• Journal of the Korea Society of Computer and Information
• /
• v.27 no.2
• /
• pp.15-23
• /
• 2022

In the field of patents, as NLP(Natural Language Processing) is a challenging task due to the linguistic specificity of patent literature, there is an urgent need to research a language model optimized for Korean patent literature. Recently, in the field of NLP, there have been continuous attempts to establish a pre-trained language model for specific domains to improve performance in various tasks of related fields. Among them, ELECTRA is a pre-trained language model by Google using a new method called RTD(Replaced Token Detection), after BERT, for increasing training efficiency. The purpose of this paper is to propose KorPatELECTRA pre-trained on a large amount of Korean patent literature data. In addition, optimal pre-training was conducted by preprocessing the training corpus according to the characteristics of the patent literature and applying patent vocabulary and tokenizer. In order to confirm the performance, KorPatELECTRA was tested for NER(Named Entity Recognition), MRC(Machine Reading Comprehension), and patent classification tasks using actual patent data, and the most excellent performance was verified in all the three tasks compared to comparative general-purpose language models.

• Annual Conference on Human and Language Technology
• /
• 2002.10e
• /
• pp.111-117
• /
• 2002

개체명 인식은 질의응답(QA), 정보 주줄(IE), 텍스트 마이닝 시스템의 성능 향상에 중요한 역할을 담당한다. 이 논문에서는 교사학습 기반의 한국어 개체명 인식에 대해 설명한다. 한국어에서 많은 개체명들이 하나 이상의 단어로 구성되어 있으며, 개체명을 구성하는 단어 사이에는 의존 관계가 존재하고, 개체명과 개체명 주위의 단어 사이에도 문맥적 의존관계를 가지고 있다. 본 논문에서는 가변길이의 개체명과 주변 문맥의 학습을 위해 트라이그램을 이용한 HMM을 사용하였으며, 자료 부족 문제를 해소하기 위해 어휘 기반이 아닌 부개체 유형 기반의 학습을 수행하였다. 학습된 개체명 인식 시스템을 이용하여 경제 분야의 신문 기사에 대한 실험 결과, 84.4%의 정확률과 90.9%의 재현률을 보였다.

• International Journal of Internet, Broadcasting and Communication
• /
• v.14 no.2
• /
• pp.55-62
• /
• 2022

A smart tourism chatbot is needed as a user interface to efficiently provide smart tourism services such as recommended travel products, tourist information, my travel itinerary, and tour guide service to tourists. We have been developed a smart tourism app and a smart tourism information system that provide smart tourism services to tourists. We also developed a smart tourism chatbot service consisting of khaiii morpheme analyzer, rule-based intention classification, and tourism information knowledge base using Neo4j graph database. In this paper, we develop the Korean and English smart tourism Name Entity (NE) datasets required for the development of the NER model using the pre-trained language models (PLMs) for the smart tourism chatbot system. We create the tourism information NER datasets by collecting source data through smart tourism app, visitJeju web of Jeju Tourism Organization (JTO), and web search, and preprocessing it using Korean and English tourism information Name Entity dictionaries. We perform training on the KoBERT-CRF NER model using the developed Korean and English tourism information NER datasets. The weight-averaged precision, recall, and f1 scores are 0.94, 0.92 and 0.94 on Korean and English tourism information NER datasets.

• Journal of the Korea Society of Computer and Information
• /
• v.23 no.1
• /
• pp.17-24
• /
• 2018

With improvement of speech recognition and natural language processing, dialog systems are recently adapted to various service domains. It became possible to get desirable services by conversation through the dialog system, but it is still necessary to improve separate modules, such as domain detection, intention detection, named entity recognition, and out-of-domain detection, in order to achieve stable service offer. When it misclassifies an in-domain sentence of conversation as out-of-domain, it will result in poor customer satisfaction and finally lost business. As there have been relatively small number of studies related to the out-of-domain detection, in this paper, we introduce a new method using a hierarchical Dirichlet process and demonstrate the effectiveness of it by experimental results on Korean dataset.

• Asia-pacific Journal of Multimedia Services Convergent with Art, Humanities, and Sociology
• /
• v.8 no.6
• /
• pp.875-885
• /
• 2018

This study analyzed the major issues of the Candlelight Rally and Taegukgi Rally registered in news comments of the politics section of the portal site from October 24, 2016 to March 19, 2017. We examined the political recognition of the two rallies with the Named Entity Recognition. The main analytical items are the responsibility for impeachment, the subject and method of settlement, and other major issues. As a result of the analysis, the comments of the Candlelight Rally focused on the impeachment support and the legal penalties of the regime ministers, and insisted on resolving the political situation through the next election after impeachment. The comments of the Taegukgi Rally focused on the rejection of the impeachment to maintain the regime and insisted on rejecting the impeachment of the Constitutional Court. The conflicts between the group that supported Candlelight Rallis and the group that supported Taegukgi rallies are predicted to last at least for the time being (Park Geun-hye's trial period) after the presidential election. After the impeachment of the President and replacement of the regime this conflict will develop into the confrontation between the pursuit of liquidation and new politics and the attempt to influence the trial of Park Geun-hye. Therefore, the efforts to integrate society in the aftermath are necessary.

