7. 2.1 一人目
• 以下は査読メールのコピー
• 査読は以下のように、要素ごとに様々な評価をしてくれる
• OVERALL EVALUATION は要素ごとのSum ではない
----------------------- REVIEW 1 ---------------------
PAPER: 43
TITLE: Fully Automated Cyclic Planning for Large-Scale Manufacturing D
AUTHORS: (anonymous)
OVERALL EVALUATION: 0 (borderline paper)
Novelty of problem addressed and/or solution proposed: 4 (good)
Technical and/or experimental quality: 3 (fair)
Breadth of interest to academics and/or practitioners: 3 (fair)
Really nice work, one of the more exciting papers I’ve reviewed
this year (will be considered for recognition): 1 (No)
Suitable for the system demonstration session?: 2 (Yes or maybe)
7
9. 2.2.1 評価
その後、評価、問題点の指摘、質問が来る
. . . the result is a negative one – SCP (the competition)
vs. ACP except for Cell-Assembly 3a N=64 and Cell-
Assembly 3c N=64. And even then, SCP could be
easily improved with simple local search to remove the
slack between batches. The second more important issue
is that I think the paper is not very well-written.
About two pages is spent explaining type-inference. . .
Cutting this out would allow the authors to . . . termi-
nology is too loosely used . . . (略)
わかりにくい文章であったので文句は言えませんが、しかし
こういった重大な誤解が低い点数(0点=borderline)に結びついて
いる。悪い文章は読む人をイラツカせる上、無用の誤解を招く。
査読者も査読者で忙しく、なるべく読む時間を掛けたくない。
9
10. 2.3 二人目
----------------------- REVIEW 2 ---------------------
PAPER: 43
TITLE: Fully Automated Cyclic Planning for Large-Scale Manufacturing D
AUTHORS: (anonymous)
OVERALL EVALUATION: 2 (accept)
Novelty of problem addressed and/or solution proposed: 4 (good)
Technical and/or experimental quality: 4 (good)
Breadth of interest to academics and/or practitioners: 5 (excellent)
Really nice work, one of the more exciting papers I’ve reviewed this y
Suitable for the system demonstration session?: 2 (Yes or maybe)
10
11. 2.3.1 親切なコメント
Comments:
• I believe that a notation for the . . . would be useful. . . .
• There are four different notations . . . I believe they all refer to
the same process, but it is not clear.
• (Question 1) There is one thing I do not understand about . . . .
• typo paramters (two times) in PDDL code
一般に、 まともな学会のまともな査読 は、typoや文章構成を
含め細かく指摘してくれる。査読を受けることは、 非常にあり
がたい貴重なもの (これは、他の査読者も同様)
11
14. 2.6 Rebuttal の書き方
Review通知メールに書いてある、「反論におけるルール」
The most effective author responses are those that respond
to specific points raised in the reviews and that
focus on clarifying specific points of misunderstand-
ing.
• 質問に答える。
• 誤解を解く。
• 余計なことをしない。
– 実際、500語では、無駄なことを書くスペースはない
14
15. 2.7 返信の重要性をクラス分け
Comments:
• . . .
• (Question 1) There is one thing I do not un-
derstand about . . . .
• (Question 2) ”enumerate all such possible pairs
” - it looks to me like . . . .
• . . .
明確に「質問」をしている。答えなくてはいけない
15
19. 3.1 Review 1
----------------------- REVIEW 1 ---------------------
PAPER: 1
TITLE: Applying Problem Decomposition to Extremely Large Planning Doma
AUTHORS: Masataro Asai and Alex Fukunaga
OVERALL EVALUATION: 3 (strong accept)
REVIEWER’S CONFIDENCE: 4 (high)
Relevance to KEPS: 5 (excellent)
Technical quality (are the claims sound and well supported?): 4 (good)
Clarity (is the paper easy to read?): 4 (good)
Novelty (is the presented idea novel?): 4 (good)
----------- REVIEW -----------
... The paper is generally well written. ...
けっこう嬉しみがヤバイ
19
20. 3.2 うれしみがヤバイのはいいけども
----------- REVIEW -----------
This is very interesting work looking at the automatic decomposition o
planning problems. Built on top of previous work such as component abs
cyclic planning, the contributions are solid, the scale of the test pr
one of the highlights. The work is well motivated and very relevant to
workshop. The paper is generally well written. I noticed a few typos,
method that try to reduce" (pp 1), "complete[Bylander, 1994] ," (pp 1
blank after complete, and remove blank before comma), and "tend to req
(pp 5).
• やはりレビューの量が査読よりかなり短い。
– 論文の主張している内容自体が少ないから
∗ 指摘できるところも少ない.
∗ (性能評価を示していない)
– ”light” な review
20
21. 3.3 Review 2
----------------------- REVIEW 2 ---------------------
PAPER: 1
TITLE: Applying Problem Decomposition to Extremely Large Planning Doma
AUTHORS: Masataro Asai and Alex Fukunaga
OVERALL EVALUATION: 2 (accept)
REVIEWER’S CONFIDENCE: 5 (expert)
Relevance to KEPS: 4 (good)
Technical quality (are the claims sound and well supported?): 4 (good)
Clarity (is the paper easy to read?): 4 (good)
Novelty (is the presented idea novel?): 4 (good)
21
22. 3.3.1 長い返信
———– REVIEW ———–
This paper describes. . . (要約)
The type of planning knowledge acquired by the techniques presented
in the paper is already used by HTN planners, most notable perhaps in
SHOP2 in the IPC-2002. . . . (40行ほど)
HTNプランナ系などの他のアイディアとの関連、HTNの自動
導出、などを強く示していただいた。
22
23. 3.4 Review 3
この方のレビューも興味深かった。
———– REVIEW ———–
This paper presents . . . (要約)
I like the paper, but have two questions that like to see answered in the
final version of the paper if it gets accepted:
• According to authors, the decomposition is good for the type of
problems such as cell-assembly, but . . . (ドメイン依存性はどう
なのか?)
• In the same line as above, what about the use of these techniques in
planning domains where the ’planning component’ is really hard. . .
(同様)
In conclusion, I think the paper can be accepted for KEPS.
23