Learning from communication data: Language in electronic business negotiations

Description
Title: Learning from communication data: Language in electronic business negotiations
Authors: Sokolova, Marina
Date: 2006
Abstract: When people communicate, language is one of the means of reaching the goal of communication. Negotiations by electronic means is an example of communication where language is the principal deal-making tool. Negotiators use language to persuade, threaten and query, aiming to get the largest piece of the pie, to reach a compromise or to find prospective partners. Here is a sample from electronic negotiations, with the original spelling, punctuation and capitalization: Seller. Dear BuTTerFLy Thanks for your offer. I see there are still some things that have to be thought about. We both come along with payment upon delivery. I could imagine a price of $3.98 and delivery 45 days, but unfortunately with the returns i cant make you any other offers. I hope you quite like this offer. Im sure an agreement will be found. Im looking forward to your respond, daisy. Buyer. To my dearest friend daisy... Thank you for your quick respond, I quite like your second offer. However I'll be more than happy if the price goes down to 3.71$ and the delivery would be within 30 days (about the payment and the return I don't have any problems with them). I'll really appreciate it if you accept the offer I just made, but if you don't, I'm sure somehow we'll come up with an agreement. yours faithfully BuTTerFLy!!!!! We apply statistical modelling and build a semantic lexicon to find the characteristics of e-negotiation data which make it unique. We find language patterns that signal of negotiator roles and success or failure of negotiations. Research in human communication shows that it is very difficult to find the characteristics of unsuccessful activities and communication corresponding to them. The interesting and promising result of this dissertation comes in the form of identifying two sets of features that characterize successful and unsuccessful communication respectively. We use these sets to represent negotiations and then classify the negotiation outcomes. The results show the advantage of the proposed feature selection approach compared with the popular statistical selection. We apply our research to the largest available collection of electronic negotiations and, when appropriate, to data of face-to-face negotiations. In the dissertation we employ methods developed for Corpus Linguistics, Natural Language Processing and Machine Learning. We investigate the ability of the methods to model and classify the data. Throughout the dissertation we examine hypotheses on language, learning and the process of electronic negotiations.
URL: http://hdl.handle.net/10393/29317
http://dx.doi.org/10.20381/ruor-12891
CollectionTh├Ęses, 1910 - 2010 // Theses, 1910 - 2010
Files
NR15045.PDF10.59 MBAdobe PDFOpen