• Proceedings of the Korea Contents Association Conference
• /
• 2010.05a
• /
• pp.425-427
• /
• 2010

Large-scaled information extraction consists of named-entity recognition, terminology extraction and relation extraction. Since all the elementary technologies have been studied independently so far, test collections for related machine learning models also have been constructed independently. As a result, it is difficult to handle scientific documents to extract both named-entities and technical terms at once. In this study, we integrate named-entities and terminologies with PLOT(Person, Location, Organization, Terminology) in a biomedical domain and construct a test collection of PLOT and relations between PLOTs.

• Proceedings of the Korean Society for Bioinformatics Conference
• /
• 2003.10a
• /
• pp.95-100
• /
• 2003

In this paper, we propose solutions to resolve the problem of many spelling variants and the problem of lack of annotated corpus for training, which are two among the main difficulties in named entity recognition in biomedical domain. To resolve the problem of spotting valiants, we propose a use of edit-distance as a feature for SVM. And we propose a use of virtual examples to automatically expand the annotated corpus to resolve the lack-of-corpus problem. Using virtual examples, the annotated corpus can be extended in a fast, efficient and easy way. The experimental results show that the introduction of edit-distance produces some improvements in protein name recognition performance. And the model, which is trained with the corpus expanded by virtual examples, outperforms the model trained with the original corpus. According to the proposed methods, we finally achieve the performance 75.80 in F-measure(71.89% in precision,80.15% in recall) in the experiment of protein name recognition on GENIA corpus (ver.3.0).

• Journal of Information Science Theory and Practice
• /
• v.2 no.4
• /
• pp.31-48
• /
• 2014

The proliferation of many interactive Topic Detection and Tracking (iTDT) systems has motivated researchers to design systems that can track and detect news better. iTDT focuses on user interaction, user evaluation, and user interfaces. Recently, increasing effort has been devoted to user interfaces to improve TDT systems by investigating not just the user interaction aspect but also user and task oriented evaluation. This study investigates the combination of the bag of words and named entities approaches implemented in the iTDT interface, called Interactive Event Tracking (iEvent), including what TDT tasks these approaches facilitate. iEvent is composed of three components, which are Cluster View (CV), Document View (DV), and Term View (TV). User experiments have been carried out amongst journalists to compare three settings of iEvent: Setup 1 and Setup 2 (baseline setups), and Setup 3 (experimental setup). Setup 1 used bag of words and Setup 2 used named entities, while Setup 3 used a combination of bag of words and named entities. Journalists were asked to perform TDT tasks: Tracking and Detection. Findings revealed that the combination of bag of words and named entities approaches generally facilitated the journalists to perform well in the TDT tasks. This study has confirmed that the combination approach in iTDT is useful and enhanced the effectiveness of users' performance in performing the TDT tasks. It gives suggestions on the features with their approaches which facilitated the journalists in performing the TDT tasks.

• MALSORI
• /
• v.68
• /
• pp.115-126
• /
• 2008

In this paper, we propose a vocabulary coverage improvement method for embedded continuous speech recognition (CSR) using knowledgebase. A vocabulary in CSR is normally derived from a word frequency list. Therefore, the vocabulary coverage is dependent on a corpus. In the previous research, we presented an improved way of vocabulary generation using part-of-speech (POS) tagged corpus. We analyzed all words paired with 101 among 152 POS tags and decided on a set of words which have to be included in vocabularies of any size. However, for the other 51 POS tags (e.g. nouns, verbs), the vocabulary inclusion of words paired with such POS tags are still based on word frequency counted on a corpus. In this paper, we propose a corpus independent word inclusion method for noun-, verb-, and named entity(NE)-related POS tags using knowledgebase. For noun-related POS tags, we generate synonym groups and analyze their relative importance using Google search. Then, we categorize verbs by lemma and analyze relative importance of each lemma from a pre-analyzed statistic for verbs. We determine the inclusion order of NEs through Google search. The proposed method shows better coverage for the test short message service (SMS) text corpus.

• Journal of Digital Contents Society
• /
• v.11 no.3
• /
• pp.375-383
• /
• 2010

In developing question-answering (QA) systems, it is hard to analyze natural language questions syntactically and semantically and to find exact answers to given query questions. In order to avoid these difficulties, we propose a new style of question-answering system that automatically generate natural language queries and can allow to search queries fit for given keywords. The key idea behind generating natural queries is that after significant sentences within text documents are applied to the named entity recognition technique, we can generate a natural query (interrogative sentence) for each named entity (such as person, location, and time). The natural query is divided into two types: simple type and sentence structure type. With the large database of question-answer pairs, the system can easily obtain natural queries and their corresponding answers for given keywords. The most important issue is how to generate meaningful queries which can present unambiguous answers. To this end, we propose two principles to decide which declarative sentences can be the sources of natural queries and a pattern-based method for generating meaningful queries from the selected sentences.