Webboard: Substring search

Author: Amar Bouchibane
Email: [email protected]
Message:
I don't know if that works as designed or unintentionally:



With an UDM_MATCH_SUBSTR word match records having a subtring as a whole word will get a lower score than records having words containing the substring.



I.e.:

The body section of record1 contains the word "Wochentage".

The body section of record2 contains the word "Woche".



When I start a search query with "Woche" record1 will get a higher score than record2.

The difference is very small (0.0X %) but I would expect that record2 has a slightly significant higher score than record1.



I've done a lot of tests and the same thing happens with other records and with other words/substrings. All strings I used in the tests hadn't synonym words.



Here's the param set I use (only relevant params):



	$res = Udm_Set_Agent_Param($udmAgent, UDM_PARAM_WORD_MATCH, UDM_MATCH_SUBSTR);

	$res = Udm_Set_Agent_Param_Ex($udmAgent, 'SubstringMatchMinWordLength', '2');

	$res = Udm_Set_Agent_Param_Ex($udmAgent, 'WordDensityFactor', 5);

	$res = Udm_Set_Agent_Param_Ex($udmAgent, 'StrictModeThreshold', '1');

	$res = Udm_Set_Agent_Param_Ex($udmAgent, 'IspellUsePrefixes', 'yes');

	$res = Udm_Set_Agent_Param_Ex($udmAgent, 'WordFormFactor', 0);

	$res = Udm_Set_Agent_Param_Ex($udmAgent, 'MinWordLength', 2);

	$res = Udm_Set_Agent_Param_Ex($udmAgent, 'MaxWordLength', 64);

	$res = Udm_Set_Agent_Param_Ex($udmAgent, 'DateFactor', 0);

	$res = Udm_Set_Agent_Param($udmAgent, UDM_PARAM_SEARCH_MODE, UDM_MODE_ALL);

	$res = Udm_Set_Agent_Param_Ex($udmAgent, 'IDFFactor', '40');

	$res = Udm_Set_Agent_Param($udmAgent, UDM_PARAM_DETECT_CLONES, UDM_DISABLED);

	$res = Udm_Set_Agent_Param($udmAgent, UDM_PARAM_PHRASE_MODE, UDM_ENABLED);

	$res = Udm_Set_Agent_Param_Ex($udmAgent, 'PopRankUseShowCnt', 'no');

	$res = Udm_Set_Agent_Param_Ex($udmAgent, 'ExcerptSize', 384);

	$res = Udm_Set_Agent_Param_Ex($udmAgent, 'ExcerptPadding', 20);

	$res = Udm_Set_Agent_Param($udmAgent, UDM_PARAM_GROUPBYSITE, UDM_DISABLED);

	$res = Udm_Set_Agent_Param($udmAgent, UDM_PARAM_MIN_WORD_LEN, 2);

	$res = Udm_Set_Agent_Param($udmAgent, UDM_PARAM_MAX_WORD_LEN, 64);

	$res = Udm_Set_Agent_Param($udmAgent, UDM_PARAM_WEIGHT_FACTOR, "00F8");

	$res = Udm_Set_Agent_Param_Ex($udmAgent, 'nwf', "0000");

	$res = Udm_Set_Agent_Param_Ex($udmAgent, 'MinCoordFactor', 0);

	$res = Udm_Set_Agent_Param_Ex($udmAgent, 'NumWordFactor', 10);

	$res = Udm_Set_Agent_Param_Ex($udmAgent, 'DocSizeWeight', 0);

	$res = Udm_Set_Agent_Param($udmAgent, UDM_PARAM_SORT_ORDER, 'R');

	$res = Udm_Set_Agent_Param_Ex($udmAgent, 'suggest', 'yes');

	$res = Udm_Set_Agent_Param_Ex($udmAgent, 'sp', '1');



WordFormFactor on the other side works great for original/fuzzy word forms.



Reply: <http://www.mnogosearch.org/board/message.php?id=21072>