Analýza knihoven jazyka Python pro vývoj NLP aplikací

Python, as a tool for NLP processing, has many libraries that handle various tasks in their own way. They have different qualities, advantages and disadvantages. This work analyzes the mentioned qualities using a criteria scheme, designed using data collected from the addressed experts. The structure of the scheme draws on the sources listed in the literature review and offers seven evaluation aspects for each library included. Libraries were included in the research after collecting data from an anonymous work environment. 77 different respondents participated through selected survey tools. It describes their benefits for the user who decides to use these libraries or chooses between them. The conclusions themselves are drawn from the data in a discussion, where the author compares the situational use of all the tools examined. All libraries, except for the "DeepPavlov" tool, which was discarded due to an insufficient sample of collected responses, are part of the final analysis. Ascending from least to most points, the libraries are sorted as follows: PyTorch, NLTK, PYNLP1, SpaCy, ChatterBot.