Paper Review Form

Section I. Overview

A. Reader Interest

1. Which category describes this manuscript?

__Practice/Application/Case Study/Experience Report

_x_Research/Technology

___Survey/Tutorial/How-To

B. Content

1. Please explain how this manuscript advances this field of research and/or contributes something new to the literature.

This paper presents a new method to discover the interesting links in multi-relational data.

1. Is the manuscript technically sound?

___Yes

___Appears to be - but didn't check completely

_x_Partially

___No

C. Presentation

1. Are the title, abstract, and keywords appropriate?

_x_Yes

___No

1. Does the manuscript contain sufficient and appropriate references?

___References are sufficient and appropriate

_x_Important references are missing; more references are needed

___Number of references are excessive

2. Does the introduction state the objectives of the manuscript in terms that encourage the reader to read on?

___Yes

_x_Could be improved

___No

4. How would you rate the organization of the manuscript? Is it focused? Is the length appropriate for the topic?

___Satisfactory

_x_Could be improved

___Poor

5. Please rate and comment on the readability of this manuscript.

___Easy to read

_x_Readable - but requires some effort to understand

___Difficult to read and understand

___Unreadable

Section II. Evaluation

Please rate the manuscript. Explain your choice.

___Award Quality

___Excellent

___Good

_x_Fair

___Poor

Section III. Detailed Comments

The new idea presented in the paper seems interesting since they explain "Unsupervised link discovery" versus traditional knowledge discovery and data mining. They focus on there classes of NLD problems; "Novel path discovery, Novel loop discovery, Significant node discovery".

To deal with novel path discovery, using the rarity to measure interestingness of set of paths and entities is a promising approach. However, to measure the interstingness, just taking "rarity" into consideration as criterion may mislead the discovery. In my opinion, using the semantic web approach, to consider the meaning of the relations rather than just "rarity" will make the discovery more sound and trustworthy. Thus prevent convicting the innocents.

The method does not rely on any preexisting or learnable pattern information and can detect novel, interesting connections that do not need to be conceived prior to the analysis.

The following citation shows the necessity of using metadata for this approach;

"For example: Is the path “A1 writes P2 and P2 cites P1” more interesting than the path “A1 writes P3, P3 is published in journal J1 and J1 also contains P1”? This query will have different answers for different points of view (spindle, source, target, global fan out), and which view is chosen will depend on the user’s focus."

In brief, this approach should be enhanced with metadata context.