Meeting with supervisor
After meeting with supervisor, Dr Yuen, Geoffrey and Wah shows their view points on my project system.
Firstly, Wah ask what if the URL share among the users are identical. For example, if User A save hk.yahoo.com as a Portal site and User B save hk.yahoo.com as a search engine, the URL will be duplicated in the database. Well, this is not a problem becuase, say in the YouTube system, when we search "Frilled Shark", more than 4 same vedio chips will be returned and, it indicated that the searching result is accurate. Dr Yuen suggested that one of the development approach in my stage is that, I can enhance the accuracy on the searching data among the various items in the system.
Afterward, Dr Yuen asked the selling point of my project, well I can hardly tell the different, I response by saying "Well, by comparing the one man work, I do better than them as a team :p" Seriously, the furture mile stoneof my project probably support the different character, espeacially the Asian fonts because, for example, the FURL.net have problem on saving the Traditional Chinese character, I think I will/need to trigger this problem.
Dr Yuen also suggested me to integrate the two algorithms into one and implementation this 'new' algorithm on doing associate analysis. These two algorithm, Apriori Algorithm and FP-Tree, have different approach.
Apriori Algorithm is base on counting the itemsets' frequency. If the frequent of itemset is low, elimnate that item(s) and the remaining enter the next stage. Finally, the resulting item(s) is/are the most likely be related among the other tags.
FP-Tree,on the other hand, is based on classifying the frequency of the tags used into the hush table. The most frequent tags will get the higher position in the hush table and lower frequent items list under them. Under this method, the completeness of the association pattern maintains, and the computational power required is lower than Apriori method.
Lastly, Dr Yuen suggested that the system should utliminately make LESS involvement, like 1. totally no input of tags by user and 2. eliminate the unneccessary user interaction on the site but the newly added data can embed to the system automatically. I think the first one can be achieved by combining the AJAX and Association Pattern suggestion like the one did by Del.icio.us. I think this function will be implement in Phase 3
Firstly, Wah ask what if the URL share among the users are identical. For example, if User A save hk.yahoo.com as a Portal site and User B save hk.yahoo.com as a search engine, the URL will be duplicated in the database. Well, this is not a problem becuase, say in the YouTube system, when we search "Frilled Shark", more than 4 same vedio chips will be returned and, it indicated that the searching result is accurate. Dr Yuen suggested that one of the development approach in my stage is that, I can enhance the accuracy on the searching data among the various items in the system.
Afterward, Dr Yuen asked the selling point of my project, well I can hardly tell the different, I response by saying "Well, by comparing the one man work, I do better than them as a team :p" Seriously, the furture mile stoneof my project probably support the different character, espeacially the Asian fonts because, for example, the FURL.net have problem on saving the Traditional Chinese character, I think I will/need to trigger this problem.
Dr Yuen also suggested me to integrate the two algorithms into one and implementation this 'new' algorithm on doing associate analysis. These two algorithm, Apriori Algorithm and FP-Tree, have different approach.
Apriori Algorithm is base on counting the itemsets' frequency. If the frequent of itemset is low, elimnate that item(s) and the remaining enter the next stage. Finally, the resulting item(s) is/are the most likely be related among the other tags.
FP-Tree,on the other hand, is based on classifying the frequency of the tags used into the hush table. The most frequent tags will get the higher position in the hush table and lower frequent items list under them. Under this method, the completeness of the association pattern maintains, and the computational power required is lower than Apriori method.
Lastly, Dr Yuen suggested that the system should utliminately make LESS involvement, like 1. totally no input of tags by user and 2. eliminate the unneccessary user interaction on the site but the newly added data can embed to the system automatically. I think the first one can be achieved by combining the AJAX and Association Pattern suggestion like the one did by Del.icio.us. I think this function will be implement in Phase 3
留言
發佈留言