• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.333-335
• /
• 2022

The development of speech recognition processing technology has been applied and used in various video and streaming services along with STT and TTS technologies. However, there are high barriers to clear written expression due to the use of dialects and overlapping of stop words, exclamations, and similar words for voice recognition of actual conversation content. In this study, for ambiguous dialects in speech recognition, we propose a speech recognition technology that applies dialect key word dictionary processing method by category and dialect prosody as speech recognition network model properties.

• ETRI Journal
• /
• v.46 no.1
• /
• pp.35-47
• /
• 2024

This study introduces CR-M-SpanBERT, a coreference resolution (CR) model that utilizes multiple embedding-based span bidirectional encoder representations from transformers, for antecedent recognition in natural language (NL) text. Information extraction studies aimed to extract knowledge from NL text autonomously and cost-effectively. However, the extracted information may not represent knowledge accurately owing to the presence of ambiguous entities. Therefore, we propose a CR model that identifies mentions referring to the same entity in NL text. In the case of CR, it is necessary to understand both the syntax and semantics of the NL text simultaneously. Therefore, multiple embeddings are generated for CR, which can include syntactic and semantic information for each word. We evaluate the effectiveness of CR-M-SpanBERT by comparing it to a model that uses SpanBERT as the language model in CR studies. The results demonstrate that our proposed deep neural network model achieves high-recognition accuracy for extracting antecedents from NL text. Additionally, it requires fewer epochs to achieve an average F1 accuracy greater than 75% compared with the conventional SpanBERT approach.

• The Transactions of the Korea Information Processing Society
• /
• v.1 no.1
• /
• pp.129-139
• /
• 1994

This paper presents a morphological analysis method for the Korean language. The characteristics and adjacency information of the words can be obtained from sentences in a large corpus. Generally a word can be analyzed to a result by applying the adjacency attributes and rules. However, we have to choose one from the several results for the ambiguous words. The collected morpheme's adjacency attributes and relations with neighbor words are recorded in a well designed dictionaries. With this information, abbreviated words as well as ambiguous words can be almost analyzed successfully. Efficiency of morphological analyzer depends on the information in the dictionaries. A morpheme dictionary and a phrase dictionary have been designed with lexical database, and necessary information extracted from the corpus is stored in the dictionaries.

• Phonetics and Speech Sciences
• /
• v.13 no.3
• /
• pp.53-64
• /
• 2021

It is known that the voicing contrast for Japanese word-initial stops is primarily realized as differences in the voice onset time (VOT). However, recent studies have reported that voiced stops are more often produced with a positive VOT than with a negative VOT among the younger generation nationwide. It is also known that post-stop F0 is associated with the stop contrast, but the degree of F0 use differs from region to region. This study explores whether the difference in post-stop F0 functions as a perceptual cue to the stop contrast along with VOT. Fifty-five college students who are native listeners from four different regions participated in two or three perception tests. The results show that VOT is a primary cue to the voiced-voiceless distinction of word-initial stops, but that the effect of post-stop F0 on the stop contrast is marginal. The post-stop F0 is involved in perception only when VOT is ambiguous, such that a sound with high F0 is more often perceived as a voiceless stop, but not vice versa. The results of this study indicate that the acoustic parameters associated with the stop contrast are not the same in production and perception, and suggest that other factors such as context, which is not an acoustic characteristic, may also be involved in the stop contrast.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.17 no.2
• /
• pp.690-696
• /
• 2016

Search engines that rely on morphological matching of user query and web document content do not support individual interests. This research proposes a personalized web search scheme that returns the results that reflect the users' query intent and personal preferences. The performance of the personalized search depends on using an effective user profiling strategy to accurately capture the users' personal interests. In this study, the user profiles are the databases of topic words and customized weights based on the recent user queries and the frequency of topic words in click history. To determine the precise meaning of ambiguous queries and topic words, this strategy uses WordNet to calculate the semantic relatedness to words in the user profile. The experiments were conducted by installing a query expansion and re-ranking modules on the general web search systems. The results showed that this method has 92% precision and 82% recall in the top 10 search results, proving the enhanced performance.

