• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.474-478
• /
• 2023

논증 마이닝이란 비정형의 텍스트 데이터에서 논증 구조와 그 요소들을 식별, 분석, 추출하는 자연어 처리의 한 분야다. 논증 마이닝의 하위 작업인 주장-증거 쌍 추출은 주어진 문서에서 자동으로 주장과 증거 쌍을 추출하는 작업이다. 본 논문에서는 효과적인 주장-증거 쌍 추출을 위해, 문서 단위의 문맥 정보를 이용하고 주장과 증거 간의 종속성을 반영하기 위한 계층적 LAN 방법을 제안한다. 실험을 통해 서로의 정보를 활용하는 종속적인 구조가 독립적인 구조보다 우수함을 입증하였으며, 최종 제안 모델은 Macro F1을 기준으로 13.5%의 성능 향상을 보였다.

• Journal of Information Processing Systems
• /
• v.19 no.5
• /
• pp.641-651
• /
• 2023

Malicious hate speech and gender bias comments are common in online communities, causing social problems in our society. Gender bias and hate speech detection has been investigated. However, it is difficult because there are diverse ways to express them in words. To solve this problem, we attempted to detect malicious comments in a Korean hate speech dataset constructed in 2020. We explored bidirectional encoder representations from transformers (BERT)-based deep learning models utilizing hyperparameter tuning, data sampling, and logits ensembles with a label distribution. We evaluated our model in Kaggle competitions for gender bias, general bias, and hate speech detection. For gender bias detection, an F1-score of 0.7711 was achieved using an ensemble of the Soongsil-BERT and KcELECTRA models. The general bias task included the gender bias task, and the ensemble model achieved the best F1-score of 0.7166.

• Journal of Intelligence and Information Systems
• /
• v.28 no.2
• /
• pp.127-146
• /
• 2022

Recently, as word embedding has shown excellent performance in various tasks of deep learning-based natural language processing, researches on the advancement and application of word, sentence, and document embedding are being actively conducted. Among them, cross-language transfer, which enables semantic exchange between different languages, is growing simultaneously with the development of embedding models. Academia's interests in vector alignment are growing with the expectation that it can be applied to various embedding-based analysis. In particular, vector alignment is expected to be applied to mapping between specialized domains and generalized domains. In other words, it is expected that it will be possible to map the vocabulary of specialized fields such as R&D, medicine, and law into the space of the pre-trained language model learned with huge volume of general-purpose documents, or provide a clue for mapping vocabulary between mutually different specialized fields. However, since linear-based vector alignment which has been mainly studied in academia basically assumes statistical linearity, it tends to simplify the vector space. This essentially assumes that different types of vector spaces are geometrically similar, which yields a limitation that it causes inevitable distortion in the alignment process. To overcome this limitation, we propose a deep learning-based vector alignment methodology that effectively learns the nonlinearity of data. The proposed methodology consists of sequential learning of a skip-connected autoencoder and a regression model to align the specialized word embedding expressed in each space to the general embedding space. Finally, through the inference of the two trained models, the specialized vocabulary can be aligned in the general space. To verify the performance of the proposed methodology, an experiment was performed on a total of 77,578 documents in the field of 'health care' among national R&D tasks performed from 2011 to 2020. As a result, it was confirmed that the proposed methodology showed superior performance in terms of cosine similarity compared to the existing linear vector alignment.

• KIPS Transactions on Software and Data Engineering
• /
• v.9 no.2
• /
• pp.53-60
• /
• 2020

Sentiment analysis is the automated process of understanding attitudes and opinions about a given topic from written or spoken text. One of the sentiment analysis approaches is a dictionary-based approach, in which a sentiment dictionary plays an much important role. In this paper, we propose a method to automatically generate Korean sentiment lexicon from the well-known English sentiment lexicon called VADER (Valence Aware Dictionary and sEntiment Reasoner). The proposed method consists of three steps. The first step is to build a Korean-English bilingual lexicon using a Korean-English parallel corpus. The bilingual lexicon is a set of pairs between VADER sentiment words and Korean morphemes as candidates of Korean sentiment words. The second step is to construct a bilingual words graph using the bilingual lexicon. The third step is to run the label propagation algorithm throughout the bilingual graph. Finally a new Korean sentiment lexicon is generated by repeatedly applying the propagation algorithm until the values of all vertices converge. Empirically, the dictionary-based sentiment classifier using the Korean sentiment lexicon outperforms machine learning-based approaches on the KMU sentiment corpus and the Naver sentiment corpus. In the future, we will apply the proposed approach to generate multilingual sentiment lexica.

