Search

Sentence compression is a natural language-processing task that produces a short paraphrase of an input sentence by deleting words from the input sentence while ensuring grammatical correctness and preserving meaningful core information. This study introduces a graph convolutional network (GCN) into a sentence compression task to encode syntactic information, such as dependency trees. As we upgrade the GCN to activate a directed edge, the compression model with the GCN layers can distinguish between parent and child nodes in a dependency tree when aggregating adjacent nodes. Furthermore, by increasing the number of GCN layers, the model can gradually collect high-order information of a dependency tree when propagating node information through the layers. We implement a sentence compression model for Korean and English, respectively. This model consists of three components: pre-trained BERT model, GCN layers, and a scoring layer. The scoring layer can determine whether a word should remain in a compressed sentence by relying on the word vector containing contextual and syntactic information encoded by BERT and GCN layers. To train and evaluate the proposed model, we used the Google sentence compression dataset for English and a Korean sentence compression corpus containing about 140,000 sentence pairs for Korean. The experimental results demonstrate that the proposed model achieves state-of-the-art performance for English. To the best of our knowledge, this sentence compression model based on the deep learning model trained with a large-scale corpus is the first attempt for Korean.

Korean and Japanese have different writing scripts but share the same Subject-Object-Verb (SOV) word order. In this study, we pre-train a language-generation model using a Masked Sequence-to-Sequence pre-training (MASS) method on Korean and Japanese monolingual corpora. When building the pre-trained generation model, we allow the smallest number of shared vocabularies between the two languages. Then, we build an unsupervised Neural Machine Translation (NMT) system between Korean and Japanese based on the pre-trained generation model. Despite the different writing scripts and few shared vocabularies, the unsupervised NMT system performs well compared to other pairs of languages. Our interest is in the common characteristics of both languages that make the unsupervised NMT perform so well. In this study, we propose a new method to analyze cross-attentions between a source and target language to estimate the language differences from the perspective of machine translation. We calculate cross-attention measurements between Korean–Japanese and Korean–English pairs and compare their performances and characteristics. The Korean–Japanese pair has little difference in word order and a morphological system, and thus the unsupervised NMT between Korean and Japanese can be trained well even without parallel sentences and shared vocabularies.

This work presents a new method that can predict, without the use of training data, whether an input essay is written on a given topic. ConceptNet is a common-sense knowledge base that is generated automatically from sentences that are extracted from a variety of document types. An essay prompt is the topic that an essay should be written about. The method that is proposed in this paper uses ConceptNet and an essay prompt to decide whether or not an input essay is off-topic. We introduce a way to find the shortest path between two nodes on ConceptNet, as well as a way to calculate the semantic similarity between two nodes. Not only an essay prompt but also a student's essay can be represented by concept nodes in ConceptNet. The semantic similarity between the concepts that represent an essay prompt and the other concepts that represent a student's essay can be used for a calculation to rank "on-topicness" ; if a low ranking is derived, an essay is regarded as off-topic. We used eight different essay prompts and a student-essay collection for the performance evaluation, whereby our proposed method shows a performance that is better than those of the previous studies. As ConceptNet enables the conduction of a simple text inference, our new method looks very promising with respect to the design of an essay prompt for which a simple inference is required.

In order to train a supervised machine learning algorithm, it is necessary to have non-biased labels and a sufficient amount of training data. However, it is difficult to collect the required non-biased labels and a sufficient amount of training data to develop an automatic English Composition scoring system. In addition, an English writing assessment is carried out using a multi-faceted evaluation of the overall level of the answer. Therefore, it is difficult to choose an appropriate machine learning algorithm for such work. In this paper, we show that it is possible to alleviate these problems through ensemble learning. The results of the experiment indicate that the ensemble technique exhibited an overall performance that was better than that of other algorithms.

