Online Text Labelling Tool

Emotion and stance detection can help to better curate journalistic content in terms of opinions and feelings of readers towards produced content. While there has been a lot of progress on these and other text classification tasks in NLP, most models are developed for English texts. Besides, most available data sets are in English. In this project, we aim to close this gap for German language focusing on the development of emotion and stance detection models for German texts.

As a part of the project, we collect a high-quality data set of journalistic text and comments annotated with stance and emotions labels. Towards this end, we develop a labelling tool which allows us to collect crowdsourced annotations for a dataset of news articles. The tool allows labelling of each paragraph of an article with an emotion and assigning a stance and emotion to the whole article. Although the labelling granularity is developed specifically for the project, the tool is easily adaptable to accommodate different labelling modularities.

The code, together with user instructions can be found here: external page https://github.com/MTC-ETH/labels-collection-tool