• Phonetics and Speech Sciences
• /
• v.12 no.4
• /
• pp.73-80
• /
• 2020

In this study, a method for measuring the jitter component in continuous speech is presented. In the conventional jitter measurement method, pitch variabilities are commonly measured from the sustained vowels. In the case of continuous speech, such as a spoken sentence, distortion occurs with the existing measurement method owing to the influence of prosody information according to the sentence. Therefore, we propose a method to reduce the pitch fluctuations of prosody information in continuous speech. To remove this pitch fluctuation component, a curve representing the fluctuation is obtained via polynomial interpolation for the pitch track in the analysis interval, and the shift is removed according to the curve. Subsequently, the variability of the pitch frequency is obtained by a method of measuring jitter from the trajectory of the pitch from which the shift is removed. To measure the effects of the proposed method, parameter values before and after the operations are compared using samples from the Kay Pentax MEEI database. The statistical analysis of the experimental results showed that jitter components from the continuous speech can be measured effectively by proposed method and the values are comparable to the parameters of sustained vowel from the same speaker.

• Phonetics and Speech Sciences
• /
• v.11 no.3
• /
• pp.77-87
• /
• 2019

The study aimed to develop Speech Range Profile (SRP) and to examine and validate its clinical application. Forty-five participants without voice disorders aged 18-29 years were compared using SRP and Voice Range Profile (VRP). The authors developed the "Fire!" paragraph as a SRP task compromising 14 sentences including all Korean spoken phonemes and sentence types. To compare SRP and VRP results, the participants read the paragraph (reading) and counted from 21 to 30 (counting) as a part of SRP tasks, and produced a vowel /a/ from low to high frequencies (gliding) and a shortened form of the VRP as a part of VRP tasks. $F0_{max}$, $F0_{min}$, $F0_{range}$, $I_{max}$, $I_{min}$, and $I_{range}$ for each task were measured and compared, showing that $F0_{max}$, $F0_{min}$, $F0_{range}$, $I_{max}$, and $I_{range}$ were not different between reading and gliding. $I_{min}$, had the lowest value in counting. It is concluded that the newly developed SRP task, reading the "Fire" paragraph, can yield a maximum phonation range similar to that found by VRP. Therefore, it is expected that voice evaluation can be effectively performed in a relatively short time by applying SRP with the "Fire" paragraph, a functional utterance task, in place of VRP, which may be difficult to measure long term or in cases of severe voice disorders.

• Journal of Intelligence and Information Systems
• /
• v.25 no.3
• /
• pp.161-177
• /
• 2019

In this paper, we study the performance improvement of the answer extraction in Question-Answering system by using sentence dependency parsing result. The Question-Answering (QA) system consists of query analysis, which is a method of analyzing the user's query, and answer extraction, which is a method to extract appropriate answers in the document. And various studies have been conducted on two methods. In order to improve the performance of answer extraction, it is necessary to accurately reflect the grammatical information of sentences. In Korean, because word order structure is free and omission of sentence components is frequent, dependency parsing is a good way to analyze Korean syntax. Therefore, in this study, we improved the performance of the answer extraction by adding the features generated by dependency parsing analysis to the inputs of the answer extraction model (Bidirectional LSTM-CRF). The process of generating the dependency graph embedding consists of the steps of generating the dependency graph from the dependency parsing result and learning the embedding of the graph. In this study, we compared the performance of the answer extraction model when inputting basic word features generated without the dependency parsing and the performance of the model when inputting the addition of the Eojeol tag feature and dependency graph embedding feature. Since dependency parsing is performed on a basic unit of an Eojeol, which is a component of sentences separated by a space, the tag information of the Eojeol can be obtained as a result of the dependency parsing. The Eojeol tag feature means the tag information of the Eojeol. The process of generating the dependency graph embedding consists of the steps of generating the dependency graph from the dependency parsing result and learning the embedding of the graph. From the dependency parsing result, a graph is generated from the Eojeol to the node, the dependency between the Eojeol to the edge, and the Eojeol tag to the node label. In this process, an undirected graph is generated or a directed graph is generated according to whether or not the dependency relation direction is considered. To obtain the embedding of the graph, we used Graph2Vec, which is a method of finding the embedding of the graph by the subgraphs constituting a graph. We can specify the maximum path length between nodes in the process of finding subgraphs of a graph. If the maximum path length between nodes is 1, graph embedding is generated only by direct dependency between Eojeol, and graph embedding is generated including indirect dependencies as the maximum path length between nodes becomes larger. In the experiment, the maximum path length between nodes is adjusted differently from 1 to 3 depending on whether direction of dependency is considered or not, and the performance of answer extraction is measured. Experimental results show that both Eojeol tag feature and dependency graph embedding feature improve the performance of answer extraction. In particular, considering the direction of the dependency relation and extracting the dependency graph generated with the maximum path length of 1 in the subgraph extraction process in Graph2Vec as the input of the model, the highest answer extraction performance was shown. As a result of these experiments, we concluded that it is better to take into account the direction of dependence and to consider only the direct connection rather than the indirect dependence between the words. The significance of this study is as follows. First, we improved the performance of answer extraction by adding features using dependency parsing results, taking into account the characteristics of Korean, which is free of word order structure and omission of sentence components. Second, we generated feature of dependency parsing result by learning - based graph embedding method without defining the pattern of dependency between Eojeol. Future research directions are as follows. In this study, the features generated as a result of the dependency parsing are applied only to the answer extraction model in order to grasp the meaning. However, in the future, if the performance is confirmed by applying the features to various natural language processing models such as sentiment analysis or name entity recognition, the validity of the features can be verified more accurately.

