add ocr filter that detects and fixes some common ocr errors, activat…

…e using the 'ocr' filter, or the new 'all' filter

cmd/subber/subber.go

@@ -20,7 +20,7 @@ var (
 	keepAds     = kingpin.Flag("keep-ads", "Do not strip advertisement captions.").Bool()
 	skipBackups = kingpin.Flag("skip-backups", "Do not make backup (.srt.org) of original .srt").Bool()
 	language    = kingpin.Flag("language", "Language.").Default("en").String()
-	filterName  = kingpin.Flag("filter", "Filter (none, caps, html).").Default("none").String()
+	filterName  = kingpin.Flag("filter", "Filter (none, caps, html, ocr, all).").Default("none").String()
 	sync        = kingpin.Flag("sync", "Synchronize captions (milliseconds).").Int()
 )
 

filter.go

@@ -7,10 +7,16 @@ import (
 // FilterCaptions pass the captions through a filter function
 func (subtitle *Subtitle) FilterCaptions(filter string) {
 	switch filter {
+	case "all":
+		subtitle.filterCapitalization()
+		subtitle.filterHTML()
+		subtitle.filterOCR()
 	case "caps":
 		subtitle.filterCapitalization()
 	case "html":
 		subtitle.filterHTML()
+	case "ocr":
+		subtitle.filterOCR()
 	case "none":
 	default:
 		fmt.Printf("Unrecognized filter name: %s\n", filter)

filter_ocr.go

@@ -0,0 +1,41 @@
+package subtitles
+
+import (
+	"strings"
+
+	log "github.com/Sirupsen/logrus"
+)
+
+var (
+	ocrErrors = map[string]string{
+		"s0 ":       "so ",
+		"g0 ":       "go ",
+		"0n ":       "on ",
+		"c0uld":     "could",
+		"s0mething": "something",
+		"l've":      "i've",
+	}
+)
+
+// filterOCR corrects some OCR mistakes
+func (subtitle *Subtitle) filterOCR() *Subtitle {
+	for _, cap := range subtitle.Captions {
+		for i, org := range cap.Text {
+			for bad, good := range ocrErrors {
+				// lower case
+				cap.Text[i] = strings.Replace(cap.Text[i], bad, good, -1)
+
+				// upper case
+				cap.Text[i] = strings.Replace(cap.Text[i], strings.ToUpper(bad), strings.ToUpper(good), -1)
+
+				// ucfirst
+				cap.Text[i] = strings.Replace(cap.Text[i], strings.Title(bad), strings.Title(good), -1)
+			}
+
+			if org != cap.Text[i] {
+				log.Println("[ocr]", org, "->", cap.Text[i])
+			}
+		}
+	}
+	return subtitle
+}

filter_ocr_test.go

@@ -0,0 +1,83 @@
+package subtitles
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestFilterOCRLower(t *testing.T) {
+
+	in := Subtitle{[]Caption{{
+		1,
+		makeTime(0, 0, 4, 630),
+		makeTime(0, 0, 6, 18),
+		[]string{"s0mething good"},
+	}}}
+
+	expected := Subtitle{[]Caption{{
+		1,
+		makeTime(0, 0, 4, 630),
+		makeTime(0, 0, 6, 18),
+		[]string{"something good"},
+	}}}
+
+	assert.Equal(t, &expected, in.filterOCR())
+}
+
+func TestFilterOCRUpper(t *testing.T) {
+
+	in := Subtitle{[]Caption{{
+		1,
+		makeTime(0, 0, 4, 630),
+		makeTime(0, 0, 6, 18),
+		[]string{"S0METHING GOOD"},
+	}}}
+
+	expected := Subtitle{[]Caption{{
+		1,
+		makeTime(0, 0, 4, 630),
+		makeTime(0, 0, 6, 18),
+		[]string{"SOMETHING GOOD"},
+	}}}
+
+	assert.Equal(t, &expected, in.filterOCR())
+}
+
+func TestFilterOCRUcFirst(t *testing.T) {
+
+	in := Subtitle{[]Caption{{
+		1,
+		makeTime(0, 0, 4, 630),
+		makeTime(0, 0, 6, 18),
+		[]string{"S0mething good"},
+	}}}
+
+	expected := Subtitle{[]Caption{{
+		1,
+		makeTime(0, 0, 4, 630),
+		makeTime(0, 0, 6, 18),
+		[]string{"Something good"},
+	}}}
+
+	assert.Equal(t, &expected, in.filterOCR())
+}
+
+func TestFilterOCREnglish(t *testing.T) {
+
+	in := Subtitle{[]Caption{{
+		1,
+		makeTime(0, 0, 4, 630),
+		makeTime(0, 0, 6, 18),
+		[]string{"l've got a feeling"},
+	}}}
+
+	expected := Subtitle{[]Caption{{
+		1,
+		makeTime(0, 0, 4, 630),
+		makeTime(0, 0, 6, 18),
+		[]string{"i've got a feeling"},
+	}}}
+
+	assert.Equal(t, &expected, in.filterOCR())
+}