We study the core technologies that enable artificial intelligence to
perform complex tasks without human intervention.

A Single Model Ensemble Framework for Neural Machine Translation using Pivot Translation

Hierarchical Retrieval with Evidence Curation for Open-Domain Financial Question Answering on Standardized Documents

Automatic Transmission for LLM Tiers: Optimizing Cost and Accuracy in Large Language Models

CARE: A Framework for Correcting Numerical Hallucinations in LLM-Generated Financial Texts

Cardinality Estimation of LIKE Predicate Queries using Deep Learning

Improving Detail in Pluralistic Image Inpainting with Feature Dequantization

DP-FROST: Differentially Private Fine-tuning of Pre-trained Models with Freezing Model Parameters.

Machine learning-based real-time prediction of duodenal stump leakage from gastrectomy in gastric cancer patients

Semi-supervised Learning for Photovoltaic Cell Defect Detection Using Module and Cell Level Labels

Improving Messenger Phishing Detection Using Heterogeneous Phishing Data

Data Augmentation for Messenger Phishing Detection Using Large Language Models

Beyond Reference: Evaluating High Quality Translations Better than Human References

Improving Domain-Specific ASR with LLM-Generated Contextual Descriptions

Enhancing Low-Resource Fine-Grained Named Entity Recognition by Leveraging Coarse-Grained Datasets

Exploring the Impact of Corpus Diversity on Financial Pretrained Language Models

THUNDER: Named Entity Recognition using a Teacher-student Model with Dual Classifiers for Strong and Weak Supervisions

Cardinality Estimation of Approximate Substring Queries using Deep Learning

Collecting Geospatial Data under Local Differential Privacy with Improving Frequency Estimation

Collecting Geospatial Data with Local Differential Privacy for Personalized Services

Dual Supervision Framework for Relation Extraction with Distant Supervision and Human Annotation

T-REX: A Topic-Aware Relation Extraction Model

TIDY: Publishing a Time Interval Dataset With Differential Privacy

Crowdsourced Truth Discovery in the Presence of Hierarchies for Knowledge Fusion

Integration of graphs from different data sources using crowdsourcing

CoCoaH: 문맥과의 관계를 고려한 혐오 발언 탐지

LLM 생성 경제 텍스트의 수치형 할루시네이션 탐지

자세기반 사람 이미지 생성을 위한 SAPDE 정규화 네트워크 활용

효과적인 스미싱 탐지를 위한 사전 학습된 언어 모델의 활용

대화형 금융정보 검색을 위한 의도 분류

경제 문서에서의 숫자 환각(Hallucination) 현상 완화를 위한 숫자 매칭

KorPatBART: 특허 문서 텍스트 생성을 위한 한국어 특허 도메인 사전학습 BART 모델

태양광 EL 검사에서의 모듈 단위 라벨을 활용한 셀 단위 불량 탐지

자연어 생성 모델을 통한 프롬프트 기반 번역 병렬 데이터 증강

언어생성 모델을 이용한 기업공시 기사 작성

시맨틱 분할 정보를 이용한 장소 분류 모델 성능 개선

세부 분류 개체명 인식을 위한 액티브 러닝 쿼리전략 분석

3차원 컨볼루션과 CBAM을 활용한 폐 결절 분류 모델 성능 개선

약한 지도 학습 기반 태양광 발전시설 고장 탐지

부가정보를 이용한 질확대경검사 딥러닝 모델

Collecting geospatial data with local differential privacy for personalized services

Collecting geospatial data under local differential privacy with improving frequency estimation

Data augmentation for neural machine translation using generative language model

TIDY: Publishing a time interval dataset with differential privacy

T-rex: A topic-aware relation extraction model

Enhancing Low-resource Fine-grained Named Entity Recognition by Leveraging Coarse-grained Datasets

Hierarchical retrieval with evidence curation for open-domain financial question answering on standardized documents

Differentially Private k-Means Clustering based on Dynamic Space Partitioning using a Quad-Tree

Cardinality estimation of like predicate queries using deep learning

Improving messenger phishing detection using heterogeneous phishing data

K-means clustering based data mining system and method using the same

Improving detail in pluralistic image inpainting with feature dequantization

Beyond reference: Evaluating high quality translations better than human references