In this paper, we present a method that can detect context-sensitive word errors and generate correction candidates. Spelling error detection is one of the most widespread research topics, however, the approach proposed in this paper is adjusted for an automated English scoring system. A common strategy in context-sensitive word error detection is using a pre-defined confusion set to generate correction candidates. We automatically generate a confusion set in order to consider the characteristics of sentences written by second-language learners. We define a word error that cannot be detected by a conventional grammar checker because of part-of-speech ambiguity, and propose how to detect the error and generate correction candidates for this kind of error. An experiment is performed on the English writings composed by junior-high school students whose mother tongue is Korean. The f1 value of the proposed method is 70.48%, which shows that our method is promising comparing to the current-state-of-the art.Keywords:Context-sensitive Word Error, Error Detection/Correction, Confusion Set, English Automatic Scoring, Grammar-level Word Error

In this paper, we introduce an automatic scoring system for middle-school level writing test based on using machine learning techniques. We discuss overall process and features for building an automatic English writing scoring system. A "concept answer" which represents an abstract meaning of text is newly introduced in order to evaluate the elaboration of a student's answer. In this work, multiple machine learning algorithms are adopted for scoring English writings. We suggest a decision process "optimal combination" which optimally combines multiple outputs of machine learning algorithms and generates a final single output in order to improve the performance of the automatic scoring. By experiments with actual test data, we evaluate the performance of overall automated English writing scoring system.

In this paper, we propose an approach that automatically aligns words extracted from paired sentences in English- Korean parallel corpus. Word-to-word mappings from aligned sentences are easily accomplished if a bilingual lexicon is provided. However, newly coined words cannot be aligned automatically since they are not registered in the lexicon. For automated alignment of newly identified words, we introduce new heuristics to align maximal number of words, using structural transfer patterns between the two languages. Provided maximum word alignments together with paired sentences in a parallel corpus, Korean learners of English can benefit from recognizing distinctions on sentential structures as well as vocabularies between the two languages.

We analyze public opinion via a sentiment analysis of tweets collected by using recent topic keywords extracted from newspaper articles. Newspaper articles collected within a certain period of time are clustered by using K-means algorithm and topic keywords for each cluster are extracted by using term frequency. A sentiment analyzer learned by a machine learning method can classify tweets according to their polarity values. We have an assumption that tweets collected by using these topic keywords deal with the same topics as the newspaper articles mentioned if the tweets and the newspapers are generated around the same time. and we tried to verify the validity of this assumption.

This paper presents a novel approach to automatically generate Korean multiword sentiment expressions by using a seed sentiment lexicon and a large-scale domainspecific corpus. A multiword sentiment expression consists of a seed sentiment word and its contextual words occurring adjacent to the seed word. The multiword sentiment expressions that are the focus of our study have a different polarity from that of the seed sentiment word. The automatically extracted multiword sentiment expressions show that 1) the contextual words should be defined as a part of a multiword sentiment expression in addition to their corresponding seed sentiment word, 2) the identified multiword sentiment expressions contain various indicators for polarity shift that have rarely been recognized before, and 3) the newly recognized shifters contribute to assigning a more accurate polarity value. The empirical result shows that the proposed approach achieves improved performance of the sentiment analysis system that uses an automatically generated lexicon.

A set of bilingual terms is one of the most important factors in building language-related applications such as a machine translation system and a cross-lingual information system. In this paper, we introduce a new approach that automatically extracts candidates of English-Korean bilingual terms by using a bilingual parallel corpus and a basic English-Korean lexicon. This approach can be useful even though the size of the parallel corpus is small. A sentence alignment is achieved first for the document-level parallel corpus. We can align words between a pair of aligned sentences by referencing a basic bilingual lexicon. For unaligned words between a pair of aligned sentences, several assumptions are applied in order to align bilingual term candidates of two languages. A location of a sentence, a relation between words, and linguistic information between two languages are examples of the assumptions. An experimental result shows approximately 71.7% accuracy for the English-Korean bilingual term candidates which are automatically extracted from 1,000 bilingual parallel corpus.

