Semester Assignment

This website is an interactive report generator for creating a parse tree on the Horizon Data.

Dependency parsing has become important for a variety of NLP applications, such as event extraction (Björne et al., 2009), error correction (Tetreault et al., 2010), and machine translation (Stein et al., 2010). As dependency parsing is often one of the last steps in a NLP pipeline, the accuracy performance has to be high-reaching. For this reason, the present paper focuses on an empirical comparison of three dependency parsers, namely AllenNLP, spaCy, and Stanford Dependency Parser, where the latter serves simultaneously as our baseline. The data set that was parsed, consists of the multilingual magazine Horizons, a Swiss research magazine. The research question that this paper will bring light into is how the different parsers perform on the same data set in terms of accuracy.

Evaluating Parsers

Based on Data from the Horizons Research Magazine, that was provided to us by the University of Zurich, the unannotated texts were loaded into a database, where we can access individual articles, for parsing.
Stanford NLP Parser (Stanza)
Stanza is our baseline parser, against which we evaluate the fscore of two other parsers.
Multilingual Suppport
Each parser that we evaluate has its own default model from the parsing providers. We evaluate the same article in three available languages.
The output of the parsing evaluation, is a table which parser is compared to the Stanford parser more suitable for the given application.

1. Choose Issue

The first step is to select an issue, with the article that you would like to parse.
Future releases
...of the application should be able to add your own text.

2. Choose Article

Article Selection
Now, you can select an article to parse.
Future releases
...of the application will have a dependent article selection, and the ability to add your own articles.

3. Start Parsing

Now Parsing:

Please make a selection


Please make a selection


German article to be displayed here!


English article to be displayed here!


French article to be displayed here!

Evaluation Results:

# Language Stanford Allen NLP SpaCy
1 DE 0 0 0
1 EN 0 0 0
3 FR 0 0 0


Here is the project team of this application.

Bootstrap Image Preview

Tobias Weisskopf

Field of Study
Computational Linguistics and Sinology @ UZH
Professional Background
Forensic Scientist @ KPMG
Fun Fact
Voluntary working on open source software project to fight COVID-19 from Taiwan #mylog14
Bootstrap Image Preview

Ariana Dragusha

Field of Study
Digital Linguistics @ UZH
Professional Background
Research Assistant @ Slavic Seminar
Fun Fact
Study the Albanian language in Switzerland in an academic setting, makes her super happy.
Bootstrap Image Preview

Rodolfo Miranda Chavez

Field of Study
General Linguistics and English @ UNIL
Professional Background
Work as a FLE (French as a Foreign language) teacher
Fun Fact
Studying German and Italian at the moment
© All copyright reserved. . Reference: [1] Data Source: provided by UZH, as part of the course Creation and Annotation of Linguistic Resources - FS2020 [2] ...