Kyuseok Shim

Seokjin Oh

Daeyoung Hong

Suyong Kwon

Keonwoong Noh

Jaeyoung Choe

Jiwon Suh

Seongwoong Oh

Seyun Ahn

Jae Hun Chung

Minsam Ko

Changheon Han

Junggab Son

Kyungtae Kang

Woohwan Jung

Geospatial data provides a lot of benefits for personalized services. However, since the geospatial data contains sensitive information about personal activities, collecting the raw data has a potential risk of leaking private information from the collectors. Recently, local differential privacy (LDP), which protects the privacy of users without trusting the collector, has been adopted to preserve privacy in many real applications. However, most of existing LDP algorithms focus on obtaining aggregated values such as mean and histogram from the collected data. In this paper, we investigate the problem of collecting the locations of individual users under LDP, and propose a perturbation mechanism designed carefully to reduce the error of each perturbed location according to the privacy budget and the domain size. In addition, we show the effectiveness of the proposed algorithm through experiments on various real datasets.

Geospatial data provides a lot of benefits for personalized services. However, since the geospatial data contains sensitive information about personal activities, collecting the raw data has a potential risk of leaking private information from the collectors. Recently, local differential privacy (LDP), which protects the privacy of users without trusting the collector, has been adopted to preserve privacy in many real applications. In this paper, we investigate the problem of collecting the locations of individual users under LDP, and propose a perturbation mechanism designed carefully to minimize the expected error of perturbed locations according to the privacy budget and the data domain. The frequency distribution of perturbed locations inevitably has a large error. To tackle the problem, we also propose a postprocessing algorithm to estimate the original frequency distribution of collected data by using convex optimization …

Despite the rapid growth in model architecture, the scarcity of large parallel corpora remains the main bottleneck in Neural Machine Translation. Data augmentation is a technique that enhances the performance of data-hungry models by generating synthetic data instead of collecting new ones. We explore prompt-based data augmentation approaches that leverage large-scale language models such as ChatGPT. To create a synthetic parallel corpus, we compare 3 methods using different prompts. We employ two assessment metrics to measure the diversity of the generated synthetic data. This approach requires no further model training cost, which is mandatory in other augmentation methods like back-translation. The proposed method improves the unaugmented baseline by 0.68 BLEU score.

Approximate substring search is a fundamental operation in many applications including substring matching [13], query refinement suggestion [24], data cleaning [34], named entity recognition [6], finding DNA subsequences [23] and spell checking [28]. With the explosive growth of data, there has been increasing demand for efficient support of approximate substring queries in database systems. The edit distance between strings is the minimum number of single character edits to transform one string to the other [9] and it

Relation extraction (RE) has been extensively studied due to its importance in real-world applications such as knowledge base construction and question answering. Most of the existing works train the models on either distantly supervised data or human-annotated data. To take advantage of the high accuracy of human annotation and the cheap cost of distant supervision, we propose the dual supervision framework which effectively utilizes both types of data. However, simply combining the two types of data to train a RE model may decrease the prediction accuracy since distant supervision has labeling bias. We employ two separate prediction networks HA-Net and DS-Net to predict the labels by human annotation and distant supervision, respectively, to prevent the degradation of accuracy by the incorrect labeling of distant supervision. Furthermore, we propose an additional loss term called disagreement penalty to enable HA-Net to learn from distantly supervised labels. In addition, we exploit additional networks to adaptively assess the labeling bias by considering contextual information. Our performance study on sentence-level and document-level REs confirms the effectiveness of the dual supervision framework.

End-to-end automatic speech recognition (E2E ASR) systems have significantly improved speech recognition through training on extensive datasets. Despite these advancements, they still struggle to accurately recognize domain specific words, such as proper nouns and technical terminologies. To address this problem, we propose a method to utilize the state-of-the-art Whisper without modifying its architecture, preserving its generalization performance while enabling it to leverage descriptions effectively. Moreover, we propose two additional training techniques to improve the domain specific ASR: decoder fine-tuning, and context perturbation. We also propose a method to use a Large Language Model (LLM) to generate descriptions with simple metadata, when descriptions are unavailable. Our experiments demonstrate that proposed methods notably enhance domain-specific ASR accuracy on real-life datasets, with LLM-generated descriptions outperforming human-crafted ones in effectiveness.

