Journal Article Event.Locky: System of Event-Data Extraction from Webpages based on Web Mining
Event.Locky: System of Event-Data Extraction from Webpages based on Web Mining

Chenyi, Liao  ,  Kei, Hiroi  ,  Katsuhiko, Kaji  ,  Ken, Sakurada  ,  Nobuo, Kawaguchi

58 ( 4 ) 2017-04-15
ISSN:1882-7764
NCID:AN00116647
Description
Nearby event data, such as those for exhibitions and sales promotions, may help users spend their free time more efficiently. However, most event data are hidden in millions of webpages, which is very time-consuming for a user to find such data. To address this issue, we use web mining that extracts event data from webpages. In this paper, we propose and discuss the implementation of Event.Locky - a system for extracting event data from webpages in a user-defined area and displaying them to a user in a spatial-temporal structure. Furthermore, we design two core algorithms for event data extraction in Event.Locky: webpage-data-record extraction and event-record classification. The former is used to convert a semi-structural HTML document into processable structured data. The latter filters out non-event data from extracted data records using machine learning. We trained and evaluated Event.Locky with an actual dataset composed by 96 restaurants and shops at Nagoya train station. As a result, our event-classification algorithm achieved an F1 score of 91.61%, an increase of 3.07% from current event-classification algorithms. The combination of our event-classification algorithm and our data-record-extraction algorithm achieved F1 score 83.96% to extract event records from webpages. That increased 1.6% from current algorithm. Finally, we discuss the feasibility of Event.Locky in an actual online environment through the implementation of a demonstration application.------------------------------This is a preprint of an article intended for publication Journal ofInformation Processing(JIP). This preprint should not be cited. Thisarticle should be cited as: Journal of Information Processing Vol.25(2017) (online)DOI http://dx.doi.org/10.2197/ipsjjip.25.321------------------------------
Nearby event data, such as those for exhibitions and sales promotions, may help users spend their free time more efficiently. However, most event data are hidden in millions of webpages, which is very time-consuming for a user to find such data. To address this issue, we use web mining that extracts event data from webpages. In this paper, we propose and discuss the implementation of Event.Locky - a system for extracting event data from webpages in a user-defined area and displaying them to a user in a spatial-temporal structure. Furthermore, we design two core algorithms for event data extraction in Event.Locky: webpage-data-record extraction and event-record classification. The former is used to convert a semi-structural HTML document into processable structured data. The latter filters out non-event data from extracted data records using machine learning. We trained and evaluated Event.Locky with an actual dataset composed by 96 restaurants and shops at Nagoya train station. As a result, our event-classification algorithm achieved an F1 score of 91.61%, an increase of 3.07% from current event-classification algorithms. The combination of our event-classification algorithm and our data-record-extraction algorithm achieved F1 score 83.96% to extract event records from webpages. That increased 1.6% from current algorithm. Finally, we discuss the feasibility of Event.Locky in an actual online environment through the implementation of a demonstration application.------------------------------This is a preprint of an article intended for publication Journal ofInformation Processing(JIP). This preprint should not be cited. Thisarticle should be cited as: Journal of Information Processing Vol.25(2017) (online)DOI http://dx.doi.org/10.2197/ipsjjip.25.321------------------------------
Full-Text

https://ipsj.ixsq.nii.ac.jp/ej/?action=repository_action_common_download&item_id=178666&item_no=1&attribute_id=1&file_no=1

Number of accesses :  

Other information