• Phonetics and Speech Sciences
• /
• v.7 no.1
• /
• pp.47-55
• /
• 2015

This study examined whether speech acceptability was changed under various conditions of prosodic manipulations. Both speech rate and voice loudness reportedly are associated with acceptability and intelligibility. Speech samples by twelve speakers with mild dysarthria were recorded. Speech rate and loudness changes were made by digitally manipulating habitual sentences. 3 different loudness levels (70, 75, & 80dB) and 4 different speech rates (normal, 20% rapidly, 20% slowly, & 40% slowly) were presented to 12 SLPs (speech language pathologists). SLPs evaluated sentence acceptability by 7-point Likert scale. Repeated ANOVA were conducted to determine if the prosodic type of resynthesized cue resulted in a significant change in speech acceptability. A faster speech rate (20% rapidly) rather than habitual and slower rates (20%, 40% slowly) resulted in significant improvement in acceptability ratings (p <.001). An increased vocal loudness (up to 80dB) resulted in significant improvement in acceptability ratings (p <.05). Speech rate and loudness changes in the prosodic properties of speech may contribute to improved acceptability.

• Proceedings of the KSPS conference
• /
• 2007.05a
• /
• pp.330-333
• /
• 2007

This paper presents a dialogue interface using the dialogue management system as a method for controlling home appliances in Home Network Services. In order to realize this type of dialogue interface, we annotated 96,000 utterance pair sized dialogue set and developed an example-based dialogue system. This paper introduces the automatic error correction module for the SMS-styled sentence. With this module we increase the accuracy of NLU(Natural Language Understanding) module. Our NLU module shows an accuracy of 86.2%, which is an improvement of 5.25% over than the baseline. The task completeness of the proposed SMS dialogue interface was 82%.

• Language and Information
• /
• v.8 no.2
• /
• pp.1-25
• /
• 2004

This paper examines two different approaches to resolving a theoretical problem which the bottom-up approach version of Discourse Representation Theory of Kamp et al. (2003) faces in dealing with anaphoric relations between pronouns and their potential antecedents in conditional sentences where consequent clauses precede their corresponding conditional clauses. In one of the approaches, every element is processed in the order of occurrence and conditional operators in a non-sentence-initial position cause the ongoing DR to split in two with the same index. The definition of accessibility is accordingly modified so that the right DR can be accessible from the left DR. In the other approach, a different type of discourse representation structure, K ${\Leftarrow}$ K, is introduced, which allows us to resolve the target problem without modifying accessibility proposed in Kamp et al. (2003). Compatibility of these two approaches with the bottom-up version of DRT is evaluated by examining their applicability to the analysis of quantified sentences where pronominal expressions precede generalized quantifiers.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.4 no.3
• /
• pp.288-291
• /
• 2004

This paper proposes an intelligent search modeling using avatar agent. This system consists of some modules such as agent interface, agent management, preprocessor, interface machine. Core-Symbol Database and Spell Checker are related to the preprocessor module and Interface Machine is connected with Best Aggregate Designer. Our avatar agent system does the indexing work that converts user's natural language type sentence to the proper words that is suitable for the specific branch information retrieval. Indexing is one of the preprocessing steps that make it possible to guarantee the specialty of user's input and increases the reliability of the result. It references a database that consists of synonym and specific branch dictionary. The resulting symbol after indexing is used for draft search by the internet search engine. The retrieval page position and link information are stored in the database. We experimented our system with the stock market keyword SAMSUNG_SDI, IBM, and SONY and compared the result with that of Altavista and Google search engine. It showed quite excellent results.

• Proceedings of the KSPS conference
• /
• 2007.05a
• /
• pp.152-154
• /
• 2007

Submucosal type cleft palate is subdivision of cleft palate. Because of late detection, the treatment - for example, the operation or the speech therapy - for the submucosal type cleft palate patient usually late. In this study, we want to find the objective characteristics of submucosal type cleft palate patient, comparing with the normal and the complete cleft palate patient. Experimental groups are 10 submucosal type cleft palate patients who got the operation in our hospital, 10 complete cleft palate patients. And, 10 normals as control group. The sentence patterns using in this study is simple 5 vowels. Using CSL program we evaluate the Formant, Bandwidth. We analized the spectral characteristics of speech signals of 3 groups, before and after the operation.

