Updated DefaultWordRunReorderWindow to 1.0

README.md

@@ -59,7 +59,7 @@ Environment variables:
 | Environment Variable | Default | Description |
 | --- | --- | --- |
 | `SERIATIM_OVERLAP_WORD_RUN_GAP` | `0.75` | Maximum gap in seconds between adjacent timed words when `resolve-overlaps` builds word-run replacement segments. Must be a positive float. |
-| `SERIATIM_OVERLAP_WORD_RUN_REORDER_WINDOW` | `0.4` | Near-start window in seconds for ordering replacement word runs shortest-first. Must be a positive float. |
+| `SERIATIM_OVERLAP_WORD_RUN_REORDER_WINDOW` | `1.0` | Near-start window in seconds for ordering replacement word runs shortest-first. Must be a positive float. |
 | `SERIATIM_BACKCHANNEL_MAX_DURATION` | `2.0` | Maximum duration in seconds for `backchannel` classification. Must be a positive float. |
 | `SERIATIM_FILLER_MAX_DURATION` | `1.25` | Maximum duration in seconds for `filler` classification. Must be a positive float. |
 
@@ -279,7 +279,7 @@ For each detected overlap group, `resolve-overlaps` uses preserved WhisperX word
 - The default word-run gap is `0.75` seconds.
 - Set `SERIATIM_OVERLAP_WORD_RUN_GAP` to a positive number of seconds to override the default.
 - Near-start replacement word runs are reordered so shorter segments come first when adjacent starts are within `SERIATIM_OVERLAP_WORD_RUN_REORDER_WINDOW`.
-- The default word-run reorder window is `0.4` seconds.
+- The default word-run reorder window is `1.0` seconds.
 - Set `SERIATIM_OVERLAP_WORD_RUN_REORDER_WINDOW` to a positive number of seconds to override the default.
 - Replacement segment text is built by joining word text with single spaces.
 - Replacement segments include `source_ref` and `derived_from`.

internal/cli/merge_test.go

@@ -8,6 +8,7 @@ import (
 	"strings"
 	"testing"
 
+	"gitea.maximumdirect.net/eric/seriatim/internal/config"
 	"gitea.maximumdirect.net/eric/seriatim/internal/model"
 	"gitea.maximumdirect.net/eric/seriatim/internal/report"
 	"gitea.maximumdirect.net/eric/seriatim/schema"
@@ -342,6 +343,8 @@ func TestMergeDetectsOverlapGroups(t *testing.T) {
 }
 
 func TestMergeResolvesOverlapGroupsWithWordRuns(t *testing.T) {
+	t.Setenv(config.WordRunReorderWindowEnv, "0.4")
+
 	dir := t.TempDir()
 	inputA := writeJSONFile(t, dir, "a.json", `{
 		"segments": [
@@ -1677,6 +1680,75 @@ func TestMergeResolveOverlapsAbsorbsNearbyContext(t *testing.T) {
 	}
 }
 
+func TestMergeResolveOverlapsPreservesAbsorbedContextPrefix(t *testing.T) {
+	dir := t.TempDir()
+	inputA := writeJSONFile(t, dir, "a.json", `{
+		"segments": [
+			{
+				"start": 7,
+				"end": 9.95,
+				"text": "full context prefix near",
+				"words": [
+					{"word": "full", "start": 7.1, "end": 7.2},
+					{"word": "context", "start": 7.3, "end": 7.4},
+					{"word": "prefix", "start": 7.5, "end": 7.6},
+					{"word": "near", "start": 9.7, "end": 9.9}
+				]
+			},
+			{
+				"start": 10,
+				"end": 11,
+				"text": "inside",
+				"words": [
+					{"word": "inside", "start": 10.5, "end": 10.7}
+				]
+			}
+		]
+	}`)
+	inputB := writeJSONFile(t, dir, "b.json", `{
+		"segments": [
+			{
+				"start": 10.2,
+				"end": 11,
+				"text": "bob",
+				"words": [
+					{"word": "bob", "start": 10.4, "end": 10.6}
+				]
+			}
+		]
+	}`)
+	speakers := writeYAMLFile(t, dir, "speakers.yml", `match:
+  - speaker: Alice
+    match: ["a.json"]
+  - speaker: Bob
+    match: ["b.json"]
+`)
+	output := filepath.Join(dir, "merged.json")
+
+	err := executeMerge(
+		"--input-file", inputA,
+		"--input-file", inputB,
+		"--speakers", speakers,
+		"--output-schema", "minimal",
+		"--output-file", output,
+	)
+	if err != nil {
+		t.Fatalf("merge failed: %v", err)
+	}
+
+	var transcript schema.MinimalTranscript
+	readJSON(t, output, &transcript)
+	aliceText := make([]string, 0)
+	for _, segment := range transcript.Segments {
+		if segment.Speaker == "Alice" {
+			aliceText = append(aliceText, segment.Text)
+		}
+	}
+	if strings.Join(aliceText, " ") != "full context prefix near inside" {
+		t.Fatalf("expected full absorbed context prefix in Alice output, got %#v", transcript.Segments)
+	}
+}
+
 func TestInvalidTimingFails(t *testing.T) {
 	tests := []struct {
 		name string

