RTVNoord

.gitignore

@@ -14,6 +14,8 @@ NU/nu
 RO/metadata
 RO/ro
 RO/xml2txt
+RTVNoord/metadata
+RTVNoord/rtvnoord
 Sargasso/metadata
 Sargasso/sargasso
 Sargasso/xml2txt
@@ -34,3 +36,4 @@ bin/week2files
 corpus
 data
 tmp
+lock

Makefile

@@ -6,6 +6,7 @@ all:
 	make -C NOS
 	make -C NU
 	make -C RO
+	make -C RTVNoord
 	make -C Sargasso
 	make -C Sikkom
 	make -C Tzum

RTVNoord/Makefile

@@ -0,0 +1,9 @@
+all: \
+	metadata \
+	rtvnoord
+
+metadata: cmd/metadata/*.go
+	go build -o $@ $^
+
+rtvnoord: cmd/rtvnoord/*.go
+	go build -o $@ $^

RTVNoord/cmd/metadata/metadata.go

@@ -0,0 +1,121 @@
+package main
+
+import (
+	e "codeberg.org/pebbe/errors"
+
+	"bufio"
+	"encoding/xml"
+	"fmt"
+	"html"
+	"os"
+	"strings"
+	"time"
+)
+
+type Item struct {
+	XMLName  xml.Name `xml:"item"`
+	UnixTime int64    `xml:"unixTime"`
+}
+
+var (
+	x        = e.ExitErr
+	escape   = html.EscapeString
+	data     = make(map[string][]string)
+	location *time.Location
+)
+
+func main() {
+	var err error
+	location, err = time.LoadLocation("Europe/Amsterdam")
+	x(err)
+
+	files, err := os.ReadDir("..")
+	x(err)
+	for _, file := range files {
+		filename := file.Name()
+		if strings.HasSuffix(filename, ".txt") {
+			doText(filename)
+		} else if strings.HasSuffix(filename, ".xml") {
+			doXml(filename)
+		}
+	}
+
+	files, err = os.ReadDir("xml")
+	x(err)
+	for _, file := range files {
+		filename := file.Name()
+		if !strings.HasSuffix(filename, ".xml") {
+			continue
+		}
+		aa := strings.Split(filename, ".")
+		base := strings.Join(aa[1:len(aa)-2], ".")
+		b, err := os.ReadFile("xml/" + filename)
+		x(err)
+		s := string(b)
+		i := strings.Index(s, "<alpino") + 1
+		i += strings.Index(s[i:], "<")
+		fp, err := os.Create("xml/" + filename + ".tmp")
+		x(err)
+		x(fp.WriteString(s[:i]))
+		x(fp.WriteString("<metadata>\n    <meta type=\"text\" name=\"source\" value=\"NU\"/>\n"))
+		for _, m := range data[base] {
+			x(fp.WriteString("    " + m + "\n"))
+		}
+		x(fp.WriteString("  </metadata>\n  "))
+		x(fp.WriteString(stripMeta(s[i:])))
+		x(fp.Close())
+		x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
+	}
+}
+
+func doText(filename string) {
+	base := filename[:len(filename)-4]
+	if _, ok := data[base]; !ok {
+		data[base] = make([]string, 0)
+	}
+	fp, err := os.Open("../" + filename)
+	x(err)
+	defer func() { x(fp.Close()) }()
+	scanner := bufio.NewScanner(fp)
+	for scanner.Scan() {
+		line := scanner.Text()
+		if !strings.HasPrefix(line, "##META") {
+			continue
+		}
+		aa := strings.Fields(line)
+		if len(aa) > 4 {
+			data[base] = append(data[base],
+				fmt.Sprintf(`<meta type="%s" name="%s" value="%s"/>`,
+					aa[1],
+					escape(aa[2]),
+					escape(strings.Join(aa[4:], " "))))
+		}
+	}
+	x(scanner.Err())
+}
+
+func doXml(filename string) {
+	base := filename[:len(filename)-4]
+	if _, ok := data[base]; !ok {
+		data[base] = make([]string, 0)
+	}
+	b, err := os.ReadFile("../" + filename)
+	x(err)
+	var item Item
+	x(xml.Unmarshal(b, &item))
+	t := time.Unix(item.UnixTime, 0).In(location)
+	data[base] = append(data[base],
+		fmt.Sprintf(`<meta type="date" name="pubdate" value="%d-%02d-%02d"/>`,
+			t.Year(),
+			int(t.Month()),
+			t.Day()))
+}
+
+func stripMeta(s string) string {
+	i1 := strings.Index(s, "<metadata>")
+	if i1 < 0 {
+		return s
+	}
+	i2 := i1 + strings.Index(s[i1:], "</metadata>") + 11
+	return s[:i1] + strings.TrimLeft(s[i2:], " \t\r\n")
+}

