Repository for the Term Paper about "Network and Sentiment Analyses on Selected Letters of the Schlegel"

This GitHub repository serves as a main collection of all scripts and digital resources for the term paper. A list of all used external packages can be found in the corresponding section.

First steps

In order to be able to visualise the links between the people in the letter corpus, firstly, I had to extract the needed metadata from the XML files and the register.xml file. For doing so, I used a Jupyter Notebook in which I implemented the functions for metadata extraction and data enrichment. When extracting information about senders and receivers from the letters, the algorithm already counts correspondence frequencies that can later be imported for the graph visualisation.

Visualisation

For visualising the results (retrieved, also, with network_preps.ipynb), I used the free open-source software Gephi. Here, my first attempts at storing the correspondence data were overthrown by the ID and labels sections Gephi has for data input. Instead of using the GND keys I previously stored in the data, I opted to using the actual names of people from the corpus to be able to use them as labels in the graph. To be frank, the first results were quite chaotic. After some adjustments and a "getting used to Gephi" time, I achieved a decent output.

Sentiment Analysis

For the sentiment analysis, I created the notebook sentiment_analysis.ipynb in which I implemented both the short preprocessing of the texts and their extraction as well as the actual analysis with textblob-de. Further information can be taken from the comments inside the code or the term paper file.

Packages (external):

The external packages and modules I used in the creation of this term paper are the following:

• lxml 4.8.0
• numpy 1.22.3
• pandas 1.4.2
• spaCy 3.3.1
• textblob-de 0.4.3