VecTop Topic Extraction

VecTop is a corpus of contextualized word embeddings labeled with topics and sub-topics. The cosine similarity predicts the topic and sub-topic labels by identifying the topics that are most closely related to the text that is being extracted from. It is therefore possible to automatically extract the topics of an unkown text and label it.

Simply put: VecTop searches through its text collection, identifying topics by comparing which texts are the most similar to your own.

In contrast to other language models such as ChatGPT, VecTop cites its source material.
VecTop offers two corpora to choose from: Spiegel Online and New York Times. Both support all languages, but is it advisable to use the Times corpus when dealing with english text and the Spiegel corpus when using german text. Also, the Times corpus extracts the topics more subtly while Spiegel is more generic. Both have their use-cases.

Extract Topics

Extracted Topics:

This text is about the following topics alongside their subtopics:

I have made my decision based on these texts which I deemed similar in context and meaning to your text:

VecTop is used by the Bundestags-Mine, an analysis tool for German governmental data powered by Natural Language Processing and AI.

Try out Legal VecTop!

Legal VecTop functions in the same way as VecTop, except that it uses a legal corpus to locate references to your input text rather than extracting topics. These sources consist of various laws, articles, and other materials which aid you in your research.
Currently, Legal VecTop is only available in the German domain.

VecTop

Extract Topics

Extracted Topics:

Try out Legal VecTop!

Extract References

Extracted References: