REPOSITORY > RESULTS

Doctoral dissertation

Advanced text processing workflows in a web-based text mining platform

Author(s): Matic Perovšek (Author), Bojan Cestnik (Supervisor), Nada Lavrač (Co-Supervisor)

Thesis defense date: 24.05.2016

Organization: MPŠ - Mednarodna podiplomska šola Jožefa Stefana

PID: 20.500.12556/ReVIS-15046

Views: 4 | Downloads: 0

Abstract

Text mining is a field of data mining, drawing its techniques from machine learning, natural
language processing, information retrieval, information extraction and knowledge management.
While text mining is a mature area of research, with numerous open source
algorithms and NLP software libraries available such as NLTK and scikit-learn text
mining and natural language processing (NLP) experiments are still difficult to reproduce,
including the difficulty of systematic comparison of algorithms. To this end, a number of
attempts have been made to develop easy-to-use workflow management systems, allowing
users to compose complex processing pipelines in a visual programming manner.
The thesis addresses several open questions and technology gaps in text mining (TM)
and natural language processing (NLP), following three main research directions: (a) development
of advanced TM and NLP infrastructures, (b) development of advanced workflows
for natural language processing and bisociative knowledge discovery, and (c) development
of advanced workflows for transforming relational data mining into text mining problems.
We developed a web-based text mining platform TextFlows, which facilitates the construction
and execution of text mining and NLP workflows. The platform serves as an
easy-to-access integration platform for the advanced text mining and NLP approaches presented
in this thesis. A survey of existing libraries and workflow management systems for
text mining is provided, outlining their abilities and defficiencies. Moreover, the dissertation
addresses several novel knowledge discovery scenarios, for which it proposes the methodology
and presents the developed reference implementation, together with their evaluation
on real-world text mining and data mining tasks. In particular, the thesis focuses on the
development of advanced workflows for natural language processing, advanced workflows
for bisociative knowledge discovery, and advanced workflows for transforming relational
data mining problems into text mining problems. The thesis also provides reference implementations
for solving the addressed scenarios in the developed TextFlows text mining
platform, which allows for workflow sharing and experiment repeatability

Attachments

Cite this work