annotate software/mpdl-services/mpiwg-mpdl-lt/bin/de/mpg/mpiwg/berlin/mpdl/lt/text/norm/lang/MpdlNormalizerLexDE.lex @ 23:e845310098ba

diverse Korrekturen
author Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
date Tue, 27 Nov 2012 12:35:19 +0100
parents
children
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
23
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
1 /*
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
2 * Normalization rules for German text
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
3 * [this is a JFlex specification]
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
4 *
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
5 * Wolfgang Schmidle
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
6 * version 2011-07-12
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
7 *
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
8 */
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
9
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
10 package de.mpg.mpiwg.berlin.mpdl.lt.text.norm.lang;
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
11
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
12 %%
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
13
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
14 %public
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
15 %class MpdlNormalizerLexDE
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
16 %type java.lang.String
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
17 %unicode
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
18
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
19 // German: de, deu, ger
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
20
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
21 %states DISP, DICT, SEARCH
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
22 %state CELEX, GRIMM
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
23
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
24 %{
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
25 private String original = "";
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
26 private String normalized = "";
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
27 private int problem = 0;
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
28
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
29 private void add (String norm) {
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
30 original += yytext();
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
31 normalized += norm;
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
32 }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
33
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
34 private static final String LB = "[\u002d\u00ad] ";
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
35 %}
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
36
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
37 hyphen = [-\u{00ad}] // hyphen and soft hyphen
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
38 LB = {hyphen} \u0020
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
39 // lb = ({hyphen} \u0020)?
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
40
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
41 END = \n
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
42
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
43 Alphabet = [abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ]
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
44
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
45 %%
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
46
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
47 // jump over empty xml elements
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
48 "<"[^><]+"/>" { add(yytext()); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
49 "-<"[^><]+"/>" { add(yytext()); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
50 "<"[^><]+"></"[^><]+">" { add(yytext()); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
51 "-<"[^><]+"></"[^><]+">" { add(yytext()); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
52
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
53 ſ { add("s"); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
54
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
55 // Fraktur
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
56
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
57 <DISP, DICT, SEARCH,
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
58 GRIMM> {
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
59
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
60 uͦ {add("u"); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
61 aͤ {add("ä"); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
62 oͤ {add("ö"); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
63 uͤ {add("ü"); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
64
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
65 }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
66
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
67 <CELEX> {
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
68
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
69 // normalize ä ö ü ß only for Celex!
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
70
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
71 ä | Ä | aͤ { add("ae"); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
72 ö | Ö | oͤ { add("oe"); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
73 ü | Ü | uͤ { add("ue"); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
74 uͦ {add("u"); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
75 ß { add("ss"); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
76
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
77 {Alphabet} { add(yytext()); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
78
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
79 . { problem = 1; add(yytext()); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
80
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
81 }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
82
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
83 <GRIMM> {
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
84
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
85 ß { add("sz"); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
86
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
87 }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
88
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
89
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
90 // default
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
91
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
92 @ { problem = 1; add(yytext()); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
93 {LB} { add(yytext()); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
94 . { add(yytext()); }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
95
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
96
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
97 <DISP> {
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
98
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
99 {END} {
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
100 switch (problem) {
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
101 case 1: return original;
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
102 default: return normalized;
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
103 }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
104 }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
105 }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
106
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
107 <DICT, CELEX, GRIMM> {
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
108
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
109 {END} {
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
110 switch (problem) {
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
111 case 1: return "";
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
112 default: return normalized.replaceAll(LB, "");
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
113 }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
114 }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
115 }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
116
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
117 <SEARCH> {
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
118
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
119 {END} {
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
120 switch (problem) {
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
121 case 1: return original;
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
122 default: return normalized.replaceAll(LB, "").toLowerCase();
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
123 }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
124 }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
125 }
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
126
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
127
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
128
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
129 /*
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
130
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
131 Annahmen:
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
132 - die Routine wird wortweise aufgerufen, mit einem \n am Ende des Strings
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
133 - Zeilenumbrüche innerhalb des Wortes sind als hyphen/soft hyphen plus space markiert
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
134
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
135 TO DO:
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
136
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
137 DE: Trennung von Deutsch und Fraktur?
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
138 DE: Celex: hyphens weg?
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
139
e845310098ba diverse Korrekturen
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff changeset
140 */