"... In PAGE 4: ... Differ- ences in the implementations, especially the pre- processing module, result in there being some terms identified by Termer that were not identified by Head Sorting. Table2 shows the number of terms identified by each method. (*Because some terms are identified by more than one technique, the percentage adds up to more than 100%.... ..."

"... In PAGE 24: ...39 and the other set with term frequencies greater than or equal to 0.39, we obtain Table3 . As shown in the table, the p-values of the index term Year in the two different value domains all fall well below 5%.... ..."

"... In PAGE 34: ...STATISTICAL APPENDIX Table1 : Bolivian Terms of Trade Index, 1980-92 (1987=100) Year Export price index hbport price index Ternn oftrade index 1980 181.0 91.... ..."

"... In PAGE 70: ...11 is shown as below: nullnullnullnull1 nullnullnullnull nullnull nullnullnullnullnull null nullnullnullnull nullnullnullnull nullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnull nullnullnullnullnull nullnull nullnullnullnullnull nullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnull nullnullnull nullnullnullnullnullnullnullnull null6.1null nullnullnull 6 null 3null Meanwhile, Table6 -1 shows the list of index terms of the three sofas chosen by customer null . Parameter frequency means the number of times an index term appears in the whole user relevance feedback set (sofa14, sofa8, and sofa46).... In PAGE 71: ...association rules from Rule Group 1, we consider the index terms from Table6 -1 as input X (see Table 5-6). Table 6-2 shows the results from the previous matching process.... In PAGE 71: ...association rules from Rule Group 1, we consider the index terms from Table 6-1 as input X (see Table 5-6). Table6 -2 shows the results from the previous matching process. Parameter frequency indicates the importance of this rule and will be considered the weight.... In PAGE 71: ... (see Table 5-6). Table 6-2 shows the results from the previous matching process. Parameter frequency indicates the importance of this rule and will be considered the weight. Table6 -2: the association rules applicable for the three relevance feedback X Index Terms of X Y Index Terms of Y support frequency Rule Group 1 Sofa Design - Metal sofa_design_4 Chair Design - Metal chair_design_3 0.405797 1 Sofa Color - Dark Neutral sofa_color_2 Chair Color - Dark Neutral chair_color_1 0.... In PAGE 71: ...521739 2 Sofa Type - Casual sofa_type_1 Chair Design - Medium Wood chair_design_2 0.442029 1 After analysis, a set of relevant index terms of chairs (null ) has been generated, as shown in Table6 -2.... In PAGE 72: ...2. The analysis shows that the three sofas selected by this customer null share the same index term sofa_color_2 , and that index term matches one association rule, which shows that in this demographic profile (Level 1-5), people who prefer sofas with attribute sofa_color_2 as X are more likely to purchase chairs with attribute chair_color_1 as Y (43% probability) (see Table6 -2). In order to observe the improvement, we highlight the chairs with attribute chair_color_1 which is Dark Neutral in field Color.... In PAGE 72: ... nullnullnull nullnullnull nullnullnull nullnullnullnull nullnull nullnullnullnullnull The order of product presentation on the chair page for customer null is remarkably changed after personalization according to the preferences in the sofa category. To exemplify how the rankings are different, the original top 15 ranking orders nullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnull (generated solely based on the demographic information of null ) are shown in Table6 -3, with the new null null null nullnullnull nullnullnull nullnull nullnullnullnullnull nullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnull shown in Table 6-4. When this customer null selects one or multiple chairs and move on to the next group of product collection (table), this process will be applied the same manner.... In PAGE 72: ... nullnullnull nullnullnull nullnullnull nullnullnullnull nullnull nullnullnullnullnull The order of product presentation on the chair page for customer null is remarkably changed after personalization according to the preferences in the sofa category. To exemplify how the rankings are different, the original top 15 ranking orders nullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnull (generated solely based on the demographic information of null ) are shown in Table 6-3, with the new null null null nullnullnull nullnullnull nullnull nullnullnullnullnull nullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnull shown in Table6 -4. When this customer null selects one or multiple chairs and move on to the next group of product collection (table), this process will be applied the same manner.... In PAGE 73: ...Table6 -3: The original nullnullnullnullnull nullnull nullnullnullnullnull nullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnull Rank ID Name Price Construction Comfort Color Design Score 1 chair19 Casual_C1_M1 799 3 3 Light Neutral Dark Wood 0.699846 2 chair24 Modern_C1_M3 799 2 3 Light Neutral Metal 0.... In PAGE 73: ...59108 15 chair3 Casual_C2_M3 999 3 3 Medium Neutral Medium Wood 0.582712 Table6 -4: The adapted nullnullnullnull nullnull nullnullnullnullnullnull nullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnull Rank ID Name Price Construction Comfort Color Design Score 1 chair19 Casual_C1_M1 799 3 3 Light Neutral Dark Wood 0.526992 2 chair30 Modern_C2_M3 999 2 3 Dark Neutral Metal 0.... In PAGE 74: ...aking. Using the same example from Section 6.2, because chair_color_1 was found to be one of the critical product attributes, the system automatically reweights this index term and rearranges the presentation order of the chairs. nullnullnull nullnullnull nullnull nullnullnullnullnull By comparing the original ranking and new ranking, listed in Table6 -3 and Table 6-4 respectively, we notice that chairs with color attributes Dark Neutral advance including chair30 moves from top 8 to top 2; chair1 from 9 to 8; chair13 from14 to 4, which even outranks chair1 .... ..."