Jana Kleckova

Automatic Dialog Acts Recognition Based on Sentence Structure

This paper deals with automatic dialog acts (DAs) recognition in Czech. Our work focuses on two applications: a multimodal reservation system and an animated talking head for hearing-impaired people. In that context, we consider the following DAs: statements, orders, investigation questions and other questions. The main goal of this paper is to propose, implement and evaluate new approaches to automatic DAs recognition based on sentence structure and prosody. Our system is tested on a Czech corpus that simulates a task of train tickets reservation. With lexical-only information, the classification accuracy is 91%. We proposed two methods to include sentence structure information, which respectively give 94% and 95%. When prosodic information is further considered, the recognition accuracy reaches 96%

Sentence Structure for Dialog Act Recognition in Czech

This paper deals with automatic dialog acts (Das) recognition in Czech based on sentence structure. We consider the following DAs: statements, orders, yes/no questions and other questions. In our previous works, we have proposed, implemented and evaluated new approaches to automatic DAs recognition based on sentence structure and prosody. The word sequences were manually transcribed. The main goal of this paper is to evaluate the performances of our approaches when these word sequences are unknown and estimated from a speech recognizer. Our system is tested on a Czech corpus that simulates a task of train tickets reservation. When manual transcription is used, classification accuracy without and with sentence structure models is 91 %, 94 % and 95 %. The recognition accuracy reaches 96 % with prosodic combination. When word sequences are estimated from a speech recognizer, the classification score is 88 % without and 91 % and 92 % with sentence structure models. The combination with prosody gives 93 % of accuracy

Semantically interoperable research medical data and meta data extraction strategy

Generally, the interoperability is the key feature to any widely used application or an information system. A system build on the basis of semantic data interpretation can also lead to an easy extensible and adaptable system, in the future. Otherwise, you have to deal with a lack of agreed terminologies or codes between standards e.g. DASTA vs. HL7, different coding structures and even variety of file formats or inconsistent database table schema. The most of these difficulties can be solved by the semantically interoperable system. We present our implementation strategy and meta data extraction methods for a research information system with an heterogeneous medical data. The medical data can have different origin, type, file format and even its version. We discuss the research information system that we primarily use for cerebrovascular brain diseases research.

Important Nonverbal Attributes for Spontaneous Speech Recognition

Understanding human emotions and their nonverbal messages is one of the most necessary and important abilities for making the next generation of human-computer interfaces (HCI) easier, more natural and effective. The main goal of this paper is to compare different methods to combine the results of both classifiers – both paralanguage and facial expressions. A prototype of the dialog system was developed in the Department of Computer Science . The proposed system is fully automatic, user-independent and real-time working. Several experiments show that the speech recognition quality is increased by using nonverbal information. The work presented in this paper was supported by the project number 2C06009.

Confidence Measures for Semi-Automatic Labeling of Dialog Acts

This paper deals with semi-supervised classifier training for automatic dialog acts (DAs) recognition. In our previous works, we have designed a dialog act recognition system for reservation applications in the Czech language. In this work, we propose to retrain this system on another corpus, for another task (broadcast news speech), in a different language (French) and with another set of dialog acts. This is realized using a semi-supervised approach based on the expectation-maximization (EM) algorithm. We show that, in the proposed experimental setup, the use of confidence measures to filter out incorrectly recognized dialog acts is required to improve the results. Two confidence measures are thus proposed and evaluated on the French broadcast news corpus. Experimental results confirm the interest of this approach for the task of training automatic dialog act classifiers.

Important prosody characteristics for spontaneous speech recognition

For languages, especially for Czech language featured by a free-word-ordering, the prosody serves a critical information for the recognition and understanding system. For some sentences the speakers style is essential to determine the core of the communication, depending on a speaker who thus emphasises a meaning of the sentence. This paper describes the first results of speakers style determination. The experiments show that the speech recognition quality is increased by the style determination by using prosody characteristics.

Inflammatory bowel disease incidence in Czech children: A regional prospective study, 2000-2015

AIM To examine the incidence and trends in pediatric inflammatory bowel diseases (IBDs) over 2000-2015 and project the incidence to 2018. METHODS A 16-year prospective study of IBD patients < 19 years of age was conducted in the Czech Republic (the Pilsen region). All incident IBD cases within a well-defined geographical area were retrieved from a prospectively collected computerized clinical database. Historical Czech data were used for comparison (1990-2001). Our catchment population was determined from the census data. We calculated the incidence by relating the number of newly diagnosed cases to the size of the pediatric population-at-risk in each calendar year. Age/sex, disease type, place of residence, and race/ethnicity were identified. RESULTS In total, 170 new IBD cases [105 Crohn’s disease (CD), 48 ulcerative colitis (UC), and 17 IBD-unclassified (IBD-U)] were identified. The median age at IBD diagnosis was 14.2 years, 59.4% were males, and 97.1% were Caucasians. A male preponderance of IBD (P = 0.026) and CD (P = 0.016) was observed. With 109209 person-years in the catchment area, the average incidence of IBD per 100000 person-years was 10.0 (6.2 for CD, 2.8 for UC, and 1.0 for IBD-U) for children aged 0 to 19 years; for those aged 0 to 15 years, the incidence rate was 7.3 (4.6 for CD, 2.0 for UC, and 0.7 for IBD-U). An increase in incidence with age was observed (P = 0.0003). Over the 16-year period, the incidence increased for IBD patients (P = 0.01) and CD in particular (P < 0.0001), whereas the incidence for UC (P = 0.09) and IBD-U (P = 0.339) remained unchanged. IBD-projected data from 2016 to 2018 were 12.1, 12.3 and 12.6 per 100000 person-years, respectively. CONCLUSION Pediatric-onset IBD incidence is around its highest point. The increase, which is particularly pronounced for CD, may be challenging to relate to causes of pediatric disease.

Are we waves or are we particles? A new insight into deep semantics in natural language processing

This paper brings conceptually new, empirically based scientific approach to a deeper understanding of human mind cognition, language acquisition, modularity of language and language origin itself. The research presented provides an interactive multilingual associative experiment as an attempt to map the Cognitive Semantic Space: (CSSES) and its basic frames of the Essential Self in the Czech language, collects and compares it to the CSSES of conceptual language view in Czech, Russian, English and potentially in other languages. We attempt to merge cognitive metaphor theory with psycholinguistics and psychoanalysis applying associative experiment methodology on the Essential Self metaphors. The research has two main goals: the first is to build an Essential Self multilingual WordNet, which serves as the basic lexical resource for Artificial Intelligence describes the core of the human nature. The second is to create a multilingual 3D semantic network.

Cerebrovascular diseases research database

Cerebrovascular diseases are one of the most common causes of death worldwide. In this paper, we analyze relationships in heterogeneous collaborating centres medical data to resolve a solution of this complex problem. Data mining is primarily based on clinical data, imaging examinations and therapeutic data stored in various data formats. The raw and mined data can be used by a registered medical doctors in the knowledge base for an evaluation of hypothesis or tests.

Collaboration and Research Information System Used for Sustainable Long-Term Research

We have to deal with some crucial difficulties when more teams collaborate in a couple of research areas at the same time. These difficulties complicates our long-term medical research and its sustainability. Each researcher use quite different data, use different tools and develops its own advanced methods for data processing. Without a commonly accepted rules there were a lot of misunderstanding, data loss or even software methods loss. Research Information System is primarily used for members collaboration and managing source data, software and research results for its life-time.