Using Transduction and Multi-view Learning to Answer Emails
Lecture notes in computer sciencePublished 1 January 2003
Michael Kockelkorn, Andreas Lüneburg, Tobias Scheffer
Citations25
SJR quartileQ2
SJR score0.35
SNIP0.55
Generate an AI Snapshot to get a quick, structured summary of this paper.
Study Snapshot
ObjectiveStudy objective
MethodsResearch methodology
PopulationPopulation studied
Sample sizeSample sizes
OutcomesStudy outcomes here
ResultsStudy results comes here
LimitationsResearch study limitations comes here
A concise AI-generated summary of the paper will appear here once you click Generate AI Snapshot.
TL;DR
This work addresses the problem of predicting which of several frequently used answers a user will choose to respond to an email, and effectively utilizes the data that is typically available in this setting: inbound and outbound emails stored on a server.
Abstract
Many organizations and companies have to answer large amounts of emails. Often, most of these emails contain variations of relatively few frequently asked questions. We address the problem of predicting which of several frequently used answers a user will choose to respond to an email. Our approach effectively utilizes the data that is typically available in this setting: inbound and outbound emails stored on a server. We take into account that there are no explicit links between inbound and corresponding outbound mails on the server. We map the problem to a semi-supervised classification problem that can be addressed by algorithms such as the transductive support vector machine and multi-view learning. We evaluate our approach using emails sent to a corporate customer service department.
Keywords
Computer Science
Pattern RecognitionThe use of the area under the ROC curve in the evaluation of machine learning algorithms
7,233 Citations1997Andrew P. Bradley
AUC exhibits a number of desirable properties when compared to overall accuracy: increased sensitivity in Analysis of Variance (ANOVA) tests; a standard error that decreased as both AUC and the number of test samples increased; decision threshold independent; and it is invariant to a priori class probabilities.
International Conference on Neural Information ProcessingAdvances in kernel methods: support vector learning
5,815 Citations1999Bernhard Schölkopf, Christopher J. C. Burges +1 more
Support vector machines for dynamic reconstruction of a chaotic system, Klaus-Robert Muller et al pairwise classification and support vector machines, Ulrich Kressel.
Combining labeled and unlabeled data with co-training
5,604 Citations1998Avrim Blum, Tom M. Mitchell
Technical reportsMaking Large-Scale SVM Learning Practical
4,317 Citations2006Thorsten Joachims
This chapter presents algorithmic and computational results developed for SVM light V 2.0, which make large-scale SVM training more practical and give guidelines for the application of SVMs to large domains.
Transductive Inference for Text Classification using Support Vector Machines
2,717 Citations1999Thorsten Joachims
An analysis of why TSVMs are well suited for text classi(cid:12)cation is presented, and an algorithm for training TSVMs e(cid:14)-ciently, handling 10,000 examples and more is proposed.
Analyzing the effectiveness and applicability of co-training
1,028 Citations2000Kamal Nigam, Rayid Ghani
It is demonstrated that when learning from labeled and unlabeled data, algorithms explicitly leveraging a natural independent split of the features outperform algorithms that do not and may out-perform algorithms not using a split.
The Case against Accuracy Estimation for Comparing Induction Algorithms
1,024 Citations1998Foster Provost, Tom Fawcett +1 more
This work describes and demonstrates what it believes to be the proper use of ROC analysis for comparative studies in machine learning research, and argues that this methodology is preferable both for making practical choices and for drawing conclusions.
The TREC-8 Question Answering Track Report
857 Citations1999Ellen M. Voorhees
An experimental comparison of naive Bayesian and keyword-based anti-spam filtering with personal e-mail messages
451 Citations2000Ion Androutsopoulos, John Koutsias +2 more
This work introduces appropriate cost-sensitive measures, and investigates at the same time the effect of attribute-set size, training-corpus size, lemmatization, and stop lists, issues that have not been explored in previous experiments.
Active + Semi-supervised Learning = Robust Multi-View Learning
288 Citations2002Ion Muslea, Steven Minton +1 more
A new multi-view algorithm, Co-EMT, which combines semi-supervised and active learning is introduced, which outperforms the other algorithms both on the parameterized problems and on two additional real world domains.
Email classification with co-training
191 Citations2011Svetlana Kiritchenko, Stan Matwin
The main problems in text classification are lack of labeled data, as well as the cost of labeling the unlabeled data, which are addressed by exploring co-training - an algorithm that uses unlabeling data along with a few labeled examples to boost the performance of a classifier.
MailCat
164 Citations1999Richard Segal, Jeffrey O. Kephart
MailCat is an intelligent assistant that heips users organize their e-mail into folders by learning each user’s mail-filing habits and providing shortcut buttons to file each message into one of these three folders.
Concept features in Re:Agent, an intelligent Email agent
120 Citations1998Gary Boone
This paper has created an intelligent email agent that can learn actions such as filtering, prioritizing, downloading to palmtops, and forwarding cmail to voicemail using automatic feature extraction, and its new feature extraction approach is based on first learning concepts present within the mail.
Munich Personal RePEc Archive (Ludwig Maximilian University of Munich)Assessing the Calibration of Naive Bayes Posterior Estimates
86 Citations2000Paul N. Bennett
Evidence is given that the posterior distribution of Naive Bayes goes to zero or one exponentially with document length, and one parametric family is investigated that attempts to downweight the growth rate.
Aberdeen University Research Archive (Aberdeen University)Using Machine Learning to Enhance Software Tools for Internet Information Management
24 Citations1996Claire Louise Green, Peter Edwards
The issues involved in the application of machine learning techniques to the management of Internet-based information are discussed, and a general architecture is presented, and how this has been instantiated in several different applications is described.
The TREC-4 Filtering Track.
23 Citations1995David Lewis
The Trec-5 filtering track, an evaluation of binary text classification systems, was a repeat of the filtering evaluation run in a trial version for TREC-4, with only the data set and participants changing.