Skip to main content

Jong-hyeok Lee

Followers

8

Following

0

Co-author

1

Public Views

Interests

Uploads

Papers by Jong-hyeok Lee

Conceptual Schema Approach to Natural Language Database Access

Proceedings of the Australasian Language …, 2003

Natural language database interfaces require translation knowledge to convert user questions into... more Natural language database interfaces require translation knowledge to convert user questions into formal database queries. Previously, translation knowledge acquisition heavily depends on human specialties such as NLP, DBMS and domain engineering, consequently undermining domain portability. This paper attempts to semi-automatically construct translation knowledge by introducing a physically-derived conceptual database schema, and by simplifying translation knowledge into two structures -classreferring documents and classconstraining selection restrictions. Based on these two structures, this paper proposes a noun translation method that employs an information retrieval framework.

Question Answering Approach Using a WordNet-based Answer Type Taxonomy

Text REtrieval Conference, 2002

POSTECH at TREC 2010 Blog Track: Top Stories Identification

This paper describes our participation in the TREC 2010 Blog Track. For the Top Stories Identific... more This paper describes our participation in the TREC 2010 Blog Track. For the Top Stories Identification Task, we explore the relationship among news events, news stories and blog posts. We first extract important news events from the TRC2 corpus using a probabilistic mixture model. Then, we propose a probabilistic approach to identify top news stories. Furthermore, we use an additional feature that can be useful in identifying top news stories. For the News Blog Post Ranking Task, we apply the Maximal Marginal Relevance method (MMR) to make the aspects of the blog posts more diverse.

Towards Semi-Supervised Learning of Automatic Post-Editing: Data-Synthesis by Infilling Mask with Erroneous Tokens

arXiv (Cornell University), Apr 8, 2022

Semi-supervised learning that leverages synthetic training data has been widely adopted in the fi... more Semi-supervised learning that leverages synthetic training data has been widely adopted in the field of Automatic post-editing (APE) to overcome the lack of human-annotated training data. In that context, data-synthesis methods to create high-quality synthetic data have also received much attention. Considering that APE takes machine-translation outputs containing translation errors as input, we propose a noising-based data-synthesis method that uses a mask language model to create noisy texts through substituting masked tokens with erroneous tokens, yet following the errorquantity statistics appearing in genuine APE data. In addition, we propose corpus interleaving, which is to combine two separate synthetic data by taking only advantageous samples, to further enhance the quality of the synthetic data created with our noising method. Experimental results reveal that using the synthetic data created with our approach results in significant improvements of APE performance upon using other synthetic data created with different existing data-synthesis methods.

Multi-Document Summarization Using Cross-Language Texts

NTCIR, 2004

Without a summarization system in source language, we try to generate a summary in source languag... more Without a summarization system in source language, we try to generate a summary in source language, using translated documents by a machine translator and a summarization system in target language. For summarizing multiple documents translated by a machine translator, we extract important sentences, and remove redundant sentences using an improved term-weighting method. It assigns weights to words, using syntactic information. According to the score of the extracted sentence, we choose sentences, and map them to Japanese sentences in original documents. Finally, we arrange Japanese sentences in chronological order, and report them as the result of our system. We submitted both a short and long type of summary, and the evaluation of our results is not good. However, our approach shows the possibility of multi-documents summarization, using crosslanguage texts.

Bring More Attention to Syntactic Symmetry for Automatic Postediting of High-Quality Machine Translations

arXiv (Cornell University), May 17, 2023

Automatic postediting (APE) is an automated process to refine a given machine translation (MT). R... more Automatic postediting (APE) is an automated process to refine a given machine translation (MT). Recent findings present that existing APE systems are not good at handling highquality MTs even for a language pair with abundant data resources, English-German: the better the given MT is, the harder it is to decide what parts to edit and how to fix these errors. One possible solution to this problem is to instill deeper knowledge about the target language into the model. Thus, we propose a linguistically motivated method of regularization that is expected to enhance APE models' understanding of the target language: a loss function that encourages symmetric self-attention on the given MT. Our analysis of experimental results demonstrates that the proposed method helps improving the state-of-the-art architecture's APE quality for high-quality MTs.

Transformer-based Automatic Post-Editing with a Context-Aware Encoding Approach for Multi-Source Inputs

arXiv (Cornell University), Aug 15, 2019

Recent approaches to the Automatic Post-Editing (APE) research have shown that better results are... more Recent approaches to the Automatic Post-Editing (APE) research have shown that better results are obtained by multi-source models, which jointly encode both source (src) and machine translation output (mt) to produce post-edited sentence (pe). Along this trend, we present a new multi-source APE model based on the Transformer. To construct effective joint representations, our model internally learns to incorporate src context into mt representation. With this approach, we achieve a significant improvement over baseline systems, as well as the state-of-the-art multi-source APE model. Moreover, to demonstrate the capability of our model to incorporate src context, we show that the word alignment of the unknown MT system is successfully captured in our encoding results.

