for line in diff:
self.logger.warning(line.rstrip())

# exit(1)

def generate_outputs(self) -> None:
"""Generate output files based on enabled features and available data."""
self.logger.info("Generating output files")
Expand Down

0 comments on commit b72928a

Please sign in to comment.