This paper explores new evaluation perspectives for image captioning and introduces a noun translation task that achieves comparative image caption generation performance by translating from a set of nouns to captions. This implies that in image captioning, all word categories other than nouns can be evoked by a powerful language model without sacrificing performance on the precision-oriented metric BLEU. The paper also investigates lower and upper bounds of how much individual word categories in the captions contribute to the final BLEU score. A large possible improvement exists for nouns, verbs, and prepositions.

This paper describes a technique to compare large text sources using word vector representations (word2vec) and dimensionality reduction (t-SNE) and how it can be implemented using Python. The technique provides a bird's-eye view of text sources, e.g. text summaries and their source material, and enables users to explore text sources like a geographical map. Word vector representations capture many linguistic properties such as gender, tense, plurality and even semantic concepts like "capital city of". Using dimensionality reduction, a 2D map can be computed where semantically similar words are close to each other. The technique uses the word2vec model from the gensim Python library and t-SNE from scikit-learn.

The main contribution of this Master's thesis is a novel way of doing text comparison using word vector representations (word2vec) and dimensionality reduction (t-SNE). This yields a bird’s-eye view of different text sources, including text summaries and their source material, and enables users to explore a text source like a geographical map.
The main goal of the thesis was to support the quality control and quality assurance efforts of a company. This goal was operationalized and subdivided into several modules. In this thesis, the Topic and Topic Comparison modules are described.
For each module, the state of the art in natural language processing and machine learning research was investigated and applied. The implementation section of this thesis discusses what each module does, how it relates to theory, how the module is implemented, the motivation for the chosen approach and self-criticism. The thesis also describes how to derive a text quality gold standard using machine learning.

My Bachelor's thesis yields feedback on advantages, disadvantages, usability problems and suggested improvements of a non-mimicking digital musical interface with an integrated music composition aid. The interface and the music composition aid were evaluated in a Thinking-aloud study with six users and analysed with a qualitative approach according to Mayring.
Columbia is a music composition aid for the iPad that focuses on harmony. The music composition aid is based on templates derived from an analysis of a set of pop songs regarding their chord progressions.