Lexical Tools

Result of Proposed Rules

I. Results

15 SD-Rules are added to the SD-Rule for evaluation. 12 of them are evaluated as good rules in the optimized set and 4 are bad rules. 2 of these bad rules are duplicated child rules. Also, in the optimized set, 2 child rules are used to replace proposed rules

SD-Rule	Rank	Precision	Instances	Source	Results
Good Rules
`se$\|verb\|zation$\|noun`	2	100.00%	1108	NOM_D	Good SD-Rule
`sation$\|noun\|ze$\|verb`	3	100.00%	1071	NOM_D	Good SD-Rule
`ility$\|noun\|le$\|adj`	9	99.94%	1626	NOM_D	Good SD-Rule
`$\|adj\|ally$\|adv`	15	99.08%	2072	ORG_D	Good SD-Rule
`ce$\|noun\|t$\|adj`	18	98.82%	847	NOM_D	Child rule `nce$\|noun\|nt$\|adj` is used
`cy$\|noun\|t$\|adj`	19	98.77%	406	NOM_D	Good SD-Rule
`e$\|verb\|ion$\|noun`	20	98.76%	2336	NOM_D	Good SD-Rule
`c$\|adj\|s$\|noun`	43	91.46%	281	ORG_D	Child rule `ic$\|adj\|is$\|noun` is used
`e$\|verb\|ing$\|noun`	45	91.43%	210	Suggestions	Good SD-Rule
`al$\|adj\|us$\|noun`	61	84.35%	262	Suggestions	Good SD-Rule
`es$\|noun\|ic$\|adj`	67	73.91%	23	Suggestions	Good SD-Rule
Bad Rules
`ian$\|adj\|ia$\|noun`	57	86.31%	263	Suggestions	Duplicated, parent rule `an$\|adj\|a$\|noun` is used
`$\|noun\|ize$\|verb`	78	59.05%	442	Suggestions	Bad SD-Rule
`ian$\|noun\|ia$\|noun`	99	0.36%	274	Suggestions	Duplicated, parent rule `an$\|noun\|a$\|noun` is a bad SD-Rule
`es$\|noun\|ic$\|noun`	101	0.00%	19	Suggestions	Bad SD-Rule

II. Observation

All 8 computer generated SD-Rules from NOM_D and ORG_D are new good rules with high precision and instances (8/8)
Only 3 (out of 7) experts' suggested SD-Rules are added to the new optimized set.
Two of expert's suggested SD-Rule are duplicates of existing child rules:
- One is a child rule of an existing good rule
- One is a child rule of an existing bad rule
The other two of experts suggested rules are bad rules

III. Further Observation on NOM_D

The top 17 SD-Rule generated from NOM_D are all good SD-Rules.

SD-Rule	Rank	Precision	Instances	status
Frequency > 200, Instance coverage > 1.00% , Accum. Coverage > 80.0%
`$\|adj\|ness$\|noun`	1	100.00%	2734	Existing SD-Rule
`se$\|verb\|zation$\|noun`	2	100.00%	1108	New SD-Rule
`sation$\|noun\|ze$\|verb`	3	100.00%	1071	New SD-Rule
`ility$\|noun\|le$\|adj`	9	99.94%	1626	New SD-Rule
`iness$\|noun\|y$\|adj`	10	99.82%	545	Existing SD-Rule
`ation$\|noun\|e$\|verb`	14	99.24%	2514	Existing SD-Rule
`ce$\|noun\|t$\|adj`	18	98.82%	847	New child rule `nce$\|noun\|nt$\|adj` is used
`cy$\|noun\|t$\|adj`	19	98.77%	406	New SD-Rule
`e$\|verb\|ion$\|noun`	20	98.76%	2336	New SD-Rule
`$\|verb\|ment$\|noun`	21	98.35%	607	Existing SD-Rule
`ication$\|noun\|y$\|verb`	22	98.33%	300	Existing SD-Rule
`ed$\|adj\|ion$\|noun`	24	98.26%	689	Existing SD-Rule
`$\|adj\|ity$\|noun`	26	97.88%	2080	Existing SD-Rule
`e$\|adj\|ity$\|noun`	30	97.54%	894	Existing SD-Rule
`$\|verb\|ion$\|noun`	40	93.71%	572	Existing SD-Rule
`$\|verb\|ing$\|noun`	42	92.53%	522	Existing SD-Rule
`$\|verb\|ation$\|noun`	48	90.29%	340	Existing SD-Rule

IV. Further Observation on ORG_D

The top 6 SD-Rules generated from ORG_D are all good SD-Rules

SD-Rule	Rank	Precision	Instances	status
Frequency > 40, Instance coverage > 1.00% , Accum. Coverage > 11.50%
`less$\|adj\|$\|noun`	11	99.64%	561	Existing SD-Rule
`$\|adj\|ally$\|adv`	15	99.08%	2072	New SD-Rule
`$\|verb\|ion$\|noun`	40	93.71%	572	Existing SD-Rule, also derived from NOM_D
`ist$\|noun\|y$\|noun`	36	95.48%	509	Existing SD-Rule
`c$\|adj\|s$\|noun`	43	91.46%	281	New child rule `ic$\|adj\|is$\|noun` is used
`ful$\|adj\|$\|noun`	50	89.21%	139	Existing SD-Rule

V. Future Work

Evaluated more SD-Rules from NOM_D and ORG_D