Details
| Originalsprache | Englisch |
|---|---|
| Titel des Sammelwerks | Research and Advanced Technology for Digital Libraries |
| Untertitel | 21st International Conference on Theory and Practice of Digital Libraries |
| Herausgeber/-innen | Yannis Manolopoulos, Jaap Kamps, Giannis Tsakonas, Lazaros Iliadis, Ioannis Karydis |
| Herausgeber (Verlag) | Springer Verlag |
| Seiten | 369-381 |
| Seitenumfang | 13 |
| ISBN (elektronisch) | 9783319670089 |
| ISBN (Print) | 9783319670072 |
| Publikationsstatus | Veröffentlicht - 2 Sept. 2017 |
| Veranstaltung | 21st International Conference on Theory and Practice of Digital Libraries, TPDL 2017 - Thessaloniki, Griechenland Dauer: 18 Sept. 2017 → 21 Sept. 2017 |
Publikationsreihe
| Name | Lecture Notes in Computer Science |
|---|---|
| ISSN (Print) | 0302-9743 |
| ISSN (elektronisch) | 1611-3349 |
Abstract
Opinionated data streams are very popular data paradigms nowadays as more and more users share their opinions online about almost everything from products to persons, brands and ideas. One of the key challenges for opinionated stream mining is dealing with concept drifts in the underlying stream population by building learners that adapt to such concept changes. Ageing is a typical way of adapting to change in a stream environment as it potentially allows us to discard outdated information from the learning models and focus on the most recent information. Most of the existing approaches follow a fixed ageing strategy which remains the same over the whole stream; for example, a fixed window size in the sliding window model or a fixed ageing factor in the damped window model. This implies that we forget at the same rate over the whole course of the stream, which is counterintuitive given the volatile nature of the stream. What is more intuitive is to forget faster in times of change so as to adapt to new data and to forget slower, or in other words, to remember more, in times of stability. In this work, we propose an informative-adaptation-to-change approach where we first detect changes in the underlying data stream and then we tune the ageing factor of the ageing-based Multinomial Naive Bayes (MNB) classifier based on the detected change. Except for the up-to-date classifier our method also outputs the points of change in the stream, therefore offering more insights to the final users.
ASJC Scopus Sachgebiete
- Mathematik (insg.)
- Theoretische Informatik
- Informatik (insg.)
Zitieren
- Standard
- Harvard
- Apa
- Vancouver
- BibTex
- RIS
Research and Advanced Technology for Digital Libraries: 21st International Conference on Theory and Practice of Digital Libraries. Hrsg. / Yannis Manolopoulos; Jaap Kamps; Giannis Tsakonas; Lazaros Iliadis; Ioannis Karydis. Springer Verlag, 2017. S. 369-381 (Lecture Notes in Computer Science).
Publikation: Beitrag in Buch/Bericht/Sammelwerk/Konferenzband › Aufsatz in Konferenzband › Forschung › Peer-Review
}
TY - GEN
T1 - Sentiment Classification over Opinionated Data Streams Through Informed Model Adaptation
AU - Iosifidis, Vasileios
AU - Oelschlager, Annina
AU - Ntoutsi, Eirini
N1 - Funding information:. The work was partially funded by the European Commission for the ERC Advanced Grant ALEXANDRIA under grant No. 339233 and by the German Research Foundation (DFG) project OSCAR (Opinion Stream Classification with Ensembles and Active leaRners).
PY - 2017/9/2
Y1 - 2017/9/2
N2 - Opinionated data streams are very popular data paradigms nowadays as more and more users share their opinions online about almost everything from products to persons, brands and ideas. One of the key challenges for opinionated stream mining is dealing with concept drifts in the underlying stream population by building learners that adapt to such concept changes. Ageing is a typical way of adapting to change in a stream environment as it potentially allows us to discard outdated information from the learning models and focus on the most recent information. Most of the existing approaches follow a fixed ageing strategy which remains the same over the whole stream; for example, a fixed window size in the sliding window model or a fixed ageing factor in the damped window model. This implies that we forget at the same rate over the whole course of the stream, which is counterintuitive given the volatile nature of the stream. What is more intuitive is to forget faster in times of change so as to adapt to new data and to forget slower, or in other words, to remember more, in times of stability. In this work, we propose an informative-adaptation-to-change approach where we first detect changes in the underlying data stream and then we tune the ageing factor of the ageing-based Multinomial Naive Bayes (MNB) classifier based on the detected change. Except for the up-to-date classifier our method also outputs the points of change in the stream, therefore offering more insights to the final users.
AB - Opinionated data streams are very popular data paradigms nowadays as more and more users share their opinions online about almost everything from products to persons, brands and ideas. One of the key challenges for opinionated stream mining is dealing with concept drifts in the underlying stream population by building learners that adapt to such concept changes. Ageing is a typical way of adapting to change in a stream environment as it potentially allows us to discard outdated information from the learning models and focus on the most recent information. Most of the existing approaches follow a fixed ageing strategy which remains the same over the whole stream; for example, a fixed window size in the sliding window model or a fixed ageing factor in the damped window model. This implies that we forget at the same rate over the whole course of the stream, which is counterintuitive given the volatile nature of the stream. What is more intuitive is to forget faster in times of change so as to adapt to new data and to forget slower, or in other words, to remember more, in times of stability. In this work, we propose an informative-adaptation-to-change approach where we first detect changes in the underlying data stream and then we tune the ageing factor of the ageing-based Multinomial Naive Bayes (MNB) classifier based on the detected change. Except for the up-to-date classifier our method also outputs the points of change in the stream, therefore offering more insights to the final users.
UR - http://www.scopus.com/inward/record.url?scp=85029591714&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-67008-9_29
DO - 10.1007/978-3-319-67008-9_29
M3 - Conference contribution
AN - SCOPUS:85029591714
SN - 9783319670072
T3 - Lecture Notes in Computer Science
SP - 369
EP - 381
BT - Research and Advanced Technology for Digital Libraries
A2 - Manolopoulos, Yannis
A2 - Kamps, Jaap
A2 - Tsakonas, Giannis
A2 - Iliadis, Lazaros
A2 - Karydis, Ioannis
PB - Springer Verlag
T2 - 21st International Conference on Theory and Practice of Digital Libraries, TPDL 2017
Y2 - 18 September 2017 through 21 September 2017
ER -