• Hwankyungkyoyuk
• /
• v.15 no.1
• /
• pp.68-82
• /
• 2002

My responses to Kim Kyung-Ok's Critique on my critique on the Hungerford approach can be summarized as follows; First, it was argued that possible confusions and misunderstandings around the concept of behavior in REB were mainly caused by Hungerford himself who has used the word in several different ways, from a bunch of overt actions to almost all kinds of responses including cognitive skills, without any clear operational definition of it for more than 20 years. It seems to be needed for future users of the word, 'Behavior' to Prevent unnecessary confusions by providing their operational definition of it. Second, REB is too ambiguous to be a legitimate goal of environmental education and too outcome-oriented to be a meaningful measure for environmental education research. Anyone who accept REB as a goal of EE or a measure for research should clearly suggest procedures and criteria for judging the environmental responsibility of actions under consideration. Third, the Hungerford approach has begun by realizing the limit of a linear traditional behavior change system and has been evolving toward a complex model with dynamic interactions among/between cognitive variables and affective variables. However, it still has one-way structural orientation toward 'Behavior' with no feedbacks. Addition of some feedback processes would make the model more flexible and realistic. Finally, both the Hines model and the Hungeford model were established based on a series of behavioristic studies including three doctoral dissertations equiped with a list of actions which were prejudged to be environmentally responsible by the researchers, not by the learners. What they were primarily interested in was not how mind functions during the learning processes but how learners' behavior can be effectively changed. Considering uncertainty and complexity associated with environmental problems, a great deal of efforts ought to be made toward more context-based and less normative studies applying cognitive psychology and quantitative approaches.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.20 no.3
• /
• pp.43-48
• /
• 2019

While reading text-only documents and finding unknown words, readers will become the focus disturbed and not be able to understand the content of the documents. Because children have little experience, it is difficult to understand correctly if the description in context is unfamiliar or ambiguous. In this paper, in order to help understand the text and increase the interest of the readers, we analyze the texts of documents and select the contents that are considered important, and implement a system that displays the most relevant images automatically from the web and links the texts and the images together. The implementation of the system divides the article into paragraphs, analyzes the text, selects important sentences for each paragraph and the important words that best represent the meaning of the important sentences, searches for images related to the words on the web, and then links the images to each of the previous paragraphs. Experiments have shown how to select important sentences and how to select important words in the sentences. As a result of the experiment, we could get 60% performance by evaluating the accuracy of the relation between three selected images and corresponding important sentences.

• Journal of the Korean Society of Costume
• /
• v.56 no.1 s.100
• /
• pp.106-120
• /
• 2006

Using aesthetics of Rosencranz as the basis, this study contains the peculiarities of 'ugliness' and the obscured conceptual meaning of 'avant-garde', 'grotesque', and 'decadence' that are being utilized under ambiguous significance are defined through modern fashion and fashion of online game characters. Forms of 'ugliness' expressed in modern fashion and in games characters display distortion of form through incongruity, unbalance, disproportion and disharmony, and with this lack of form and expressional imprecision, both contain the elements of comical characteristics of vulgarity and repugnance. The difference in 'ugliness' between modern fashion and game character fashion is, the significance of 'ugliness' being expressed in modern fashion challenges new concepts by refusing tradition and recovering the human nature that has become turbid. On the other hand, 'ugliness' in game character fashion complements the story of the game that uses legends, fantasies or novels as its basis. Opposed to the significance of recovering human nature that is displayed in modern fashion, in order to terminate the opposing game character, the fashion of game characters exaggerates the form of modern fashion with added brutality. In addition, with the advantage of virtual reality that allows a more flexible expression than in the real word, images created are more sensational and excessive use of grotesque images are being expressed.

• Proceedings of the Korean Society for Language and Information Conference
• /
• 2007.11a
• /
• pp.405-414
• /
• 2007

This paper describes a preliminary work on prosody modeling aspect of a text-to-speech system for Thai. Specifically, the model is designed to predict symbolic markers from text (i.e., prosodic phrase boundaries, accent, and intonation boundaries), and then using these markers to generate pitch, intensity, and durational patterns for the synthesis module of the system. In this paper, a novel method for annotating the prosodic structure of Thai sentences based on dependency representation of syntax is presented. The goal of the annotation process is to predict from text the rhythm of the input sentence when spoken according to its intended meaning. The encoding of the prosodic structure is established by minimizing speech disrhythmy while maintaining the congruency with syntax. That is, each word in the sentence is assigned a prosodic feature called strength dynamic which is based on the dependency representation of syntax. The strength dynamics assigned are then used to obtain rhythmic groupings in terms of a phonological unit called foot. Finally, the foot structure is used to predict the durational pattern of the input sentence. The aforementioned process has been tested on a set of ambiguous sentences, which represents various structural ambiguities involving five types of compounds in Thai.

• Journal of Information Science Theory and Practice
• /
• v.1 no.3
• /
• pp.33-46
• /
• 2013

Despite limited success, today's information retrieval (IR) systems are not intelligent or reliable. IR systems return poor search results when users formulate their information needs into incomplete or ambiguous queries (i.e., weak queries). Therefore, one of the main challenges in modern IR research is to provide consistent results across all queries by improving the performance on weak queries. However, existing IR approaches such as query expansion are not overly effective because they make little effort to analyze and exploit the meanings of the queries. Furthermore, word sense disambiguation approaches, which rely on textual context, are ineffective against weak queries that are typically short. Motivated by the demand for a robust IR system that can consistently provide highly accurate results, the proposed study implemented a novel topic detection that leveraged both the language model and structural knowledge of Wikipedia and systematically evaluated the effect of query disambiguation and topic-based retrieval approaches on TREC collections. The results not only confirm the effectiveness of the proposed topic detection and topic-based retrieval approaches but also demonstrate that query disambiguation does not improve IR as expected.