Topic Evolution Model

TEM is a graph-based temporal topic modeling for very small corpora. Detailed information about TEM can be found in the following publication

Liebe L, Baum J, Cech T, Scheibel W, and Dollner J (2024). Detecting and Comparing LLM Capabilities to Human Writers through Linguistic Analysis

Usage

To use TEM, first build the model executable and term distance server Docker container.

• Build the model:

  mkdir build && cd build
  cmake .. && make

  An executable will be compiled into build/topic_evolution_model.
• Build the term distance server:

  make -C term-distance build

  Note that this may take a while since it downloads word2vec models (total of around 8GB).

Once this is set up, you can start using TEM.

• Run the term distance server, e.g. with

  make -C term-distance run

  This will expose the server on localhost:8000. To stop the server, run

  make -C term-distance kill

• You can then run the model executable (in build/topic_evolution_model) directly or use the provided Python interface in the script/ directory, which also features parallelization.
  Note that you need to specify the URL to the term distance server via the environment variable TEM_WORD_DISTANCE_ENDPOINT, e.g. if running the executable directly

  TEM_WORD_DISTANCE_ENDPOINT=http://localhost:8000/similarity \
    build/topic_evolution_model [...]