User:Francis Tyers/MT - Revision history

Francis Tyers at 17:39, 8 June 2008

2008-06-08T17:39:25Z

Francis Tyers at 21:55, 7 June 2008

2008-06-07T21:55:32Z

Francis Tyers at 11:48, 7 June 2008

2008-06-07T11:48:30Z

Francis Tyers: /* Review */

2008-04-09T06:29:10Z

Review

Francis Tyers: /* Review */

2008-04-08T12:21:06Z

Review

Francis Tyers: /* Review */

2008-04-08T09:56:04Z

Review

Francis Tyers: /* Review */

2008-04-08T09:40:24Z

Review

Francis Tyers: /* Review */

2008-04-05T19:00:55Z

Review

Francis Tyers: /* Review */

2008-04-05T18:22:45Z

Review

Francis Tyers: /* Review */

2008-04-05T17:40:28Z

Review

@@ Line 1: / Line 1: @@
 Schubert, E., Schaffert, S., and Bry, F. (2005). "Structure-preserving difference search for xml documents". ''Proceedings of the Extreme Markup Languages 2005 Conference, 1-5 August 2005, Montréal, Quebec, Canada''
+==Summary==
 ;Structure-Preserving Difference Search for XML Documents
@@ Line 24: / Line 26: @@
 sub-structures.
-The algorithm used is Dijkstra's algorithm, where the edge weights are calculated by different cost
+The algorithm used is Dijkstra's algorithm, where the edge weights of the search graph are calculated by
+different cost functions, where each node is a nodeset correspondence and each edge is a transition which
-functions. The authors outline four possible cost functions.
+adds adds a new mapping to the nodeset correspondene. The authors outline four possible cost functions. The
+first and most basic simply takes into account the number of sub-structures not retained in a mapping between
+nodes. Others provide for credits for unavoidably dropped mappings and calculation of incomplete mappings.
+The nodeset correspondence is considered maximal if there are no other nodeset correspondences which contain
+mappings between more nodes.
+The authors provide some details on experimental results in the form of comparative examples from both their
+prototype and from Logilab xmldiff. They comment that presenting detailed results is difficult as it would
+involve the inclusion of long XML documents, impractical in such a paper. They state that the output from their
+prototype is more likely to reflect changes made by humans than xmldiff.
+They conclude the paper by giving some prospects for further research and development, including methods of improving
+the efficiency of the algorithm (search tree pruning and divide and conquer strategies) and taking advantage of
+other available information (attributes and DTDs).
 ==Review==
@@ Line 58: / Line 74: @@
 Comments on Paper:
-The introduction clearly lays out the problem. The state of the art section is fairly well written, with a couple of
+The introduction clearly lays out the problem. The state of the art section is well written, with a couple of
 minor quibbles that should be taken care of. The explanation of the methodology is adequate.
+The description of the data structures used (document graphs) and the quality measures (retained relations, node
+similarity etc.) is good except the part describing the search graph, which could do with some clarification. Perhaps
+it would be worth describing the "basic cost function" before describing the search graph and then outlining the remaining
+cost functions afterwards.
 In the results section I find that the paper focusses too much on the formats of the output rather than the actual
 character of the output itself.
 Regarding formatting, the style is less than optimal, equations are typeset too small, code fragments overlap the
@@ Line 153: / Line 172: @@
 * p.8  Comments about Dijkstra's algorithm, e.g. named, described, taught etc. are probably superfluous.
-* p.9  It isn't entirely clear how the search graph is constructed, it may benefit from a diagram, such
+* p.9  It isn't entirely clear how the search graph is constructed, it would benefit from a diagram, such
        as those describing "retained relations" and "maximal nodeset correspondences".

@@ Line 155: / Line 155: @@
 * p.9  It isn't entirely clear how the search graph is constructed, it may benefit from a diagram, such
        as those describing "retained relations" and "maximal nodeset correspondences".
+* p.9  "credits" data → "credit" data
 * p.10 The sentence starting "Informally..." is too long, split it into two, probably around the "and".

@@ Line 12: / Line 12: @@
 The authors present their method of "structure preserving difference", which instead of trying to find
-the smallest possible edit script, attempts to maximise the size of retained sub-structures maintained in
+the smallest possible edit script, attempts to maximise the size of retained sub-structures (that is
-changing one document to another. In order to calculate this, they model the document as a graph,
+sub-trees of the XML document tree) maintained in changing one document to another. In order to calculate
-where relations other than simple parent-child can be taken into account, for example ancestor-descendent,
+this, they model the document as a graph, where relations other than simple parent-child can be taken into
-and sibling relationships.
+account, for example ancestor-descendent, and sibling relationships.
-Retained sub-structures are calculated as nodeset correspondences (set of mappings between equal nodes)
+Retained sub-structures are calculated as nodeset correspondences (set of mappings between equivalent nodes)
 between the original, and modified document. Two nodes are considered equal when they match a node
 similarity relation, which is user-specified. The most simple node similarity relation would be to consider
@@ Line 152: / Line 152: @@
 * p.8  Comments about Dijkstra's algorithm, e.g. named, described, taught etc. are probably superfluous.
+* p.9  It isn't entirely clear how the search graph is constructed, it may benefit from a diagram, such
+       as those describing "retained relations" and "maximal nodeset correspondences".
 * p.10 The sentence starting "Informally..." is too long, split it into two, probably around the "and".
@@ Line 210: / Line 213: @@
 * p.13 ISBN is an acronym, should it not be in upper case? The "N" stands for "Number", so "ISBN number"
        is redundant.
+* p.14 A link should be given to libxml, and the version required specified.
 * p.14 The xmldiff output looks considerably easier to read than the XUpdate syntax, as the examples are