The goal of this research is to develop an error-driven noun-connection rules which is used forbreaking complicate nouns in Korean morphology analysis module. We collected complicate nouns from Web sites, and analyzed them by CnuMa. Whenever we find errors from outputs of the analyzer, we writenoun-connection rules to correct the errors. The noun-connection rules are devised by considering left/rightcontexts in compound nouns. The error-driven noun-connection rules are helpful in improving precision andrecall of a Korean morphology analyzer, CnuMa by 2.8% and 10.8%, respectively. Key words: morphological analysis, nouns connection table, compound noun †교신저자(한국교통대학교 컴퓨터정보공학과, E-mail: leesw@ut.ac.kr, Tel: 043-841-5464)1 충남대학교 정보통신공학과, E-mail: kjoolee@cnu.ac.kr, Tel: 042-821-5662 1. 서 론 한국어 형태소 분석 기술은 검색엔진의 색인어 추출시스템, 자동 기계번역 시스템, 정보추출 시스템, 자동 문서 클러스터링 등 자연언어처리 기술 응용 분야에서 가장 기본이 되는 요소기술이며 형태소 분석의 오류가 그 응용프로그램의 오류에 직결되므로 상당한 정확성이 요구된다. 무한한 복합명사 생성이 가능한 한국어 특성 때문에 복합명사 분석은 한국어 형태소 분석에서 가장 어려운 문제 중의 하나이다[1].형태소 분석 방법은 최장일치 head-tail 분석법[2], CYK 알고리즘에 기반한 tabular parsing과 접속정보를 이용한 방법[3]이 있으며, 음절 정보를 이용한 방법[4]도 있다. 대부분의 형태소 분석 방법에서 사용하는 가장 중요한 정보는 각 형태소의 배열 규칙을 표현하고 있는 접속정보이다. 이 형태소별 접속정보가 올바른 품사열을 결정하기 위해 사용된다. 일반적으로, 접속정보는 대량의 말뭉치에서 자동으로 추출하거나 수동으로 구축하여 사용한다. 이와 같은 접속정보 중, 각 형태소별로 왼쪽에 나열될 수 있는 형태소의 종류와 오른쪽에 나열될 수 있는 형태소의 종류를 정리해 놓은 것이

Recently, statistical machine translation (SMT) has received many attention with ease of its implementation and maintenance. The goal of our works is to build bidirectional Korean-Japanese SMT system by using MOSES [1] system. We use Korean-Japanese bilingual corpus which is aligned per sentence to train the translation model and use a large raw corpus in each language to train each language model. The proposed system shows results comparable to those of a rule-based machine translation system. Most of errors are caused by noises occurred in each processing stage.

This paper presents an automated scoring system which grades students' English writing tests. The system provides a score and diagnostic feedback to students without human's efforts. Target users are Korean students in junior high schools who learn English as a second language. The system takes a single English sentence as its input. Dealing with a single sentence as an input has some advantages on comparing the input with the answers given by human teachers and giving detailed feedback to the students. The system was developed and tested with the real test data collected through English tests given to third grade students in junior high school. Scoring requires two steps of the process. The first process is analyzing the input sentence in order to detect possible errors, such as spelling errors and syntactic errors. The second process is comparing the input sentence with given answers to identify the differences as errors. To evaluate the performance of the system, the output produced by the system is compared with the result provided by human raters. The score agreement value between a human rater and the system is quite close to the value between two human raters.

In this paper, we present a system that classifies comments on a news article into a user opinion called a polarity (positive or negative). The system is a kind of document classification system for comments and is based on machine learning techniques like support vector machine. Unlike normal documents, comments have their body that can influence classifying their opinions as polarities. In this paper, we propose a feature weighting scheme using such characteristics of comments and several resources for opinion classification. Through our experiments, the weighting scheme have turned out to be useful for opinion classification in comments on Korean news articles. Also Korean character n-grams (bigram or trigram) have been revealed to be helpful for opinion classification in comments including lots of Internet words or typos. In the future, we will apply this scheme to opinion analysis of comments of product reviews as well as news articles.

