Inductive Modeling in Subjectivity/Sentiment Analysis (case study: dialog processing). Angels Catena, Mikhail Alexandrov, Xavier Blanco, Natalia Ponomareva

Abstract. Subjectivity/sentiment analysis is an area of natural language processing, which aims to determine people opinions or sentiments with respect to some subject or event. This paper presents a methodology for constructing empirical formulae based on lexico-syntactic properties of people utterances in order to evaluate their politeness, satisfaction and competence. This methodology includes three steps: 1) Linguistic analysis, where a set of linguistic indicators (LIs) regarding each characteristic is selected; 2) Manual annotations, where documents are manually estimated by experts; 3) Inductive modeling using the Ivakhnenko method of model self-organization (IMMSO) in order to find an optimal model describing dependency between LIs and manual estimations. The suggested methodology is applied to a real set of dialogs between passengers and Directory Inquires in Barcelona railway station.

Keywords. Subjectivity Analysis, Sentiment Analysis, Inductive Modeling.

References.

1. Alexandrov, M., Blanco, X., Ponomareva, N., Rosso, P. (2007): Constructing Empirical Models for Automatic Dialog Parameterization. In: “Text, Speech and Dialogue”, Springer, LNCS, N_4629, pp. 455-462

2. Ivakhnenko, A.: Inductive Method of Model Self-Organization of Complex Systems. Kiev, Naukova Dumka Publ.1982 (rus.)

3. Ivakhnenko, A., Stepashko, V.: Noise Immunity of Modeling. Kiev, Naukova Dumka Publ.1985 (rus.)

4. Ivakhnenko, A., Savchenko, E.: Investigation of Efficiency of Additional Determination Method of the Model Choice in the Modelling Problems with Application of GMDH Algorithms. Journ. of Inform. and Automat. Sciences, Begell House Inc, 2008, Vol. 40, N_3, pp. 4-22

5. Ponomareva, N., Blanco, X. (2008): Regression Model for Politeness Estimation Trained on Examples. In "NooJ Applications", Proc. of the Intern. Conf. NooJ-2007. Cambridge Scholars Publishing, UK, 2008, 7 pp.

6. Ponomareva, N., Catena, A.: Linguistic Models Construction and Analysis for Satisfaction Estimation. In: “Text, Speech, Dialog”, Springer, LNCS, 2009 (to be published)

7. Stepashko, V.: Method of Critical Variances as an Analytical Tool of the Inductive Modeling Theory. Journ. of Inform. and Automat. Sciences, Begell House Inc, 2008, Vol. 40, N_3, pp. 47-58

8. Webloging conferences: www.icwsm.org/2008/; www.icwsm.org/2009/; www.lsi.us.es/womsa09/