Korytkowski Marcin. - Artificial Intelligence and Soft Computing: 16th International Conference, ICAISC 2017, Zakopane, Poland, June 11-15, 2017, Proceedings, Part I

Artificial Intelligence and Soft Computing: 16th International Conference, ICAISC 2017, Zakopane, Poland, June 11-15, 2017, Proceedings, Part I — read online for free the complete book (whole text) full work

Below is the text of the book, divided by pages. System saving the place of the last page read, allows you to conveniently read the book "Artificial Intelligence and Soft Computing: 16th International Conference, ICAISC 2017, Zakopane, Poland, June 11-15, 2017, Proceedings, Part I" online for free, without having to search again every time where you left off. Put a bookmark, and you can go to the page where you finished reading at any time.

Light

Font size:

↓

↑

Georgia Tahoma Arial Verdana

Reset

Interval:

↓

↑

Bookmark:

Make

1234567...296

Neural Networks and Their Applications

Springer International Publishing AG 2017

Leszek Rutkowski , Marcin Korytkowski , Rafa Scherer , Ryszard Tadeusiewicz , Lotfi A. Zadeh and Jacek M. Zurada (eds.) Artificial Intelligence and Soft Computing Lecture Notes in Computer Science 10245 10.1007/978-3-319-59063-9_1

Author Profiling with Classification Restricted Boltzmann Machines

Mateusz Antkiewicz 1, Marcin Kuta 1 and Jacek Kitowski 1

(1)

Department of Computer Science, Faculty of Computer Science, Electronics and Telecommunications, AGH University of Science and Technology, Al. Mickiewicza 30, 30-059 Krakow, Poland

Marcin Kuta

Email:

Abstract

This paper discusses author profiling of English-language mails and blogs using Classification Restricted Boltzmann Machines. We propose an author profiling framework with no need for handcrafted features and only minor use of text preprocessing and feature engineering. The classifier achieves competitive results when evaluated with the PAN-AP-13 corpus: 36.59% joint accuracy, 57.83% gender accuracy and 59.17% age accuracy. We also examine the relations between discriminative, generative and hybrid training methods.

Keywords

Author profiling Restricted Boltzmann Machines Classification Restricted Boltzmann Machines Discriminative training Generative training Hybrid training

Introduction

The author profiling problem is a classification task where, based on an input document, the goal is to construct a profile describing the authors gender, age, native language, personality traits, etc. Author profiling comes into play when no set of candidate authors is available ( ). Thus it may be perceived as a variation of the authorship attribution problem. Author profiling finds applications in marketing (user segmentation), linguistic forensics (profiling of authors of blackmail or offensive posts), terrorism prevention, suicide attempt detection and analysis.

The quality of authorship attribution and author profiling systems depends on a set of features applicable to the examined documents. Currently, such features have to be handcrafted (which may require years of experience) and extracted, selected or composed from core features. The number of applied features may run into the millions [].

This paper presents an author profiling system which does not rely on handcrafted features or feature engineering. This represents the first application of Restricted Boltzmann Machines to the author profiling problem. In this paper we consider two dimensions of the author profile, i.e., gender and age. Two possible values for gender and three age groups are considered, yielding possible profiles.

The contribution of this paper is as follows:

• creation of author profiling application which does not require feature engineering or feature selection and performs only a tiny amount of text preprocessing
• comparison of generative and discriminative training in author profiling tasks
• evaluation of the performance of Restricted Boltzmann Machines compared to heavy feature-engineering approaches and classical machine learning classification algorithms applied to the PAN author profiling task.

Author Profile Dimensions

The author profiling problem is fully defined only when the applicable profile dimensions are specified. Author profile dimensions include gender, age, native language, personality traits, emotional state and health.

In languages such as Polish the difference between text authored by males and females may be easy to observe due to the syntactic peculiarities of inflective languages. In other languages, like English, it is much harder to infer the authors gender. Typical style features useful for gender discrimination are determiners ( a , an , the ) and prepositions, which are preferred by men. Women, on the other hand, tend to use pronouns more frequently. Content features characteristic of men include words related to technology, whereas for women words associated with personal life and relationships are more prevalent. Koppel [] was able to classify a subset of documents from the British National Corpus with regard to gender, achieving the accuracy of approximately 80%.