In this paper, we introduce a system that summarizes product evaluation through linguistic analysis to effectively utilize explosively increasing product reviews. Our system analyzes polarities of product reviews by product features, based on which customers evaluate each product like `design` and `material` for a skirt product category. The system shows to customers a graph as a review summary that represents percentages of positive and negative reviews. We build an opinion word dictionary for each product feature through context based automatic expansion with small seed words, and judge polarity of reviews by product features with the extracted dictionary. In experiment using product reviews from online shopping malls, our system shows average accuracy of 69.8% in extracting judgemental word dictionary and 81.8% in polarity resolution for each sentence.

The proposed automated scoring system for English writing tests provides an assessment result including a score and diagnostic feedback to test-takers without human's efforts. The system analyzes an input sentence and detects errors related to spelling, syntax and content similarity. The scoring model has adopted one of the statistical approaches, a regression tree. A scoring model in general calculates a score based on the count and the types of automatically detected errors. Accordingly, a system with higher accuracy in detecting errors raises the accuracy in scoring a test. The accuracy of the system, however, cannot be fully guaranteed for several reasons, such as parsing failure, incompleteness of knowledge bases, and ambiguous nature of natural language. In this paper, we introduce an error-weighting technique, which is similar to term-weighting widely used in information retrieval. The error-weighting technique is applied to judge reliability of the errors detected by the system. The score calculated with the technique is proven to be more accurate than the score without it.

In this paper, we introduce an automatic product feature extracting system that improves the efficiency of product review analysis. Our system consists of 2 parts: a review collection and correction part and a product feature extraction part. The former part collects reviews from internet shopping malls and revises spoken style or ungrammatical sentences. In the latter part, product features that mean items that can be used as evaluation criteria like 'size' and 'style' for a skirt are automatically extracted by utilizing term statistics in reviews and web documents on the Internet. We choose nouns in reviews as candidates for product features, and calculate degree of association between candidate nouns and products by combining inner association degree and outer association degree. Inner association degree is calculated from noun frequency in reviews and outer association degree is calculated from co-occurrence frequency of a candidate noun and a product name in web documents. In evaluation results, our extraction method showed an average recall of 90%, which is better than the results of previous approaches.Keywords:Product Review, Product Feature Extraction, Feature Based Summarization, Term Statistics, Electronic Commerce

The purpose of developing an automated scoring system for English composition is to score English writing tests and to give diagnostic feedback to the test-takers without human`s efforts. The system developed through our research detects grammatical errors of a single sentence on morphological, syntactic and semantic stages, respectively, and those errors are calculated into the final score. The error detecting stages are independent from one another, which causes duplicating the identical errors with different labels at different stages. These duplicated errors become a hindering factor to calculating an accurate score. This paper presents a solution to detecting the duplicated errors and improving an accuracy in calculating the final score by eliminating one of the errors.

The purpose of developing an automated scoring system for English composition is to score the tests for writing English sentences and to give feedback on them without human`s efforts. This paper presents an automated system to score English composition, whose input is a single sentence, not an essay. Dealing with a single sentence as an input has some advantages on comparing the input with the given answers by human teachers and giving detailed feedback to the test takers. The system has been developed and tested with the real test data collected through English tests given to the third grade students in junior high school. Two steps of the process are required to score a single sentence. The first process is analyzing the input sentence in order to detect possible errors, such as spelling errors, syntactic errors and so on. The second process is comparing the input sentence with the given answer to identify the differences as errors. The results produced by the system were then compared with those provided by human raters.

This paper presents an automated English scoring system designed to help non-native speakers of English, Korean-speaking learners in particular. The system is developed to help the 3rd grade students in junior high school improve their English grammar skills. Without human`s efforts, the system identifies grammar errors in English sentences, provides feedback on the detected errors, and scores the sentences. Detecting grammar errors in the system requires implementing a special type of rules in addition to the rules to parse grammatical sentences. Error production rules are implemented to analyze ungrammatical sentences and recognize syntactic errors. The rules are collected from the junior high school textbooks and real student test data. By firing those rules, the errors are detected followed by setting corresponding error flags, and the system continues the parsing process without a failure. As the final step of the process, the system scores the student sentences based on the errors detected. The system is evaluated with real English test data produced by the students and the answers provided by human teachers.

