Oracle text clustering - use of Stemming

Question

Hi,
I am using K-means to cluster a set of documents. But I am unable to set the clustering algorithm parameters to use stemming for the tokens. The clustering algorithm uses 'move', 'moving', 'moves', 'moved' as separate words and clusters the documents into different clusters. I would like to group all the documents that contain 'move', 'moving', 'moves', 'moved' into a single group by using the stem 'move'. I am unable to do this so far. In case any of you have some ideas, please suggest.
I use the following preferences and attributes to create a text index:
BEGIN
CTX_DDL.DROP_PREFERENCE ('test_lex');
CTX_DDL.CREATE_PREFERENCE ('test_lex', 'BASIC_LEXER');
CTX_DDL.SET_ATTRIBUTE ('test_lex', 'INDEX_STEMS', 'ENGLISH');
END;
drop index temp_idx;
CREATE index temp_idx ON temp(text1) indextype is CTXSYS.CONTEXT parameters ('WORDLIST CTXSYS.BASIC_WORDLIST LEXER test_lex SYNC (ON COMMIT)');
And below is the code I use to cluster the documents:
create table temp0 (docid NUMBER, clusterid NUMBER, score NUMBER);
create table temp1 (clusterid NUMBER, descript varchar2(4000), label varchar2(200), sze number, quality_score number, parent number);
begin
ctx_ddl.drop_preference('my_cluster');
ctx_ddl.create_preference('my_cluster','KMEAN_CLUSTERING');
ctx_ddl.set_attribute('my_cluster','CLUSTER_NUM','10');
ctx_ddl.set_attribute('my_cluster','STEM_ON','FALSE');
ctx_output.start_log('my_log');
ctx_cls.clustering('temp_idx','seq','temp0','temp1','my_cluster');
ctx_output.end_log;
end;
Thanks!

Answer 1

Make the following true
ctx_ddl.set_attribute('my_cluster','STEM_ON','FALSE');
i.e.
ctx_ddl.set_attribute('my_cluster','STEM_ON','TRUE');
and then create the clusters. Also as you have already done, the lexer should have INDEX_STEM on