RTVNoord/cmd/rtvnoord/rtvnoord.go

@@ -0,0 +1,290 @@
+package main
+
+import (
+	e "codeberg.org/pebbe/errors"
+
+	"encoding/json"
+	"encoding/xml"
+	"fmt"
+	"io"
+	"net/http"
+	"os"
+	"path/filepath"
+	"strings"
+	"time"
+)
+
+type Rss struct {
+	XMLName xml.Name `xml:"rss"`
+	Items   []ItemT  `xml:"channel>item"`
+}
+
+type ItemT struct {
+	PubDate  string `xml:"pubDate"`
+	UnixTime int64  `xml:"unixTime"`
+	Guid     string `xml:"guid"`
+	Link     string `xml:"link"`
+	Data     []byte `xml:",innerxml"`
+}
+
+type Doc struct {
+	Type  string   `json:"@type"`
+	Cat   string   `json:"articleSection"`
+	Tags  []string `json:"keywords"`
+	Title string   `json:"headline"`
+	Text  string   `json:"articleBody"`
+}
+
+type GItem struct {
+	ArticleBody    string   `json:"articleBody"`
+	ArticleSection []string `json:"articleSection"`
+}
+
+var (
+	p     = e.PanicErr
+	w     = e.WarnErr
+	agent = "AhrefsBot/7.0"
+)
+
+func exists(filename string) bool {
+	_, err := os.Stat(filename)
+	return err == nil
+}
+
+func fileDate(filename string) string {
+	b, err := os.ReadFile(filename)
+	if err != nil {
+		return ""
+	}
+	s := string(b)
+	i1 := strings.Index(s, "<unixTime>") + 10
+	i2 := strings.Index(s, "</unixTime>")
+	if i2 < i1 {
+		return ""
+	}
+	return s[i1:i2]
+}
+
+func main() {
+	defer func() {
+		if e.Panicked {
+			_ = recover()
+			os.Exit(1)
+		}
+	}()
+
+	myLock := "/net/corpora/nlnieuws/RTVNoord/lock"
+	mkLock(myLock)
+	defer func() {
+		_ = os.Remove(myLock)
+	}()
+
+	resp, err := http.Get("https://www.rtvnoord.nl/rss/index.xml")
+	p(err)
+	body, err := io.ReadAll(resp.Body)
+	p(err)
+	p(resp.Body.Close())
+
+	var rss Rss
+	p(xml.Unmarshal(body, &rss))
+
+	if len(rss.Items) == 0 {
+		p(fmt.Errorf("len(rss.Items) == 0"))
+	}
+
+	for _, item := range rss.Items {
+		t, err := time.Parse(time.RFC1123Z, item.PubDate)
+		if err != nil {
+			t, err = time.Parse(time.RFC1123, item.PubDate)
+		}
+		p(err)
+		year, week := t.ISOWeek()
+		dirname := fmt.Sprintf("/net/corpora/nlnieuws/RTVNoord/%d/%02d", year, week)
+		if exists(dirname + "/lock") {
+			continue
+		}
+		guid := strings.TrimPrefix(item.Guid, "https://www.rtvnoord.nl/")
+		guid = strings.TrimRight(guid, "/-")
+		guid = strings.ReplaceAll(guid, "/", "_")
+		filename := dirname + "/" + guid
+
+		ts := fmt.Sprintf("%d", t.Unix())
+		needUpdate := fileDate(filename+".xml") != ts
+
+		p(os.MkdirAll(dirname, 0777))
+		func() {
+			var ok bool
+			defer func() {
+				if e.Panicked {
+					fmt.Fprintln(os.Stderr, "----", filename)
+					fmt.Fprintln(os.Stderr, "----", item.Link)
+				}
+				if !ok {
+					_ = os.Remove(filename + ".xml")
+				}
+			}()
+			fp, err := os.Create(filename + ".xml")
+			p(err)
+			p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
+			p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
+			p(fp.Write(item.Data))
+			p(fp.WriteString("</item>\n"))
+			p(fp.Close())
+			p(os.Chtimes(filename+".xml", t, t))
+			ok = doArticle(filename, item.Link, t, needUpdate)
+		}()
+	}
+}
+
+func doArticle(filename string, url string, timestamp time.Time, needUpdate bool) bool {
+	if exists(filename + ".skip") {
+		return true
+	}
+	if needUpdate {
+		_ = os.Remove(filename + ".err")