Over the past few years, various domain-specific pretrained language models (PLMs) have been proposed and have outperformed general-domain PLMs in specialized areas such as biomedical, scientific, and clinical domains. In addition, financial PLMs have been studied because of the high economic impact of financial data analysis. However, we found that financial PLMs were not pretrained on sufficiently diverse financial data. This lack of diverse training data leads to a subpar generalization performance, resulting in general-purpose PLMs, including BERT, often outperforming financial PLMs on many downstream tasks. To address this issue, we collected a broad range of financial corpus and trained the Financial Language Model (FiLM) on these diverse datasets. Our experimental results confirm that FiLM outperforms not only existing financial PLMs but also general domain PLMs. Furthermore, we provide empirical evidence that this improvement can be achieved even for unseen corpus groups.

Log data from mobile devices generally contain a series of events with temporal information including time intervals which consist of the start and finish times. However, the problem of releasing differentially private time interval datasets has not been tackled yet. A time interval dataset can be represented by a two dimensional (2D) histogram. Most of the methods to publish 2D histograms partition the data into rectangular spaces to reduce the aggregated noise error for range queries. However, the existing algorithms to publish 2D histograms suffer from the structural error when applied to time interval datasets. To reduce the aggregated noise errors and suppress the increase in the structural error, we propose the TIDY (publishing Time Intervals via Differential privacY) algorithm. We use the frequency vectors as a compact representation of the time interval dataset. After applying the Laplace mechanism to the …

Existing works for truth discovery in categorical data usually assume that claimed values are mutually exclusive and only one among them is correct. However, many claimed values are not mutually exclusive even for functional predicates due to their hierarchical structures. Thus, we need to consider the hierarchical structure to effectively estimate the trustworthiness of the sources and infer the truths. We propose a probabilistic model to utilize the hierarchical structures and an inference algorithm to find the truths. In addition, in the knowledge fusion, the step of automatically extracting information from unstructured data (e.g., text) generates a lot of false claims. To take advantages of the human cognitive abilities in understanding unstructured data, we utilize crowdsourcing to refine the result of the truth discovery. We propose a task assignment algorithm to maximize the accuracy of the inferred truths. The performance study with real-life datasets confirms the effectiveness of our truth inference and task assignment algorithms.

Data integration is the process of identifying pairs of records from different databases that refer to the same entity in the real world. It has been extensively studied with regard to entity resolution, record linkage, duplicate detection or network alignment. With the increasing use of crowdsourcing platforms as a means of assessing queries manually at low cost, many studies have begun to consider ways to exploit crowdsourcing systems for efficient data integration. In this paper, we present an efficient algorithm to integrate two graphs collected from different sources using crowdsourcing systems. Given two graphs, we repeatedly select a query node from a graph and request a human annotator to find its matching node from the other graph, which is considered to be the one indicating the same entity as the query node. The proposed method is to choose the query nodes that would increase the precision the most if it is …

Document-level relation extraction (RE) has recently received a lot of attention. However, existing models for document-level RE have similar structures to the models for sentence-level RE. Thus, they still do not consider some unique characteristics of the new problem setting. For example, in Wikipedia, there is a title for each page and it usually represents the topic entity that is mainly described on the page. In many cases, the topic entity is omitted in the text. Thus, existing RE models often fail to find the relations with the omitted topic entity. To tackle the problem, we propose a Topic-aware Relation EXtraction (T-REX) model. To extract the relations with the (possibly omitted) topic entity, the proposed model first encodes the topic entity by aggregating the information of all its mentions in the document. Then it finds the relations between the topic entity and each mention of other entities. Finally, the output layer …

Named Entity Recognition (NER) frequently suffers from the problem of insufficient labeled data, particularly in fine-grained NER scenarios. Although K-shot learning techniques can be applied, their performance tends to saturate when the number of annotations exceeds several tens of labels. To overcome this problem, we utilize existing coarse-grained datasets that offer a large number of annotations. A straightforward approach to address this problem is pre-finetuning, which employs coarse-grained data for representation learning. However, it cannot directly utilize the relationships between fine-grained and coarse-grained entities, although a fine-grained entity type is likely to be a subcategory of a coarse-grained entity type. We propose a fine-grained NER model with a Fine-to-Coarse (F2C) mapping matrix to leverage the hierarchical structure explicitly. In addition, we present an inconsistency filtering method to eliminate coarse-grained entities that are inconsistent with fine-grained entity types to avoid performance degradation. Our experimental results show that our method outperforms both K-shot learning and supervised learning methods when dealing with a small number of fine-grained annotations.