• Journal of the Korean Chemical Society
• /
• v.63 no.4
• /
• pp.266-279
• /
• 2019

The purpose of this study was to investigate the inquiry tendency of North Korean chemistry textbook by Romey's method. We analyzed the inquiry tendencies of texts, figures, questions, activities, and summaries by Romey's method. And the following results were compared with those of previous South Korean studies, which method was the same. The 20,017 texts, 541 figures, 140 questions, 243 activities, and 25 summaries in 5 chapters of chemistry textbook of the Higher-level middle school 1 in the Kim Jong-Un Era were analyzed. Results were as follows: texts were authoritarian tendency, figures were authoritarian tendency, questions were excessive inquiry tendency, activities were inquiry tendency, and chapter summaries were drastic authoritarian tendency. The inquiry tendency of North Korean chemistry textbook in higherlevel middle school showed similar tendencies as the textbooks of South Korean chemistry I textbook in the 6th National Science Curriculum. The results of this study are expected to be utilized as basic information and data to understand the inquiry teaching that North Korea science education is aiming at.

• Science of Emotion and Sensibility
• /
• v.24 no.1
• /
• pp.91-104
• /
• 2021

This study explored users' emotional states identified from the wake-up words -"Hi, KIA!"- using a machine learning algorithm considering the user interface of passenger cars' voice. We targeted four emotional states, namely, excited, angry, desperate, and neutral, and created a total of 12 emotional scenarios in the context of car driving. Nine college students participated and recorded sentences as guided in the visualized scenario. The wake-up words were extracted from whole sentences, resulting in two data sets. We used the soundgen package and svmRadial method of caret package in open source-based R code to collect acoustic features of the recorded voices and performed machine learning-based analysis to determine the predictability of the modeled algorithm. We compared the accuracy of wake-up words (60.19%: 22%~81%) with that of whole sentences (41.51%) for all nine participants in relation to the four emotional categories. Accuracy and sensitivity performance of individual differences were noticeable, while the selected features were relatively constant. This study provides empirical evidence regarding the potential application of the wake-up words in the practice of emotion-driven user experience in communication between users and the artificial intelligence system.

• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.1
• /
• pp.1-8
• /
• 2021

In the political field of news articles, there are polarized and biased characteristics such as conservative and liberal, which is called political bias. We constructed keyword-based dataset to classify bias of news articles. Most embedding researches represent a sentence with sequence of morphemes. In our work, we expect that the number of unknown tokens will be reduced if the sentences are constituted by subwords that are segmented by the language model. We propose a document embedding model with subword tokenization and apply this model to SVM and feedforward neural network structure to classify the political bias. As a result of comparing the performance of the document embedding model with morphological analysis, the document embedding model with subwords showed the highest accuracy at 78.22%. It was confirmed that the number of unknown tokens was reduced by subword tokenization. Using the best performance embedding model in our bias classification task, we extract the keywords based on politicians. The bias of keywords was verified by the average similarity with the vector of politicians from each political tendency.

• Journal of the Korean Society for information Management
• /
• v.23 no.2
• /
• pp.229-243
• /
• 2006

This study aims to suggest an effective way to enhance question-answer(QA) document retrieval performance by reconstructing queries based on the structural features in the QA documents. QA documents are a structured document which consists of three components : question from a questioner, short description on the question, answers chosen by the questioner. The study proposes the methods to reconstruct a new query using by two major structural parts, question and answer, and examines which component of a QA document could contribute to improve query performance. The major finding in this study is that to use answer document set is the most effective for reconstructing a new query. That is, queries reconstructed based on terms appeared on the answer document set provide the most relevant search results with reducing redundancy of retrieved documents.

