SD-Rule Transaction Details: 2016 to 2017

The detail transaction of SD-Rules are described as below:

• The following table shows the transcation on the 11 new propsoed SD-Rules in 2017.

  Computer Generated SD-Rules
  ID	Proposed New Rule	Source	Results	Rank & Rule 2016	Rank & Rule 2017	Type	Count Change	Accu. Count
  01-CG1	sation$|noun|zed$|adj	nomD	Good	None	5: sation$|noun|zed$|adj	New in 2017	+1	83
  02-CG2	ed$|adj|zation$|noun	nomD	Good	None	6: ed$|adj|zation$|noun	New in 2017	+1	84
  03-CG3	$|noun|tous$|adj	orgD	Good	None	28: $|noun|tous$|adj	New in 2017	+1	85
  04-CG4	$|noun|y$|noun	orgD	Good	39: graph$|noun|graphy$	48: h$|noun|hy$|noun	Parent-1-Child	+0	85
  05-CG5	sity$|noun|us$|adj	nomD	Good	53: osity$|noun|ous$|adj	58: osity$|noun|ous$|adj	Parent-1-Child	+0	85
  06-CG6	e$|verb|tion$|noun	nomD	Good	None	59: e$|verb|tion$|noun	New in 2017	+1	86
  07-CG7	ous$|adj|y$|noun	nomD	Good	68: ous$|adj|y$|noun	72: ous$|adj|y$|noun	Duplicate	+0	86
  08-CG8	$|noun|ish$|adj	orgD	Bad	None	88: $|noun|ish$|adj	New in 2017	+0	86
  09-CG9	$|noun|fully$|adv	orgD	Bad	None	118: $|noun|fully$|adv	New in 2017	+0	86
  Expert-Suggested SD-Rules
  10-ES1	er$|noun|ress$|noun	Experts	Bad	None	95: er$|noun|ress$|noun	New in 2017	+0	86
  11-ES2	$|noun|ty$|adj	Experts	Bad	None	97: $|noun|ty$|adj	New in 2017	+0	86

  All 82 good SD-Rules in 2016 are evaluated as good rules in 2017. They could be identical, or replaced by the parent-rules or child-rules.

• Good SD-Rules count in Optimal Set:
  • 2016 has 82 good rules while 2017 has 86 good rules in optimate set:
  • From the evaluation, 6 of 11 new rules are good (4 bad; 1 duplicated). Why is the total number of good SD-Rule only increased by 4 (from 82 to 86), not 88 (82 + 6)? It is because:
    • 2 of good new rules are the parent-rule of 1 existing rules (+0).
    • 4 new rules have no parent-child relationshion with existing rule (+4)

    • So, tolal change is 6-2 = 4.

• Good Rules comparison (2016-2017):

  Type	2016	2017	Details
  No Change	80	80	...
  Good Rule turn bad	0	0	N/A
  Parent-1-Child	2	2	2016 2017 39: graph$|noun|graphy$ 48: h$|noun|hy$|noun 53: osity$|noun|ous$|adj 58: sity$|noun|us$|adj
  New in 2017	0	4	5: sation$|noun|zed$|adj 6: ed$|adj|zation$|noun 28: $|noun|tous$|adj 59: e$|verb|tion$|noun
  Total	82	86

• In our process, we only analyze parent-child hierachy for those SD-Rules has parent-child relationship co-exist in the collected set because it is very expensive (time comsuming) to evaluate all parent-child rules. Shoule we modify the processes as:
  • Normalize all SD-Rules to it's root-parent-rule.
  • Analyze parent-child-hieracy for all SD-Rules.
  
  in 2017, we spent 2 weeks to evaluated 17 parents rules. If we modify to this process, there will be 101 parents rules, very expensive!!

The conclusion is the optimized set of SD-Rules is very steady as we expected. We believe this is one of the component that implies that Lexicon is a good representative subset of general English.