Topics

Semantic analysis serves as the essential part of our research. We develope algorithms trying to discover hidden relationships between words and text spans according to words distribution in the corpora. Until now, we have successfully used this research in many areas (e.g. to improve the Language modeling, Named entity recognition, Sentiment analysis, Document classification, etc.).
View details
Language models are crucial for many tasks in NLP. The goal of a language model is to estimate the probability of a given word sequence. Automatic speech recognition, optical character recognition, machine translation, and other areas heavily depend on the performance of the underlying language model.
View details
Sentiment Analysis is the detection of attitudes. The basic task is to automatically decide whether a piece of text (e.g. a review, a tweet, a blog post, or a general document) is positive or negative. Also the attitude’s polarity as well as the target, source, or complex types are detected.
View details
Automatic summarization is the process of reducing a set of text documents in order to create a summary that retains the most important points of the original documents.
View details
Automatic document classification becomes very important for information organization and storage because of the fast increasing amount of electronic text documents and the rapid growth of the World Wide Web. We use primarily deep neural networks with particular focus on multi-label categorization.
View details
Given the sequence of dialogue utterances (units of speech), the task is to assign them the labels representing their function in the dialogue. For the classification of the spoken utterances, we use mostly deep neural network and we also focus on multilinguality of . Most recently, our research aims for the automatic processing of dialogues in comic books.
View details
Artificial neural networks are computing systems inspired by the biological neural networks that constitute animal brains. Such systems learn to do tasks by considering examples, generally without task-specific programming. We use various models of neural networks in wide spectrum of tasks.
View details
Named Entity Recognition (NER) automatically identifies words or phrases of a special meaning in texts and classifies them into groups (e.g. persons, organizations, products, dates, cities, coutnries, product names). We have successfully employed a modern BERT-like model focused on Czech language.
View details
Since we often work with scanned documents and other image materials such as historical texts or maps, we employ methods of image processing. We successfuly work with modern deep convolutional and fully-convolutional neural networks as well as standard computer vision algorithms (binarization, connected component analysis, edge detection and many others). We have also experience with OCR models based on convolutional and recurrent neural networks.
View details
Back to Top