+		_ = os.Remove(filename + ".html")
+		_ = os.Remove(filename + ".json")
+		_ = os.Remove(filename + ".txt")
+	} else {
+		// voor sommige berichten is geen .txt, alleen .json
+		if exists(filename + ".json") {
+			return true
+		}
+	}
+
+	time.Sleep(2 * time.Second)
+
+	req, err := http.NewRequest("GET", url, nil)
+	p(err)
+	req.Header.Set("User-Agent", agent)
+
+	client := &http.Client{}
+	resp, err := client.Do(req)
+	p(err)
+	body, err := io.ReadAll(resp.Body)
+	p(err)
+	p(resp.Body.Close())
+
+	s := string(body)
+	ok := true
+	var doc Doc
+	skip := 0
+	for {
+		var i2 int
+		var s2 string
+		i1 := strings.Index(s[skip:], `type="application/ld+json"`) + skip
+		if i1 < 0 {
+			ok = false
+			break
+		} else {
+			i1 += strings.Index(s[i1:], `>`) + 1
+			i2 = i1 + strings.Index(s[i1:], `</script>`)
+			if i2 < i1 {
+				ok = false
+				break
+			} else {
+				s2 = s[i1:i2]
+			}
+		}
+		if err = json.Unmarshal([]byte(s2), &doc); err != nil {
+			_ = w(err, url)
+			skip = i2
+			continue
+		}
+		if doc.Type == "Organization" {
+			skip = i2
+			continue
+		}
+		s = s2
+		break
+	}
+	if !ok {
+		_ = w(fmt.Errorf("script jsonld not found: %s", url))
+
+		fp, err := os.Create(filename + ".err")
+		p(err)
+		p(fmt.Fprintf(fp, "script jsonld not found: %s\n", url))
+		p(fp.Close())
+		p(os.Chtimes(filename+".err", timestamp, timestamp))
+
+		fp, err = os.Create(filename + ".html")
+		p(err)
+		p(fp.Write(body))
+		p(fp.Close())
+		p(os.Chtimes(filename+".html", timestamp, timestamp))
+
+		return false
+	}
+	fp, err := os.Create(filename + ".json")
+	p(err)
+	p(fp.WriteString(s))
+	p(fp.Close())
+	p(os.Chtimes(filename+".json", timestamp, timestamp))
+
+	// text bevat kopjes zonder punt aan het eind
+	lines := strings.Split(doc.Text, "\n")
+	for i, line := range lines {
+		lines[i] = addEnd(fixSpace(line))
+	}
+	text := strings.Join(lines, "") + "\n"
+
+	fp, err = os.Create(filename + ".txt")
+	p(err)
+	if len(doc.Tags) == 0 {
+		p(fmt.Fprintln(fp, "##META text tag ="))
+	} else {
+		for _, tag := range doc.Tags {
+			p(fmt.Fprintf(fp, "##META text tag = %s\n", strings.ToLower(fixSpace(tag))))
+		}
+	}
+	if doc.Cat == "" {
+		p(fmt.Fprintln(fp, "##META text cat ="))
+	} else {
+		p(fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(doc.Cat)))
+	}
+
+	p(fp.WriteString(addEnd(doc.Title)))
+
+	p(fp.WriteString(text))
+	p(fp.Close())
+	p(os.Chtimes(filename+".txt", timestamp, timestamp))
+
+	return true
+}
+
+func addEnd(s string) string {
+	s = strings.TrimSpace(s)
+	n := len(s)
+	if n == 0 {
+		return ""
+	}
+	if n > 0 {
+		if strings.ContainsAny(s[n-1:], ".!?") {
+			return s + "\n"
+		}
+	}
+	if n > 1 {
+		s2 := s[n-2:]
+		if s2 == `."` || s2 == `!"` || s2 == `?"` || s2 == `.'` || s2 == `!'` || s2 == `?'` {
+			return s + "\n"
+		}
+	}
+	return s + ".\n"
+}
+
+func fixSpace(s string) string {
+	return strings.Join(strings.Fields(s), " ")
+}
+
+func mkLock(filename string) {
+	pid := os.Getpid()
+	link := fmt.Sprintf("%s.%d", filepath.Base(filename), pid)
+	p(os.Symlink(link, filename))
+
+	name, err := os.Readlink(filename)
+	p(err)
+
+	if name != link {
+		p(fmt.Errorf("wrong lock name %q, should be %q", name, link))
+	}
+}