• Proceedings of the Korean Society for Language and Information Conference
• /
• 2001.06a
• /
• pp.141-144
• /
• 2001

본 논문의 목적은 ‘다중 동사성 명사 구문’(Multiple Verbal Noun Construe-tions)의 논항실현 양상을 이론 중립적으로 고찰해 보고, 이 분석을 제약기반 문법 이론인 최근의 핵 심어주도 구구조문법 (Head-driven Phrase Structure Grammar)틀 안에서, 특히 다중계승위 계를 가정하는 제약기반 어휘부를 기반으로 형식화해 논항의 실현과정을 기술하고 설명하는 것이다. 우선 일본어의 유사한 현상을 분석한 Grimshaw & Mester (1988)의 격실현 양상에 관한 일반화를 기반으로 한국어 동사성명사구문의 논항실현 양상을 ‘논항전이’ (argument transfer)라는 이론적 장치를 이용해 형식화할 수 있음을 보이고, 동사성 합성명사의 논항구조를 만들기 위해 ‘논항합성’(argument composition)이라는 이론적 장치를 제안한다. 나아가서 다중 동사성 명사구문의 논항실현 과정에서 보이는 겹격표지 현상을 ‘격 복사’(case copying)를 제안해 동사성 명사의 격표지가 합성 명사에서 분리되어 문장단위에서 실현될 때 동일한 격을 복사해 실현한다는 점을 주장하고자 한다. 이 주장을 뒷받침하기 위해 수동과 능동 등 문법기능의 변화현상에서 하위범주화된 요소들의 격변화가 자의적이 아님을 실례를 들어 보여 주고자 한다. 일본어의 경동사 (light verbs)에 관한 분석 인 Grimshaw Meste, (1988) 이래 한국어에서도 이와 유사한 구문에 대한 재조명이 활발하게 이루어져 왔다 (Ryu (1993b), 채희락 (1996), Chae (1997) 등 참조). 한국어에서 ‘하다’와 동사성명사(verbal nouns)가 결합하여 이루어진 ‘동사성명사구문’ (Verbal Noun Constructions)에 대한 기존의 논의는 대부분 하나의 동사성 명사가 ‘하다’나 ‘되다등 소위 문법기능을 바꾸는 ‘경동사’들과 결합하여 복합술어가 되는 문법적 현상에 초점이 맞춰져 있었다. 그와 비교해서 동사성 명사의 어근이 두 개 이상 결합하여 동사성명사들끼리 합성명사(compound nouns)를 이루고 그 동사성 합성명사가 문법기능의 변화를 바꾸는 ‘경동사’와 결합하여 이루어진 복합술어에 대해서는 논의가 거의 없는 형편이다. 특히 이 지적은 핵심어주도 구절구조문법틀 내에서는 논란의 여지가 없다. 본 논문의 대상은 바로 이러한 합성 동사성명사의 논항구조와 동사성명사에 의해 하위범주화된 논항들의 문법적 실현양상이다.

• The Transactions of the Korea Information Processing Society
• /
• v.5 no.3
• /
• pp.681-687
• /
• 1998

The paper presents a hybrid mcthod for disambiguation of the verb meaning in the machine translation. The presented verb translation algorithm is to perform the concept-based method and the statistics-based method simultaneously. It uses a collocation dictionary, WordNct and the statistical information extracted from corpus. In the transfer phase of the machine translation, it tries to find the target word of the source verb. If it fails, it refers to Word Net to try to find it by calculating word similarities between the logical constraints of the source sentence and those in the collocation dictionary. At the same time, it refers to the statistical information extracted from corpus to try to find it by calculating co-occurrence similarity knowledge. The experimental result shows that the algorithm performs more accurate verb translation than the other algorithms and improves accuracy of the verb translation by 24.8% compared to the collocation-based method.

• Journal of the Institute of Convergence Signal Processing
• /
• v.20 no.1
• /
• pp.31-36
• /
• 2019

A chatbot is an important area of mobile service, which understands informal data of a user as a conversational form and provides a customized service information for user. However, there is still a lack of a service way to fully understand the user's natural language typed query dialogue. Therefore, in this paper, we extract meaningful words, such a region, a food category, and a restaurant name from user's dialogue sentences for recommending a restaurant. and by comparing the extracted words against the contents of the knowledge database that is built from the hashtag for recommending a restaurant in SNS, and provides user target information having statistically much the word-similarity. In order to evaluate the performance of the restaurant recommendation chatbot system implemented in this paper, we measured the accessibility of various user query information by constructing a web-based mobile environment. As a results by comparing a previous similar system, our chabot is reduced by 37.2% and 73.3% with respect to the touch-count and the cutaway-count respectively.