parse text without explicit endtag - antlr

Question

The problem of parsing a dictionary-entry (see examples below) is based on
not having explicit start- and end-tags but:
the end-tag of one element is already the start-tag of the following element
or: the start-tag is not a syntactical element, but it is the current parsing state (so it depends on what you have already "seen" in the inputstream)
example 1, simple entry:
wordWithoutSpace [phonetic information]
definition as everything until colon: example sentence until EOF
example 2, entry with multi-definition:
wordWithoutSpace [phonetic information]
1. first definition until colon: example sentence until second definition
2. second definition until colon: example sentence until EOF
As I would say it in words or pseudo-code:
dictionary-entry :
word = .+ ' ' // catch everything as word until you see a space
phon = '[' .+ ']' // then follows phonetic, which is everything in brackets
(MultipleMeaning | UniqueMeaning)
MultipleMeaning : Int '.' definition= .+ ':' // a MultipleMeaning has a number
// before the definition
UniqueMeaning : definition= .+ ':'
I've tried the Lexer with gates (antlr-version: 3.2)
#members {
int cs = 0; // current state
}
#lexer::header {
package main;
}
Word :
{cs==0}?=> .+ ' ' {cs=1;} // in this state everything until
; // Space belongs to the Word, now go to Phon-mode
Phon :
{cs==1}?=> '[' .+ ']' {cs=2;} // everything in brackets is phonetic-information
; // after you have seen this go to next state
MultiDef :
{cs==2}?=> Int '.' .+ ':' {cs=3;}
;
Def :
{cs==2}?=> .+ ':' {cs=3;}
;
fragment
Digit :
'0'..'9';
Int :
Digit Digit*;
The TestLexer:
import org.antlr.runtime.ANTLRStringStream;
import org.antlr.runtime.CharStream;
import org.antlr.runtime.Token;
public class TestLexer {
public static void main(String[] args) {
String str = "Word [phon]1.definition:";
CharStream input = new ANTLRStringStream(str);
DudenLexer lexer = new DudenLexer(input);
Token token;
while ((token = lexer.nextToken())!=Token.EOF_TOKEN) {
System.out.println("Token: "+token);
}
}
}
The problems I have:
I get the error-msg: line 1:0 rule Def failed predicate: {cs==2}?
I don't know if this is even the right way to do?
I am stucked for about three days with this and am very thankful for any help and hints.
Thank you,
Tom