I’m a postgraduate student that is working on the topic of the coherence evaluation of Ukrainian texts using machine learning techniques. The coherence of a text implies its thematic integrity, an ability of a text to convey its communication purpose to a reader. A coherent text is easier to read and understand. The example of coherent and incoherent documents is shown below.
The measurement of the coherence of a text is utilized during text generation. Moreover, an incoherent speech can indicate the first symptoms of mental illness.
I’d like to find out whether this direction is interesting for you. I’m interested in the implementation of research results in order to prove the practical meaning of this investigation. Below I provide you with links for corresponding works. Some of the works aren’t available for free access due to journal restrictions. In the case of need, I can share them privately. It should be mentioned that the Transformer-based approach outperforms others. The corresponding pre-trained model is available at PyPi: https://pypi.org/project/coherence-ua/
- AUTOMATED METHODS OF COHERENCE EVALUATION OF UKRAINIAN TEXTS USING MACHINE LEARNING TECHNIQUES (http://pp.isofts.kiev.ua/ojs1/article/view/421) - RNN, CNN, and graph-based solutions for the Ukrainian language
- Coherence Evaluation Method Based on Analyses of Phrases Graph (https://ieeexplore.ieee.org/document/9239252) - a graph-based approach for the Ukrainian language, private only
- Evaluation of the Coherence of Ukrainian Texts Using a Transformer Architecture - Transformer-based NN for the Ukrainian language, private mode
- Assessment of Text Coherence by Constructing the Graph of Semantic, Lexical, and Grammatical Consistancy of Phrases of Sentences (https://link.springer.com/article/10.1007/s10559-020-00309-7) - a graph-based approach for the English language, private only
- Assessment of text coherence based on the cohesion estimation (https://www.researchgate.net/publication/345758533_Assessment_of_text_coherence_based_on_the_cohesion_estimation) - graph-based method verification for the English, Chinese, and Arabic texts