internal/config/config.go

@@ -17,7 +17,7 @@ const (
 	DefaultPreprocessingModules   = "validate-raw,normalize-speakers,trim-text"
 	DefaultPostprocessingModules  = "detect-overlaps,resolve-overlaps,backchannel,filler,coalesce,detect-overlaps,autocorrect,assign-ids,validate-output"
 	DefaultOverlapWordRunGap      = 0.75
-	DefaultWordRunReorderWindow   = 0.4
+	DefaultWordRunReorderWindow   = 1.0
 	DefaultCoalesceGap            = 3.0
 	DefaultCoalesceGapValue       = "3.0"
 	DefaultBackchannelMaxDuration = 2.0

internal/config/config_test.go

@@ -191,7 +191,7 @@ func TestOverlapWordRunGapRejectsInvalidEnvOverride(t *testing.T) {
 	}
 }
 
-func TestWordRunReorderWindowDefaultsTo04(t *testing.T) {
+func TestWordRunReorderWindowDefaultsTo1(t *testing.T) {
 	t.Setenv(WordRunReorderWindowEnv, "")
 	dir := t.TempDir()
 	input := writeTempFile(t, dir, "input.json")

internal/overlap/resolve.go

@@ -171,7 +171,7 @@ func resolveGroup(in model.MergedTranscript, group model.OverlapGroup, refToInde
 	speakers := groupSpeakerOrder(group, segmentsBySpeaker)
 	resolved := resolvedGroup{}
 	for speakerIndex, speaker := range speakers {
-		timedWords, untimedWords := gatherResolutionWords(segmentsBySpeaker[speaker], expandedStart, expandedEnd)
+		timedWords, untimedWords := gatherResolutionWords(segmentsBySpeaker[speaker])
 		if len(timedWords) == 0 {
 			continue
 		}
@@ -303,7 +303,7 @@ func groupSpeakerOrder(group model.OverlapGroup, segmentsBySpeaker map[string][]
 	return speakers
 }
 
-func gatherResolutionWords(segments []model.Segment, groupStart float64, groupEnd float64) ([]resolutionWord, []resolutionWord) {
+func gatherResolutionWords(segments []model.Segment) ([]resolutionWord, []resolutionWord) {
 	timedWords := make([]resolutionWord, 0)
 	untimedWords := make([]resolutionWord, 0)
 	sequence := 0
@@ -321,9 +321,6 @@ func gatherResolutionWords(segments []model.Segment, groupStart float64, groupEn
 				untimedWords = append(untimedWords, candidate)
 				continue
 			}
-			if word.End <= groupStart || word.Start >= groupEnd {
-				continue
-			}
 			timedWords = append(timedWords, candidate)
 		}
 	}

internal/overlap/resolve_test.go

@@ -72,7 +72,7 @@ func TestResolveCreatesChronologicalWordRunSegments(t *testing.T) {
 	}
 }
 
