• Journal of information and communication convergence engineering
• /
• v.11 no.4
• /
• pp.268-273
• /
• 2013

In this paper, we propose a document classification model using Web documents as a part of the training corpus in order to resolve the imbalance of the training corpus size per category. For the purpose of retrieving the Web documents closely related to each category, the proposed document classification model calculates the matching score between word features and each category, and generates a Web search query by combining the higher-ranked word features and the category title. Then, the proposed document classification model sends each combined query to the open application programming interface of the Web search engine, and receives the snippet results retrieved from the Web search engine. Finally, the proposed document classification model adds these snippet results as Web documents to the training corpus. Experimental results show that the method that considers the balance of the training corpus size per category exhibits better performance in some categories with small training sets.

• Journal of Korea Game Society
• /
• v.18 no.5
• /
• pp.113-122
• /
• 2018

This paper describes an approach to building and evaluating a sentiment dictionary using a Web corpus in the game domain. To build a sentiment dictionary, we collected vocabulary based on game-related web documents from a domestic portal site, using the Twitter Korean Processor. From the collected vocabulary, we selected the words whose POS are tagged as either verbs or adjectives, and assigned sentiment score for each selected word. To evaluate the constructed sentiment dictionary, we calculated F1 score with precision and recall, using Korean-SWN that is based on English Senti-word Net(SWN). The evaluation results show that average F1 scores are 0.85 for adjectives and 0.77 for verbs, respectively.

• Journal of the Korea Society of Computer and Information
• /
• v.17 no.9
• /
• pp.29-38
• /
• 2012

In this paper, we propose a named entity and event annotation tool for cultural heritage information corpus construction. Focusing on time, location, person, and event suitable for cultural heritage information management, the annotator writes the named entities and events with the proposed tool. In order to easily annotate the named entities and the events, the proposed tool automatically annotates the location information such as the line number or the word number, and shows the corresponding string, formatted as both bold and italic, in the raw text. For the purpose of reducing the costs of the manual annotation, the proposed tool utilizes the patterns to automatically recognize the named entities. Considering the very little training corpus, the proposed tool extracts simple rule patterns. To avoid error propagation, the proposed patterns are extracted from the raw text without any additional process. Experimental results show that the proposed tool reduces more than half of the manual annotation costs.

• Journal of Korea Game Society
• /
• v.17 no.2
• /
• pp.135-144
• /
• 2017

Sentiment analysis refers to a way of analyzing the writer's subjective opinions or feelings through text. For effective sentiment analysis, it is essential to build emotional word polarity lexicon. This paper introduces a crowdsourcing-based game that we have developed for efficiently building a polarity lexicon in Korean. First, we collected a corpus from the relating Internet communities using a crawler, and we classified them into words using the Twitter POS analyzer. These POS-tagged words are provided as a form of mobile platform based tagging game in which the players voluntarily tagged the polarities of the words, and then the result was collected into the database. So far we have tagged the polarities of about 1200 words. We expect that our research can contribute to the Korean sentiment analysis research especially in the game domain by collecting more emotional word data in the future.

• Annual Conference on Human and Language Technology
• /
• 2012.10a
• /
• pp.99-103
• /
• 2012

의미 역 결정 (Semantic Role Labeling)은 문장 내의 술어-논항 요소들의 의미 관계를 결정하는 과정이다. 이를 위해서는 의미 표지 부착 말뭉치가 필요하지만 한국어의 경우 이 데이터가 매우 부족한 상황이다. 본 논문에서는 한국어 Proposition Bank(이하 PropBank) 말뭉치와 세종 용언 격틀 말뭉치 구축을 위한 의미 표지 부착 작업에 대해 설명한다. 표지 부착 작업은 말뭉치의 의존 관계를 사람이 파악하여 적절한 의미 역 태그를 다는 과정이고, 이 과정으로부터 얻은 말뭉치는 의미 역 결정을 위한 기계 학습 방법론의 훈련 자료로 이용된다. 이 과정에서 필요한 구문 표지 부착 밀뭉치로는 한국전자통신연구원의 구문표지 부착 말뭉치를, 그리고 언어자원으로는 한국어 PropBank의 frame file과 세종 용언 격틀 사전을 사용한다.

• Annual Conference on Human and Language Technology
• /
• 2012.10a
• /
• pp.35-39
• /
• 2012

의미역 결정에 있어 의미 표지 부착 말뭉치는 필수적이지만 한국어 의미 표지 부착 말뭉치는 영어나 중국어와 같은 언어에 비하여 구축이 미비한 상황이다. 본 논문에서는 한국어 의미 분석을 위한 한국어 Proposition Bank(이하 PropBank)와 세종 의미 표지 부착 말뭉치의 구축을 위한 소프트웨어 도구를 개발하였다. 본 논문에서 구현한 도구는 문장 성분의 의존관계를 이용하여 주어진 술어에 대한 논항을 찾아주고, PropBank 프레임 파일과 세종 용언 격틀 사전을 활용하여 사용자가 능률적으로 한국어 PropBank와 세종 의미 표지 부착 말뭉치를 구축할 수 있도록 하였다.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.12 no.11
• /
• pp.5278-5283
• /
• 2011

One approach for very high speed korean morphological analysis is to use pre-built morphological results in dictionary. It pays the high cost to build this morphological pattern dictionary manually, besides the dictionary may contain errors. This paper proposes a method to generate morphological patterns automatically using Korean morphological synthesis. The experiment shows that we automatically generate 86% morphological patterns for analyzing Korean sentences. It takes 52.68 seconds for the morphological system using the patterns to analyze 403MB Korean corpus on 2.8GHz Window system.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.18 no.4
• /
• pp.693-700
• /
• 2023

Sentiment analysis, a branch of natural language processing that classifies and identifies subjective opinions and emotions in text documents as positive or negative, can be used for various promotions and services through customer preference analysis. To this end, recent research has been conducted utilizing various techniques in machine learning and deep learning. In this study, we propose an optimal language model by comparing the accuracy of sentiment analysis for movie, product, and game reviews using existing RNN-based models and recent Transformer-based language models. In our experiments, LMKorBERT and GPT3 showed relatively good accuracy among the models pre-trained on the Korean corpus.

• Journal of the Korea Society of Computer and Information
• /
• v.17 no.10
• /
• pp.25-33
• /
• 2012

In this paper, we propose a rule-based model to correct errors in a speech recognition result in the mobile device environment. The proposed model considers the mobile device environment with limited resources such as processing time and memory, as follows. In order to minimize the error correction processing time, the proposed model removes some processing steps such as morphological analysis and the composition and decomposition of syllable. Also, the proposed model utilizes the longest match rule selection method to generate one error correction candidate per point, assumed that an error occurs. For the purpose of deploying memory resource, the proposed model uses neither the Eojeol dictionary nor the morphological analyzer, and stores a combined rule list without any classification. Considering the modification and maintenance of the proposed model, the error correction rules are automatically extracted from a training corpus. Experimental results show that the proposed model improves 5.27% on the precision and 5.60% on the recall based on Eojoel unit for the speech recognition result.