Multi-encoder Transformer Network for Automatic Post-Editing

This paper describes the POSTECH's submission to the WMT 2018 shared task on Automatic Post-Editi... more This paper describes the POSTECH's submission to the WMT 2018 shared task on Automatic Post-Editing (APE). We propose a new neural end-to-end post-editing model based on the transformer network. We modified the encoder-decoder attention to reflect the relation between the machine translation output, the source and the postedited translation in APE problem. Experiments on WMT17 English-German APE data set show an improvement in both TER and BLEU score over the best result of WMT17 APE shared task. Our primary submission achieves -4.52 TER and +6.81 BLEU score on PBSMT task and -0.13 TER and +0.40 BLEU score for NMT task compare to the baseline.

Transformer-based Automatic Post-Editing Model with Joint Encoder and Multi-source Attention of Decoder

This paper describes POSTECH's submission to the WMT 2019 shared task on Automatic Post-Editing (... more This paper describes POSTECH's submission to the WMT 2019 shared task on Automatic Post-Editing (APE). In this paper, we propose a new multi-source APE model by extending Transformer. The main contributions of our study are that we 1) reconstruct the encoder to generate a joint representation of translation (mt) and its src context, in addition to the conventional src encoding and 2) suggest two types of multi-source attention layers to compute attention between two outputs of the encoder and the decoder state in the decoder. Furthermore, we train our model by applying various teacher-forcing ratios to alleviate exposure bias. Finally, we adopt the ensemble technique across variations of our model. Experiments on the WMT19 English-German APE data set show improvements in terms of both TER and BLEU scores over the baseline. Our primary submission achieves -0.73 in TER and +1.49 in BLEU compared to the baseline, and ranks second among all submitted systems.

Unlimited Vocabulary Grapheme to Phoneme Conversion for

This paper describes a grapheme-to-phoneme conversion method using phoneme connectivity and CCV c... more This paper describes a grapheme-to-phoneme conversion method using phoneme connectivity and CCV conversion rules. The method consists of mainly four modules including morpheme normalization, phrase-break detection, morpheme to phoneme conversion and phoneme connectivity check. The morpheme normalization is to replace non-Korean symbols into standard Korean graphemes. The phrase-break detector assigns phrase breaks using part-of-speech (POS) information. In the morpheme-to-phoneme conversion module, each morpheme in the phrase is converted into phonetic patterns by looking up the morpheme phonetic pattern dictionary which contains candidate phonological changes in boundaries of the morphemes. Graphemes within a morpheme are grouped into CCV patterns and converted into phonemes by the CCV conversion rules. The phoneme connectivity table supports grammaticality checking of the adjacent two phonetic morphemes. In the experiments with a corpus of 4,973 sentences, we achieved 99.9% of the graphemeto-phoneme conversion performance and 97.5% of the sentence conversion performance. The full Korean TTS system is now being implemented using this conversion method.

Unlimited Vocabulary Grapheme to Phoneme Conversion for Korean TTS

arXiv (Cornell University), Jun 10, 1998

This paper describes a grapheme-to-phoneme conversion method using phoneme connectivity and CCV c... more This paper describes a grapheme-to-phoneme conversion method using phoneme connectivity and CCV conversion rules. The method consists of mainly four modules including morpheme normalization, phrase-break detection, morpheme to phoneme conversion and phoneme connectivity check. The morpheme normalization is to replace non-Korean symbols into standard Korean graphemes. The phrase-break detector assigns phrase breaks using part-of-speech (POS) information. In the morpheme-to-phoneme conversion module, each morpheme in the phrase is converted into phonetic patterns by looking up the morpheme phonetic pattern dictionary which contains candidate phonological changes in boundaries of the morphemes. Graphemes within a morpheme are grouped into CCV patterns and converted into phonemes by the CCV conversion rules. The phoneme connectivity table supports grammaticality checking of the adjacent two phonetic morphemes. In the experiments with a corpus of 4,973 sentences, we achieved 99.9% of the graphemeto-phoneme conversion performance and 97.5% of the sentence conversion performance. The full Korean TTS system is now being implemented using this conversion method.

Combination Approaches in Information Retrieval: Words vs. N-grams and Query Translation vs. Document Translation

NTCIR, 2004

This paper reports our proposal and experimental results at the NTCIR-4 CLIR task. For monolingua... more This paper reports our proposal and experimental results at the NTCIR-4 CLIR task. For monolingual information retrieval, we use a combination strategy that integrates words and n-grams at the ranked list level. In combining words and n-grams, we concentrate on generating several ranked lists showing different retrieval characteristics on word and n-gram indexes by incorporating feedback schemes. For cross-language information retrieval, we attempt a dictionary-based bi-directional combination of query translation and document translation. For both query translation and document translation, their naïve translation is used. Experimental evaluations on CJK monolingual and KC/KJ cross-language retrieval give promising results on our combination approaches: words vs. ngrams, and query translation vs. document translation.

