Xapian discuss - Sep 2008 - Spelling correction in Omega

I'm trying to add spelling correction to Omega (1.0.7) by defining a  
$suggestion OmegaScript keyword with the following patch, but  
Xapian::QueryParser::get_corrected_query_string() always returns an  
empty string. I'm using a database with tens of thousands of  
documents, each with hundreds of words, so I would expect that a  
misspelling like "comptuer" should readily be corrected to
"computer,"
but it's not. I've tried numerous misspelled queries but none of them  
yield a correction string, so I suspect I'm doing something wrong.  
Does anyone have any ideas?

Thank you,
Yevgeny Binder

--------------------------

--- query.cc.orig	2008-07-15 23:42:17.000000000 -0500
+++ query.cc	2008-09-03 20:17:02.000000000 -0500
@@ -236,7 +236,11 @@
      }

      try {
-	query = qp.parse_query(query_string);
+	query = qp.parse_query(query_string,
+			qp.FLAG_PHRASE |
+			qp.FLAG_PARTIAL |
+			qp.FLAG_AUTO_MULTIWORD_SYNONYMS |
+			qp.FLAG_SPELLING_CORRECTION);
      } catch (Xapian::QueryParserError &e) {
  	error_msg = e.get_msg();
  	return BAD_QUERY;
@@ -799,6 +803,7 @@
  CMD_stoplist,
  CMD_sub,
  CMD_substr,
+CMD_suggestion,
  CMD_terms,
  CMD_thispage,
  CMD_time,
@@ -918,6 +923,7 @@
  T(stoplist,	   0, 0, N, Q), // return list of stopped terms
  T(sub,		   2, 2, N, 0), // subtract
  T(substr,	   2, 3, N, 0), // substring
+T(suggestion,	   0, 0, N, Q), // misspelled word correction suggestion
  T(terms,	   0, 0, N, M), // list of matching terms
  T(thispage,	   0, 0, N, M), // page number of current page
  T(time,		   0, 0, N, M), // how long the match took (in seconds)
@@ -1767,6 +1773,9 @@
  		value = args[0].substr(start, len);
  		break;
  	    }
+	    case CMD_suggestion:
+	    value = qp.get_corrected_query_string();
+		break;
  	    case CMD_terms:
  		if (enquire) {
  		    // list of matching terms

On Fri, Sep 05, 2008 at 04:20:44PM -0500, Yevgeny Binder
wrote:
> I'm trying to add spelling correction to Omega (1.0.7) by defining a  
> $suggestion OmegaScript keyword with the following patch, but  
> Xapian::QueryParser::get_corrected_query_string() always returns an  
> empty string. I'm using a database with tens of thousands of  
> documents, each with hundreds of words, so I would expect that a  
> misspelling like "comptuer" should readily be corrected to
"computer,"
> but it's not. I've tried numerous misspelled queries but none of
them
> yield a correction string, so I suspect I'm doing something wrong.  
You also need to add the spelling targets to the database at index time.

It would be nice if Omega supported this - probably just by default, as
I suspect the overhead is fairly small.  So if you get it working,
please submit a patch.

Cheers,
    Olly

On Sep 12, 2008, at 11:43 PM, Olly Betts wrote:
> You also need to add the spelling targets to the database at index  
> time.
Thanks, Olly, that did the trick. Omega has spelling suggestion now.
> It would be nice if Omega supported this - probably just by default,  
> as
> I suspect the overhead is fairly small.  So if you get it working,
> please submit a patch.
Done. I left it as an opt-in feature for now, but it would be simple  
to make it enabled by default.

- Yevgeny