[Gzz-commits] manuscripts/Paper paper.tex

[Janne V. Kujala]

CVSROOT:        /cvsroot/gzz
Module name:    manuscripts
Changes by:     Janne V. Kujala <address@hidden>        03/03/29 06:44:21

Modified files:
        Paper          : paper.tex 

Log message:
        experiment stuff

CVSWeb URLs:
http://savannah.gnu.org/cgi-bin/viewcvs/gzz/manuscripts/Paper/paper.tex.diff?tr1=1.61&tr2=1.62&r1=text&r2=text

Patches:
Index: manuscripts/Paper/paper.tex
diff -u manuscripts/Paper/paper.tex:1.61 manuscripts/Paper/paper.tex:1.62
--- manuscripts/Paper/paper.tex:1.61    Sat Mar 29 06:19:17 2003
+++ manuscripts/Paper/paper.tex Sat Mar 29 06:44:21 2003
@@ -1214,7 +1214,7 @@
 }
 \end{figure}
 
-\subsection{Recognizability and memorizability}
+\subsection{Recognizability experiment}
 
 %In our application visual texture discrimination is not as
 %much of an issue as memorizability and recognizability of
@@ -1227,23 +1227,27 @@
 
 Experiments on black-and-white %(faces,) 
 ink blots and snow crystals
-\cite{goldstein71visualrecognition} show that
+\cite{goldstein71visualrecognition} have shown that
 complex, unfamiliar pictures can be remembered and recognized 
 and that recognition
 performance decreases very little over time.
 
 In order to evaluate the recognizability of our procedurally generated
-textures, 
+textures, we need to have an appropriate comparison point.
+%
+%- need to compare to something
+%- can't generate an infinite supply 
+%  of images easily and there are too much semantics
+Pictures of natural objects would not provide a good comparsion,
+because they cannot be generated in infinite amounts from seed
+numbers and they would esasily yield undesirable semantic associations.
+Thus, we simply use plain solid color backgrounds as a baseline for comparison
+even though the colors of even a small set of randomly chosen colors would 
+most likely not be discriminable.
 
-- need to compare to something
+%even though know solid colors not really recognizable.
 
-- images not good, can't generate infinite supply easily, too much
-  semantics
-
-- thus, use solid colors even though know solid colors not really recognizable.
-
-
-Surely it would be hard to remember, e.g., 100 textures.
+Surely it would be hard to remember a large number of textures, too.
 However, the user does not have to remember all the textures;
 it suffices to learn the textures of the most often used documents
 and distinguish them from the rest:
@@ -1267,25 +1271,33 @@
 Each square represents a document, and the area of each square is scaled
 to its rate of accesses.
 The diagram shows 2000 documents weighted with Zipf's law with exponent 1.1.
-The 15 most important documents account XXX\% of the accesses.
+The 15 most important documents account for 50\% of the accesses.
+% 0.50469672124463749
 }
 \end{figure}
 
-Thus, the experiment measured the recognition of only 15 textures,
-and 
+Thus, we chose to measure the recognition of only 15 target textures
+or colors in the experiment. 
+Our hypthesis is that the texture backgrounds are more recognizable
+than the solid colors.
 
-Furthermore, because the texture appearance has no correlation
+%and 
+%
+
+XXX: Furthermore, because the texture appearance has no correlation
 with the document content, 
 the textures of any two important files are similar only by chance.
 
-We have conducted an experiment in a similar setting comparing
-the recognition performance of our textures with solid color backgrounds.
-
+%We have conducted an experiment in a similar setting comparing
+%the recognition performance of our textures with solid color backgrounds.
 
+\subsubsection{Method}
 
-Five participants performed the experiment in both conditions
-(in random order). 
+\emph{Participants.}
+Five participants naïve to the hypothesis and our texturing work
+performed the experiment in both conditions.
 
+\emph{Design and materials.}
 15 target backgrounds and 15 distractor backgrounds were randomly
 chosen for both conditions.
 The distribution of the solid colors was the same that is used for 
@@ -1293,13 +1305,18 @@
 lighness tail was de-emphasized to increase the otherwise low
 discriminability of very light, unsaturated colors.
 
+\emph{Procedure.}
+Each participant performed the test individually in both conditions
+in a random order.
 First, the 15 target backgrounds were shown sequentially, 
 in a random order, 5 seconds each.
 Then, recognition was tested by showing the 15 target backgrounds
-and 15 distractor backgrounds in a random order and 
+and the 15 distractor backgrounds in a random order and 
 having the participant answer ``old'' or ``new'' for each one. 
 The time for answers was not limited.
 
+\subsubsection{Results}
+
 The results are summarized in Table~\ref{tableresults} 
 
 \begin{table}
@@ -1330,16 +1347,22 @@
 %use mutual information or something like that. In colors, the subject
 %was basically guessing!!!
 
-% RT less for correct responses: [F(1,4) = 5.2, p = .082]
 Thus, the recognition performance of the textures was good while
 the solid colors do not have enough variation for unambiguous 
-recognition (without additional cues).
-The observed difference in recognition performance is statistically 
+recognition.
+A repeated measures analysis of variance indicates that
+the observed difference in the recognition performance is statistically 
 significant [F(1,4) = 19.0, p = .012].
+% RT less for correct responses: [F(1,4) = 5.2, p = .082]
 
 %Our experience shows that at least the most recurring textures 
 %are not only recognized but can 
 %also easily be associated with the document content.
+
+The solid colors could be more recognizable with additional cues,
+such as the text of the document, but 
+
+- fragments and not title page etc...
 
 XXX: refs?