Annotation Guidelines for Chinese-Korean Word Alignment

Language Resources and Evaluation, May 1, 2008

For a language pair such as Chinese and Korean that belong to entirely different language familie... more For a language pair such as Chinese and Korean that belong to entirely different language families in terms of typology and genealogy, finding the correspondences is quite obscure in word alignment. We present annotation guidelines for Chinese-Korean word alignment through contrastive analysis of morpho-syntactic encodings. We discuss the differences in verbal systems that cause most of linking obscurities in annotation process. Systematic comparison of verbal systems is conducted by analyzing morpho-syntactic encodings. The viewpoint of grammatical category allows us to define consistent and systematic instructions for linguistically distant languages such as Chinese and Korean. The scope of our guidelines is limited to the alignment between Chinese and Korean, but the instruction methods exemplified in this paper are also applicable in developing systematic and comprehensible alignment guidelines for other languages having such different linguistic phenomena.

POSTECH's Statistical Machine Translation Systems for NTCIR-9 PatentMT Task (English-to-Japanese)

NTCIR, 2011

Subtopic mining using simple patterns and hierarchical structure of subtopic candidates from web documents

Information Processing and Management, Nov 1, 2015

The intention gap between users and queries results in ambiguous and broad queries. To solve thes... more The intention gap between users and queries results in ambiguous and broad queries. To solve these problems, subtopic mining has been studied, which returns a ranked list of possible subtopics according to their relevance, popularity, and diversity. This paper proposes a novel method to mine subtopics using simple patterns and a hierarchical structure of subtopic candidates. First, relevant and various phrases are extracted as subtopic candidates using simple patterns based on noun phrases and alternative partial-queries. Second, a hierarchical structure of the subtopic candidates is constructed using sets of relevant documents from a web document collection. Finally, the subtopic candidates are ranked considering a balance between popularity and diversity using this structure. In experiments, our proposed methods outperformed the baselines and even an external resource based method at high-ranked subtopics, which shows that our methods can be effective and useful in various search scenarios like result diversification.

Semantic Relation Extraction using Pattern Pairs Sharing a Term

Journal of KIISE:Computing Practices and Letters, 2009

Constructing an ontology using a mass corpus begins with an automatic semantic relation extractio... more Constructing an ontology using a mass corpus begins with an automatic semantic relation extraction. A general method regards words appearing between terms as patterns which are used to extract semantic relations. However, previous approaches consider only one sentence to extract a pattern, so they cannot extract semantic relations for terms in different sentences. This paper proposes a semantic relation extraction method using pairs of patterns sharing a term, where each pattern is extracted using one of the seed term pair satisfying the target relation. In our experiments, we achieved the accuracy 83.75% improving previous methods by 7.5% in is- relation and the accuracy 83.75% improved by 5% in part-of relation. We also present a possibility of improving the recall by the relative recall.

Subtopic Mining Based on Head-Modifier Relation and Co-occurrence of Intents Using Web Documents

Springer eBooks, 2013

This paper proposes a method that mines subtopics using the head-modifier relation and co-occurre... more This paper proposes a method that mines subtopics using the head-modifier relation and co-occurrence of users' intents from web documents in Japanese. We extracted subtopics using the simple patterns based on the head-modifier relation between the query and its adjacent words, and returned the ranked list of subtopics by the proposed score equation. We re-ranked subtopics according to the intent co-occurrence measure. Our method achieved good performance than the baseline methods and suggested queries from the major web search engine. The results of our method will be useful in various search scenarios, such as query suggestion and result diversification.

A discriminative reordering parser for IWSLT 2013

We participated in the IWSLT 2013 Evaluation Campaign for the MT track for two official direction... more We participated in the IWSLT 2013 Evaluation Campaign for the MT track for two official directions: German$English. Our system consisted of a reordering module and a statistical machine translation (SMT) module under a pre-ordering SMT framework. We trained the reordering module using three scalable methods in order to utilize training instances as many as possible. The translation quality of our primary submissions were comparable to that of a hierarchical phrasebased SMT, which usually requires a longer time to decode.

Forest-to-string translation using binarized dependency forest for IWSLT 2012 OLYMPICS task

We participated in the OLYMPICS task in IWSLT 2012 and submitted two formal runs using a forest-t... more

Transformer-based Screenplay Summarization Using Augmented Learning Representation with Dialogue Information

Proceedings of the Third Workshop on Narrative Understanding, 2021

Screenplay summarization is the task of extracting informative scenes from a screenplay. The scre... more Screenplay summarization is the task of extracting informative scenes from a screenplay. The screenplay contains turning point (TP) events that change the story direction and thus define the story structure decisively. Accordingly, this task can be defined as the TP identification task. We suggest using dialogue information, one attribute of screenplays, motivated by previous work that discovered that TPs have a relation with dialogues appearing in screenplays. To teach a model this characteristic, we add a dialogue feature to the input embedding. Moreover, in an attempt to improve the model architecture of previous studies, we replace LSTM with Transformer. We observed that the model can better identify TPs in a screenplay by using dialogue information and that a model adopting Transformer outperforms LSTM-based models.