Mercurial > hg > mpdl-group

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

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
diverse Korrekturen
author Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
date Tue, 27 Nov 2012 12:35:19 +0100
parents
children
comparison
equal deleted inserted replaced
22:6a45a982c333 23:e845310098ba
1 /*
2 * Normalization rules for English text
3 * [this is a JFlex specification]
4 *
5 * Wolfgang Schmidle
6 * version 2011-07-12
7 *
8 */
9
10 package de.mpg.mpiwg.berlin.mpdl.lt.text.norm.lang;
11
12 %%
13
14 %public
15 %class MpdlNormalizerLexEN
16 %type java.lang.String
17 %unicode
18
19 // 1.5 English: en
20
21 %states DISP, DICT, SEARCH
22
23 %{
24 private String original = "";
25 private String normalized = "";
26 private int problem = 0;
27
28 private void add (String norm) {
29 original += yytext();
30 normalized += norm;
31 }
32
33 private static final String LB = "[\u002d\u00ad] ";
34 %}
35
36 hyphen = [-\u{00ad}] // hyphen and soft hyphen
37 LB = {hyphen} \u0020
38 // lb = ({hyphen} \u0020)?
39
40 END = \n
41
42 %%
43
44 // jump over empty xml elements
45 "<"[^><]+"/>" { add(yytext()); }
46 "-<"[^><]+"/>" { add(yytext()); }
47 "<"[^><]+"></"[^><]+">" { add(yytext()); }
48 "-<"[^><]+"></"[^><]+">" { add(yytext()); }
49
50 <DISP, DICT, SEARCH> {
51
52 ſ { add("s"); }
53
54 }
55
56
57 // default
58
59 @ { problem = 1; add(yytext()); }
60 {LB} { add(yytext()); }
61 . { add(yytext()); }
62
63
64 <DISP> {
65
66 {END} {
67 switch (problem) {
68 case 1: return original;
69 default: return normalized;
70 }
71 }
72 }
73
74 <DICT> {
75
76 {END} {
77 switch (problem) {
78 case 1: return "";
79 default: return normalized.replaceAll(LB, "");
80 }
81 }
82 }
83
84 <SEARCH> {
85
86 {END} {
87 switch (problem) {
88 case 1: return original;
89 default: return normalized.replaceAll(LB, "").toLowerCase();
90 }
91 }
92 }
93
94
95 /*
96
97 Annahmen:
98 - die Routine wird wortweise aufgerufen, mit einem \n am Ende des Strings
99 - Zeilenumbrüche innerhalb des Wortes sind als hyphen/soft hyphen plus space markiert
100
101 TO DO:
102
103 EN: vollständig?
104
105 */