RTVNoord/txt2corpus.sh

@@ -0,0 +1,66 @@
+#!/bin/bash
+
+set -e
+
+unset CDPATH
+PATH=/net/corpora/nlnieuws/bin:/net/aps/bin:$PATH
+export TZ=Europe/Amsterdam
+. /net/aps/etc/alpino-activate.sh > /dev/null
+
+if [ "$1" = "" ]
+then
+    ds=`ISOWeek -7`
+else
+    case "$1" in
+        2[0-9][0-9][0-9]-[0-5][0-9])
+            ds=$1
+            ;;
+        *)
+            echo INVALID
+            exit 1
+            ;;
+    esac
+fi
+
+dp=${ds//-//}
+
+corpus=/net/corpora/nlnieuws/RTVNoord/corpus/$ds
+
+cd /net/corpora/nlnieuws/RTVNoord/$dp
+
+ln -s lock.$$ lock
+if [ "`readlink lock`" != lock.$$ ]
+then
+    echo Getting lock failed
+    exit 1
+fi
+
+rm -fr out
+mkdir out
+
+rm -f $corpus.lines
+for i in *.txt
+do
+    b=`basename $i .txt`
+    perl -p -e 's/^\s*//; s/^##META.*\n//' $i | tokenize.sh \
+        | perl -e '$n = 0; while(<>) { $n++; print("rtvnoord.'$b'.$n|$_"); }' \
+        >> $corpus.lines
+done
+
+cd out
+mkdir xml
+Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
+
+../../../metadata
+
+cd xml
+alto -o $corpus.data.dz *.xml 2> /dev/null
+
+# telling per bericht, niet per zin
+/net/corpora/nlnieuws/namen.sh -x C -s $corpus.data.dz > $corpus.cat.txt
+/net/corpora/nlnieuws/namen.sh -x T -s $corpus.data.dz > $corpus.tag.txt
+
+cd ../..
+rm -fr out
+
+rm -f lock

collect.sh

@@ -40,7 +40,7 @@ cd /net/corpora/nlnieuws/data
 declare -A parts
 parts[alles]='.'
 parts[algemeen]='NOS|NU|NieuwsNL|RO|Sargasso'
-parts[groningen]='GG|Sikkom'
+parts[groningen]='GG|RTVNoord|Sikkom'
 parts[AT5]='AT5'
 parts[GG]='GG'
 parts[NOS]='NOS'