Another important dimension in author profiling is age. Usually, age groups are defined as intervals with lower and upper age bounds provided. Age profiling exploits the observation that people tend to apply different vocabulary and style as they grow older. To-date attempts to discriminate documents with respect to age groups have yielded promising results. For instance, age classification performed with a corpus composed of blog notes and three age groups (1317, 2327, 3347) resulted in an accuracy of 76% [].

Restricted Boltzmann Machines

Restricted Boltzmann Machines (RBMs) have been successfully applied to various problems including image classification, speech recognition and user rating of movies.

Restricted Boltzmann Machine [].

Fig. 1.

Restricted Boltzmann Machine

The energy of RBM in configuration ( x , h ) is defined as:

(1)

Probability p ( x , h ) of RBM configuration ( x , h ) is expressed in terms of energy E ( x , h ) of this configuration, and is given by:

(2)

In the above equation is a normalization constant referred to as the partition function, .

RBMs have no lateral connections between nodes in the same layer. This lack of lateral connections makes inference easier as conditional probabilities factorize, e.g., .

Classification RBM (ClassRBM) is an extension of the RBM model which operates as a standalone classifier. It does not require an external classifier such as SVM, or to be a part of deeper multilayered architecture to perform classification.

The architecture of the Classification RBM is shown in Fig.. ClassRBM is equipped with an additional layer, , encoding the true class of input x . In our case, layer

Next page

1234567...296

Light

Font size:

↓

↑

Georgia Tahoma Arial Verdana

Reset

Interval:

↓

↑

Bookmark:

Make

Similar books «Artificial Intelligence and Soft Computing: 16th International Conference, ICAISC 2017, Zakopane, Poland, June 11-15, 2017, Proceedings, Part I»

Look at similar books to Artificial Intelligence and Soft Computing: 16th International Conference, ICAISC 2017, Zakopane, Poland, June 11-15, 2017, Proceedings, Part I. We have selected literature similar in name and meaning in the hope of providing readers with more options to find new, interesting, not yet read works.

Nikhil Marriwala

Soft Computing for Intelligent Systems: Proceedings of ICSCIS 2020

Pedro A. Castillo (editor)

Applications of Evolutionary Computation: 24th International Conference, EvoApplications 2021, Held as Part of EvoStar 2021, Virtual Event, April 7–9, ... (Lecture Notes in Computer Science, 12694)

Markus Vincze

Computer Vision Systems: 13th International Conference, ICVS 2021, Virtual Event, September 22-24, 2021, Proceedings

Charles Fox

Towards Autonomous Robotic Systems: 22nd Annual Conference, TAROS 2021, Lincoln, UK, September 8–10, 2021, Proceedings

Igor Farkaš (editor)

Artificial Neural Networks and Machine Learning – ICANN 2021: 30th International Conference on Artificial Neural Networks, Bratislava, Slovakia, ... I

De-Shuang Huang

Intelligent Computing Theories and Application: 17th International Conference, ICIC 2021, Shenzhen, China, August 12–15, 2021, Proceedings, Part II

Bernabé Dorronsoro

Optimization and Learning: 4th International Conference, OLA 2021, Catania, Italy, June 21-23, 2021, Proceedings

Allan Tucker

Artificial Intelligence in Medicine: 19th International Conference on Artificial Intelligence in Medicine, AIME 2021, Virtual Event, June 15–18, 2021, ...

Christian Hochberger (editor)

Architecture of Computing Systems: 34th International Conference, ARCS 2021, Virtual Event, June 7–8, 2021, Proceedings

Sabu M. Thampi (editor)

Advances in Computing and Network Communications: Proceedings of CoCoNet 2020, Volume 1: 735 (Lecture Notes in Electrical Engineering, 735)

Loia Vincenzo

Fuzzy Logic and Soft Computing Applications: 11th International Workshop, WILF 2016, Naples, Italy, December 19-21, 2016, Revised Selected Papers

Amsaleg Laurent

MultiMedia Modeling: 23rd International Conference, MMM 2017, Reykjavik, Iceland, January 4-6, 2017, Proceedings, Part II

Discussion, reviews of the book Artificial Intelligence and Soft Computing: 16th International Conference, ICAISC 2017, Zakopane, Poland, June 11-15, 2017, Proceedings, Part I and just readers' own opinions. Leave your comments, write what you think about the work, its meaning or the main characters. Specify what exactly you liked and what you didn't like, and why you think so.