-func TestResolveIncludesWordsByIntervalIntersection(t *testing.T) {
+func TestResolvePreservesAllWordsFromSelectedOverlapSegment(t *testing.T) {
 	merged := model.MergedTranscript{
 		Segments: []model.Segment{
 			segmentWithWords(
@@ -101,11 +101,98 @@ func TestResolveIncludesWordsByIntervalIntersection(t *testing.T) {
 	if len(got.Segments) != 1 {
 		t.Fatalf("segment count = %d, want 1", len(got.Segments))
 	}
-	if got.Segments[0].Text != "left-edge inside right-edge" {
+	if got.Segments[0].Text != "before left-edge inside right-edge after" {
 		t.Fatalf("text = %q", got.Segments[0].Text)
 	}
 }
 
+func TestResolvePreservesPrefixWordsFromAbsorbedContextSegment(t *testing.T) {
+	merged := model.MergedTranscript{
+		Segments: []model.Segment{
+			segmentWithWords(
+				"a.json",
+				0,
+				"Alice",
+				7,
+				9.5,
+				word("prefix", 7.2, 7.4),
+				word("near", 9.2, 9.4),
+			),
+			segmentWithWords("a.json", 1, "Alice", 10, 12, word("inside", 10.5, 10.7)),
+			segmentWithWords("b.json", 0, "Bob", 10.2, 11.2, word("bob", 10.4, 10.6)),
+		},
+		OverlapGroups: []model.OverlapGroup{
+			group(1, 10, 12, []string{"a.json#1", "b.json#0"}, []string{"Alice", "Bob"}),
+		},
+	}
+
+	got, _, err := Resolve(merged, 10, 0.4, 3)
+	if err != nil {
+		t.Fatalf("resolve failed: %v", err)
+	}
+	if gotTexts(got.Segments) != "prefix near inside,bob" {
+		t.Fatalf("segment texts = %s", gotTexts(got.Segments))
+	}
+}
+
+func TestResolvePreservesSuffixWordsFromAbsorbedContextSegment(t *testing.T) {
+	merged := model.MergedTranscript{
+		Segments: []model.Segment{
+			segmentWithWords("a.json", 0, "Alice", 10, 12, word("inside", 10.5, 10.7)),
+			segmentWithWords(
+				"a.json",
+				1,
+				"Alice",
+				12.5,
+				16,
+				word("near", 12.7, 12.9),
+				word("suffix", 15.6, 15.8),
+			),
+			segmentWithWords("b.json", 0, "Bob", 10.2, 11.2, word("bob", 10.4, 10.6)),
+		},
+		OverlapGroups: []model.OverlapGroup{
+			group(1, 10, 12, []string{"a.json#0", "b.json#0"}, []string{"Alice", "Bob"}),
+		},
+	}
+
+	got, _, err := Resolve(merged, 10, 0.4, 3)
+	if err != nil {
+		t.Fatalf("resolve failed: %v", err)
+	}
+	if gotTexts(got.Segments) != "bob,inside near suffix" {
+		t.Fatalf("segment texts = %s", gotTexts(got.Segments))
+	}
+}
+
+func TestResolveOutOfWindowWordsAffectWordRunBoundaries(t *testing.T) {
+	merged := model.MergedTranscript{
+		Segments: []model.Segment{
+			segmentWithWords(
+				"a.json",
+				0,
+				"Alice",
+				7,
+				9.5,
+				word("early", 7.2, 7.4),
+				word("near", 9.2, 9.4),
+			),
+			segmentWithWords("a.json", 1, "Alice", 10, 12, word("inside", 10.5, 10.7)),
+			segmentWithWords("b.json", 0, "Bob", 10.2, 11.2, word("bob", 10.4, 10.6)),
+		},
+		OverlapGroups: []model.OverlapGroup{
+			group(1, 10, 12, []string{"a.json#1", "b.json#0"}, []string{"Alice", "Bob"}),
+		},
+	}
+
+	got, _, err := Resolve(merged, 0.75, 0.4, 3)
+	if err != nil {
+		t.Fatalf("resolve failed: %v", err)
+	}
+	if gotTexts(got.Segments) != "early,near,bob,inside" {
+		t.Fatalf("segment texts = %s", gotTexts(got.Segments))
+	}
+}
+
 func TestResolveIncludesContextWordsAroundOverlapWindow(t *testing.T) {
 	merged := model.MergedTranscript{
 		Segments: []model.Segment{