@@ Line 126: / Line 126: @@
        → "us to model" or "the modelling of"
+* p.6  In the equation S1 ⊂ ~, '~' is undefined. The next page defines it, but this definition should be
+       given before the usage, or at least on the same page.
 * p.7  The <em> and <i> tags are not as far as I know semantically equal. They belong to different groups of

← Older revision		Revision as of 12:21, 8 April 2008
Line 63:		Line 63:

	In the results section I find that the paper focusses too much on the formats of the output rather than the actual		In the results section I find that the paper focusses too much on the formats of the output rather than the actual
	character of the output itself.		character of the output itself.

@@ Line 205: / Line 205: @@
 * p.13 Remove exclamation mark, for emphasis, use bold or italic face.
-* p.13 ISBN is an acronym, should it not be in upper case, the "N" stands for "Number", so "ISBN number"
+* p.13 ISBN is an acronym, should it not be in upper case? The "N" stands for "Number", so "ISBN number"
        is redundant.
@@ Line 214: / Line 214: @@
 * p.16 It might be interesting to note that Figure 16 is quite similar to what would be expected from a
        traditional "diff" using diffutils.
 </pre>

@@ Line 61: / Line 61: @@
 minor quibbles that should be taken care of. The explanation of the methodology is adequate.
+In the results section I find that the paper focusses too much on the formats of the output rather than the actual
+character of the output itself.
@@ Line 156: / Line 159: @@
        → credit value ...
+* p.12 The part about installation is a bit orphaned and not particularly relevant. It could be removed.
+* p.12 Does "optimal result" here refer to "shortest edit distance", if so this should be replaced, if
+       not it should be explained. Change "it (now) uses an" to "recently it has begun to use an" for
+       ease of reading.
 * p.12 It would be nice with the first example to display the output of GNU diff for comparison.
@@ Line 193: / Line 202: @@
        ----
+* p.13 Remove exclamation mark, for emphasis, use bold or italic face.
+* p.13 ISBN is an acronym, should it not be in upper case, the "N" stands for "Number", so "ISBN number"
+       is redundant.
+* p.14 The xmldiff output looks considerably easier to read than the XUpdate syntax, as the examples are
+       contrived anyway, it might be better to convert the XUpdate syntax to the logilab syntax for
+       comparative purposes. This would probably serve better than "human readable" descriptions.
+* p.16 It might be interesting to note that Figure 16 is quite similar to what would be expected from a
+       traditional "diff" using diffutils.

← Older revision		Revision as of 19:00, 5 April 2008
Line 193:		Line 193:

	----		----


	</pre>		</pre>

@@ Line 59: / Line 59: @@
 The introduction clearly lays out the problem. The state of the art section is fairly well written, with a couple of
-minor quibbles that should be taken care of. The
+minor quibbles that should be taken care of. The explanation of the methodology is adequate.
 Regarding formatting, the style is less than optimal, equations are typeset too small, code fragments overlap the
@@ Line 155: / Line 157: @@
        → credit value ...
+* p.12 It would be nice with the first example to display the output of GNU diff for comparison.
-*
+       ----
+       $ diff -uw /tmp/1~ /tmp/2~
+       --- /tmp/1~      2008-04-05 20:30:14.000000000 +0100
+       +++ /tmp/2~      2008-04-05 20:30:10.000000000 +0100
+       @@ -11,10 +11,10 @@
+            <title2>
+              and the Order of the Phoenix
+             </title2>
+       -    <edition>Hardcover</edition>
+       +    <edition>Paperback</edition>
+            <author>J. K. Rowling</author>
+       -    <price>19.79</price>
+       -    <isbn>043935806X</isbn>
+       +    <price>8.49</price>
+       +    <isbn>0439358078</isbn>
+          </book>
+          <book>
+            <rank>2</rank>
+       @@ -24,9 +24,9 @@
+            <title2>
+              and the Order of the Phoenix
+            </title2>
+       -    <edition>Paperback</edition>
+       +    <edition>Hardcover</edition>
+            <author>J. K. Rowling</author>
+       -    <price>8.99</price>
+       -    <isbn>0439358078</isbn>
+       +    <price>17.99</price>
+       +    <isbn>043935806X</isbn>
+          </book>
+        </charts>
+       ----
 </pre>

@@ Line 151: / Line 151: @@
+* p.11 ... credits value ...
+       → credit value ...
+*