• 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.

• 한국어정보학회:학술대회논문집
• /
• 2016.10a
• /
• pp.14-19
• /
• 2016

본 논문은 문장에서의 어절 간 의존관계가 성립될 때 의존소와 지배소가 어떠한 관계를 가지는지 의존 관계명을 부착하는 모델을 제안한다. 국내에서 한국어 의존구문분석에 관한 연구가 활발히 진행되고 있지만 의존 관계만을 결과로 제시하고 의존 관계명을 제공하지 않는 경우가 많았다. 따라서 본 논문에서는 의존경로(Dependency Path)와 음절의 의존 관계명 분포를 반영하는 음절 임베딩를 이용한 의존 관계명 부착모델을 제안한다. 문장에서 나올 수 있는 최적의 입력 열인 의존 경로(Dependency Path)를 순차 레이블링에서 좋은 성능을 나타내고 있는 bidirectional LSTM-CRFs의 입력 값으로 사용하여 의존 관계명을 결정한다. 제안된 기법은 자질에 대한 많은 노력 없이 의존 경로에 따라 어절 및 음절 단어표상(word embedding)만을 사용하여 순차적으로 의존 관계명을 부착한다. 의존 경로를 사용하지 않고 전체 문장의 어절 순서를 바탕으로 자질을 추출하여 CRFs로 분석한 기존 모델보다 의존 경로를 사용했을 때 4.1%p의 성능향상을 얻었으며, 의존 관계명 분포를 반영하는 음절 임베딩을 사용한 bidirectional LSTM-CRFs는 의존 관계명 부착에 최고의 성능인 96.01%(5.21%p 개선)를 내었다.

• Annual Conference on Human and Language Technology
• /
• 2016.10a
• /
• pp.14-19
• /
• 2016

본 논문은 문장에서의 어절 간 의존관계가 성립될 때 의존소와 지배소가 어떠한 관계를 가지는지 의존 관계명을 부착하는 모델을 제안한다. 국내에서 한국어 의존구문분석에 관한 연구가 활발히 진행되고 있지만 의존 관계만을 결과로 제시하고 의존 관계명을 제공하지 않는 경우가 많았다. 따라서 본 논문에서는 의존 경로(Dependency Path)와 음절의 의존 관계명 분포를 반영하는 음절 임베딩를 이용한 의존 관계명 부착 모델을 제안한다. 문장에서 나올 수 있는 최적의 입력 열인 의존 경로(Dependency Path)를 순차 레이블링에서 좋은 성능을 나타내고 있는 bidirectional LSTM-CRFs의 입력 값으로 사용하여 의존 관계명을 결정한다. 제안된 기법은 자질에 대한 많은 노력 없이 의존 경로에 따라 어절 및 음절 단어표상(word embedding)만을 사용하여 순차적으로 의존 관계명을 부착한다. 의존 경로를 사용하지 않고 전체 문장의 어절 순서를 바탕으로 자질을 추출하여 CRFs로 분석한 기존 모델보다 의존 경로를 사용했을 때 4.1%p의 성능향상을 얻었으며, 의존 관계명 분포를 반영하는 음절 임베딩을 사용한 bidirectional LSTM-CRFs는 의존 관계명 부착에 최고의 성능인 96.01%(5.21%p 개선)를 내었다.

• Korean Journal of Remote Sensing
• /
• v.37 no.4
• /
• pp.719-731
• /
• 2021

This study proposes a two-stage hybrid classification model for crop classification using multi-temporal remote sensing images; the model combines feature embedding by using an autoencoder (AE) with a convolutional neural network (CNN) classifier to fully utilize features including informative temporal and spatial signatures. Long short-term memory (LSTM)-based AE (LAE) is fine-tuned using class label information to extract latent features that contain less noise and useful temporal signatures. The CNN classifier is then applied to effectively account for the spatial characteristics of the extracted latent features. A crop classification experiment with multi-temporal unmanned aerial vehicle images is conducted to illustrate the potential application of the proposed hybrid model. The classification performance of the proposed model is compared with various combinations of conventional deep learning models (CNN, LSTM, and convolutional LSTM) and different inputs (original multi-temporal images and features from stacked AE). From the crop classification experiment, the best classification accuracy was achieved by the proposed model that utilized the latent features by fine-tuned LAE as input for the CNN classifier. The latent features that contain useful temporal signatures and are less noisy could increase the class separability between crops with similar spectral signatures, thereby leading to superior classification accuracy. The experimental results demonstrate the importance of effective feature extraction and the potential of the proposed classification model for crop classification using multi-temporal remote sensing images.