mavengence / bachelor-thesis-textgeneration-th.ohm

IT-Based textgeneration with the use of NLP methods. A text summarization task is conducted with the amazon fine food review dataset from Kaggle. This task is done by attention and lstm neural networks.

Business Information Systems and Management

Bachelor Thesis

University of Applied Science | Georg Simon OHM

Abstract

Text Summarization can be a powerful tool to reduce the amount of time for reading documents, articles or even research papers. The thesis is divided into a larger state of the art part and a shorter prototype part. The state of the art part examines the concepts of text generation and text summarization with the focus on my prototype. Most concepts are introduced in order to fully understand how my prototype is able to achieve the text generation, except for some advanced thinking outside the box concepts, which cannot be applied by me, because it would exceed this thesis. My prototype is trained on the Amazon-finefood-reviews from www.kaggle.com and the results are evaluated on the Rouge and BLEU scores. In the end, I further introduce some performance enhancements.

Content

• Python Notebook as iPython
• Bachelor Thesis as PDF
• Presentation as PDF
• Methoden Matrix as PDF

Getting Started

Clone the repository, to get my Bachelor Thesis as PDF and LateX.

git clone https://github.com/Mavengence/Bachelor-Thesis-Textgeneration-TH.OHM.git

Prerequisites

• You need git installed to get the source
• If you want to compile the report or the presentation by urself, you need a LaTex Compiler for your OS and an IDE which makes things easier

Author

• Tim Löhr - Coding, Report, Presentation - GitHub Mavengence

License

This work and all its parts are (protected by copyright). Any use outside the narrow limits of copyright law without the author’s consent is prohibited and liable to prosecution. This applies in particular to duplications, translations, microfilming as well as storage and processing in electronic systems.

Acknowledgments

I would like to thank Prof. Dr. Alfred Holl for a really good supervising of my Bachelor Thesis and Prof. Dr. Florian Gallwitz for being my second supervisor.