• English Language & Literature Teaching
• /
• v.18 no.1
• /
• pp.23-43
• /
• 2012

This study investigates the effects of three different types of instruction, namely form-based, comprehension-based, and production-based on the development of Korean university students' (n=137) comprehension and production of English relative clauses (RCs). The extent of improvements was analyzed by administering pre-and post-tests consisting of two comprehension tests (selecting the right form of RCs and the right picture descriptions) and one production test (combining two sentences). Findings of this study suggest that all three types of instruction increased participants' comprehension and productions of RCs. However, there appeared differential effects by the instruction type. It was found production-based instruction was most effective in promoting comprehension, followed by comprehension-based instruction. Comprehension-based instruction worked best with the development of production, suggesting that the effects of comprehension training did not only work for increasing comprehension skills, but also transfer to production skills. The type or level of tasks employed for each instruction appeared to play an important role in causing such results. Form-based instruction displayed the lowest improvements in both comprehension and production of RCs. A sentence-combination task employed for form-based instruction appear to result in mere explicit rule explanations without chances to notice rules in context or use their knowledge in practice.

• ETRI Journal
• /
• v.41 no.6
• /
• pp.771-781
• /
• 2019

We propose the VS³-NET model to solve the task of question answering questions with machine-reading comprehension that searches for an appropriate answer in a given context. VS³-NET is a model that trains latent variables for each question using variational inferences based on a model of a simple recurrent unit-based sentences and self-matching networks. The types of questions vary, and the answers depend on the type of question. To perform efficient inference and learning, we introduce neural question-type models to approximate the prior and posterior distributions of the latent variables, and we use these approximated distributions to optimize a reparameterized variational lower bound. The context given in machine-reading comprehension usually comprises several sentences, leading to performance degradation caused by context length. Therefore, we model a hierarchical structure using sentence encoding, in which as the context becomes longer, the performance degrades. Experimental results show that the proposed VS³-NET model has an exact-match score of 76.8% and an F1 score of 84.5% on the SQuAD test set.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.18 no.4
• /
• pp.451-455
• /
• 2008

In this paper we embody a system that recognizes the printed alphabet of two font types (Batang, Dodum) inputted by a black-and-white CCD camera and converts it into an editable text form. The image of the inputted printed sentences is binarized, then the rows of each sentence are separated through the vertical projection using the Histogram method, and the height of the characters are normalized to 48 pixels. With the reverse application of the basic principle of the Nonogram puzzle to the individual normalized character, the character is covered with the pixel-based squares, representing the characteristics of the character as the numerical information of the Nonogram puzzle in order to recognize the character through the comparison with the standard pattern information. The test of 2609 characters of font type Batang and 1475 characters of font type Dodum yielded a 100% recognition rate.

• Journal of Biomedical Engineering Research
• /
• v.45 no.1
• /
• pp.1-9
• /
• 2024

Communication with others plays an important role in human social interaction and information exchange in modern society. However, some individuals have difficulty in communicating due to dysarthria. Therefore, it is necessary to develop effective diagnostic techniques for early treatment of the dysarthria. In the present study, we propose a mobile device-based methodology that enables to automatically classify dysarthria type. The light-weight CNN model was trained by using the open audio dataset of Korean patients with dysarthria. The trained CNN model can successfully classify dysarthria into related subtype disease with 78.8%~96.6% accuracy. In addition, the user-friendly mobile application was also developed based on the trained CNN model. Users can easily record their voices according to the selected inspection type (e.g. word, sentence, paragraph, and semi-free speech) and evaluate the recorded voice data through their mobile device and the developed mobile application. This proposed technique would be helpful for personal management of dysarthria and decision making in clinic.

• Lingua Humanitatis
• /
• v.1 no.1
• /
• pp.235-251
• /
• 2001

According to Chomskyan grammar, humans can generate and understand an unbounded number of grammatical sentences. Against the background of pure and idealized linguistic competence, this linguistic productivity is argued and understood. In actual utterances, however, there are many limitations of productivity but they are said to come from the general constraints on performances such as capacity of short term memory or attention. In this paper I discuss a problem raised against idealized productivity. I argue that linguistic productivity idealizes our linguistic competence too much. By separating idealized competence from the various constraints of performance, Chomskyan theorists can argue for unlimited productivity. However, the absolute distinction between grammar (pure competence) and parser (actual psychological processes) makes little sense when we explain the low acceptability(intelligibility) of center embedded sentences. Usually, the problem of center embedded sentence is explained in terms of memory shortage or other performance constraints. To explain the low acceptability, however, we need to assume specialized memory structure because the low acceptability occurs only with a specific type of syntactic pattern. 1 argue that this special memory structure should not be considered as a general performance constraint. It is a domain specific (specifically linguistic) constraints and an intrinsic part of human language processing. Recent development of Chomskyan grammar, i.e., minimalist approach seems to close the gap between pure competence and this type of specialized constraints. Chomsky's earlier approach of generative grammar focuses on end result of the generative derivation. However, economy principle (of minimalist approach) focuses on actual derivational processes. By having less mathematical or less idealized grammar, we can come closer to the actual computational processes that build syntactic structure of a sentence. In this way, we can have a more concrete picture of our linguistic competence, competence that is not detached from actual computational processes.