Retrieval-augmented generation (RAG) based large language models (LLMs) are widely used in finance for their excellent performance on knowledge-intensive tasks. However, standardized documents (eg, SEC filing) share similar formats such as repetitive boilerplate texts, and similar table structures. This similarity forces traditional RAG methods to misidentify near-duplicate text, leading to duplicate retrieval that undermines accuracy and completeness. To address these issues, we propose the Hierarchical Retrieval with Evidence Curation (HiREC) framework. Our approach first performs hierarchical retrieval to reduce confusion among similar texts. It first retrieve related documents and then selects the most relevant passages from the documents. The evidence curation process removes irrelevant passages. When necessary, it automatically generates complementary queries to collect missing information. To evaluate our approach, we construct and release a Large-scale Open-domain Financial (LOFin) question answering benchmark that includes 145,897 SEC documents and 1,595 question-answer pairs. Our code and data are available at https://github. com/deep-over/LOFin-bench-HiREC.

최근 공개되는 데이터에 적용하는 다양한 프라이버시 보호 기법들이 연구가 되어왔다. 그 중 차분 프라이버시는 본래의 데이터에 확률적인 노이즈를 더하여 공격자의 사전 지식에 상관없이 개인 정보를 보호한다. 기존 차분 프라이버시를 만족하는 k-평균 클러스터링은 데이터로부터 차분 프라이버시를 만족하는 히스토그램 형태로 바꾼 뒤. k-평균 클러스터링 알고리즘을 수행한다. 하지만 이는 데이터의 분포와 상관없이 등간격으로 히스토그램을 만들기 때문에 노이즈가 삽입되는 버킷이 많아지는 단점이 있다. 이를 해결하기 위해 본 논문에서는 데이터의 분포를 더 적은 버킷으로 나타낼 수 있는 쿼드 트리를 이용하여 히스토그램을 만든 뒤 k-평균을 찾는 알고리즘을 제안한다. 또한, 실험을 통해 기존의 알고리즘보다 더 좋은 성능을 가지는 것을 보인다.

Cardinality estimation of LIKE predicate queries has an important role in the query optimization of database systems. Traditional approaches generally use a summary of text data with some statistical assumptions. Recently, the deep learning model for cardinality estimation of LIKE predicate queries has been investigated. To provide more accurate cardinality estimates and reduce the maximum estimation errors, we propose a deep learning model that utilizes the extended N -gram table and the conditional regression header. We next investigate how to efficiently generate training data. Our LEADER (LikE predicate trAining Data gEneRation) algorithms utilize the shareable results across the relational queries corresponding to the LIKE predicates. By analyzing the queries corresponding to LIKE predicates, we develop an efficient join method and utilize the join order for fast query execution and maximal sharing of …

Recently, as messenger phishing has been occurring more frequently, the need for its detection has increased; however, datasets for messenger phishing detection are publicly unavailable. In this paper, we address the data scarcity problem of the newly collected messenger phishing dataset by leveraging various types of pre-existing auxiliary phishing data. Experimental results demonstrate that the error rate decreased by up to 1.81 % and the F1 score improved by 7.35% when smishing and voice phishing data are used. These findings confirm that integrating heterogeneous phishing data can mitigate the data scarcity problem and enhance messenger phishing detection performance.

A method of performing K-means clustering by a data mining system is provided. The method includes generating a plurality of initial buckets by dividing data including a plurality of points each being expressed in coordinate infor mation, reflecting a count noise in a number of points included in each of the initial buckets and then generating a plurality of new buckets by dividing at least one initial bucket among the initial buckets based on a first threshold and a second threshold, generating a plurality of final buckets from the plurality of initial buckets or the plurality of new buckets, generating a histogram including section information for each of the final buckets and a number of points included in each of the final buckets in which the count noise is reflected, and performing K-means clustering on the histogram based on a number of clusters.

Purpose This study aimed to develop a machine learning (ML) model for real-time prediction of duodenal stump leakage (DSL) following gastrectomy in patients with gastric cancer (GC) using a comprehensive set of clinical variables to improve postoperative outcomes and monitoring efficiency. Methods A retrospective analysis was conducted on 1,107 patients with GC who underwent gastrectomy at Pusan National University Yangsan Hospital between 2019 and 2022. One hundred eighty-nine features were extracted from each patient record, including demographic data, preoperative comorbidities, and blood test outcomes from the subsequent seven postoperative days (POD). Six ML algorithms were evaluated: Logistic Regression (LR), K-nearest neighbors (KNN), Support Vector Machine (SVM), Random Forest (RF), Extreme Gradient Boosting (XGB), and Neural Network (NN). The models predicted DSL occurrence preoperatively and on POD 1, 2, 3, 5, and 7. Performance was assessed using the Area Under the Receiver Operating Characteristic Curve (AUROC) and Recall@K. Results Among the 1,107 patients, 29 developed DSL. XGB demonstrated the highest AUROC score (0.880), followed by RF (0.858), LR (0.823), SVM (0.819), NN (0.753), and KNN (0.726). The RF achieved the best Recall@K score of 0.643. Including additional POD features improved the predictive performance, with the AUROC value increasing to 0.879 on POD 7. The confidence scores of the model indicated that the DSL predictions became more reliable over time. Conclusion The study concluded that ML models, notably the XGB algorithm, can effectively …

Pluralistic Image Inpainting (PII) offers multiple plausible solutions for restoring missing parts of images and has been successfully applied to various applications including image editing and object removal. Recently, VQGANbased methods have been proposed and have shown that they significantly improve the structural integrity in the generated images. Nevertheless, the state-of-the-art VQGANbased model PUT faces a critical challenge: degradation of detail quality in output images due to feature quantization. Feature quantization restricts the latent space and causes information loss, which negatively affects the detail quality essential for image inpainting. To tackle the problem, we propose the FDM (Feature Dequantization Module) specifically designed to restore the detail quality of images by compensating for the information loss. Furthermore, we develop an efficient training method for FDM which drastically …

In Machine Translation (MT) evaluations, the conventional approach is to compare a translated sentence against its human-created reference sentence. MT metrics provide an absolute score (eg, from 0 to 1) to a candidate sentence based on the similarity with the reference sentence. Thus, existing MT metrics give the maximum score to the reference sentence. However, this approach overlooks the potential for a candidate sentence to exceed the reference sentence in terms of quality. In particular, recent advancements in Large Language Models (LLMs) have highlighted this issue, as LLM-generated sentences often exceed the quality of human-written sentences. To address the problem, we introduce the Residual score Metric (ResuMe), which evaluates the relative quality between reference and candidate sentences. ResuMe assigns a positive score to candidate sentences that outperform their reference sentences, and a negative score when they fall short. By adding the residual scores from ResuMe to the absolute scores from MT metrics, it can be possible to allocate higher scores to candidate sentences than what reference sentences are received from MT metrics. Experimental results demonstrate that ResuMe enhances the alignments between MT metrics and human judgments both at the segment-level and the system-level.

AutoAI Lab

교수님 소개

정우환

연구실 소개

연구실 키워드

알럼나이가 간 회사

연구실 지원 방법

대우 조건

연도별 논문 인용수

저널 & 학회 논문 출판 수

연구실 논문 목록

A Single Model Ensemble Framework for Neural Machine Translation using Pivot Translation

Hierarchical Retrieval with Evidence Curation for Open-Domain Financial Question Answering on Standardized Documents

Automatic Transmission for LLM Tiers: Optimizing Cost and Accuracy in Large Language Models

연구실 보유 특허

대학별 입시끈

우리 연구실 모집

교수님 소개

정우환

연구실 키워드

알럼나이가 간 회사

연구실 지원 방법

대우 조건

연도별 논문 인용수

저널 & 학회 논문 출판 수

연구실 논문 목록

A Single Model Ensemble Framework for Neural Machine Translation using Pivot Translation

Hierarchical Retrieval with Evidence Curation for Open-Domain Financial Question Answering on Standardized Documents

Automatic Transmission for LLM Tiers: Optimizing Cost and Accuracy in Large Language Models

연구실 보유 특허

최근 3년간 연구실적 높은 연구실

대학별 입시끈

우리 연구실 모집