In this paper, we present preliminary work on recognizing affect from a Korean textual document by using a manually built affect lexicon and adopting natural language processing tools. A manually built affect lexicon is constructed in order to be able to detect various emotional expressions, and its entries consist of emotion vectors. The natural language processing tools analyze an input document to enhance the accuracy of our affect recognizer. The performance of our affect recognizer is evaluated through automatic classification of song lyrics according to moods.

In this paper, we present a Korean to Chinese/English/Japanese multilingual Machine Translation (MT) system of closed captions for Digital Television (DTV). Preliminary experiments of our closed caption translation with existing base MT systems had shown unsatisfactory result. In order to achieve more accurate translation with the base MT systems, we adopted live resources of multilingual Named Entities and their translingual equivalences from the Web. We also utilize the program information, which the terrestrial broadcasters offer through DTV transport stream, in order to use program specific dictionaries, including the names of characters, locations and organizations. Two more components are adopted for reducing the ambiguities of parsing and word sense disambiguation; sentence simplification for long sentence segmentation and dynamic domain identification for automatic domain dictionary stacking. With these integrated approaches, we could raise the Mean Opinion Score (MOS) of translation accuracy by 0.40 higher than the base MT systems.

As the limited resolution and small screen restrict the amount of information to be displayed on PDAs (personal digital assistants), compressed display techniques can be effectively used. This paper presents the development of an intelligent compressed display system for 240/spl times/320 resolution PDAs that are most commonly used model. The proposed system automatically generates compressed documents fit to the 240/spl times/320 resolution from any size input data. Through experiments with extensive real world dataset, the proposed compression system has been shown to be promising.

In this paper, we present a pilot system that tn compress a Korean sentence automatically using knowledge extracted from news articles and their headlines. A sot of compressed sentences can be presented as an abstraction of a document. As a compressed sentence is of headline-style, it could be easily displayed on small devices, such as mobile phones and other handhold devices. Our compressing system has shown to be promising through a preliminary experiment.

It is not always possible to find a global parse for an input sentence owing to problems such as errors of a sentence, incompleteness of lexicon and grammar. Partial parsing is an alternative approach to respond to these problems. Partial parsing techniques try to recover syntactic information efficiently and reliably by sacrificing completeness and depth of analysis. One of the difficulties in partial parsing is how the grammar might be automatically extracted. In this paper we present a method of automatically extracting partial parsing rules from a tree-annotated corpus using the decision tree method. Our goal is deterministic global parsing using partial parsing rules, in other words, to extract partial parsing rules with higher accuracy and broader expansion. First, we define a rule template that enables to learn a subtree for a given substring, so that the resultant rules can be more specific and stricter to apply. Second, rule candidates extracted from a training corpus are enriched with contextual and lexical information using the decision tree method and verified through cross-validation. Last, we underspecify non-deterministic rules by merging substructures with ambiguity in those rules. The learned grammar is similar to phrase structure grammar with contextual and lexical information, but allows building structures of depth one or more. Thanks to automatic learning, the partial parsing rules can be consistent and domain-independent. Partial parsing with this grammar processes an input sentence deterministically using longest-match heuristics, and recursively applies rules to an input sentence. The experiments showed that the partial parser using automatically extracted rules is not only accurate and efficient but also achieves reasonable coverage for Korean.

To understand complex signaling pathways and networks, it is necessary to develop a formal and structured representation of the available information in a format suitable for analysis by software tools. Due to the complexity and incompleteness of the current biological knowledge about cell signaling, such a device must be able to represent cellular pathways at differing levels of details, one level of information abstract enough to convey an essential signaling flow while hiding its details and another level of information detailed enough to explain the underlying mechanisms that account for the signaling flow described at a more abstract level. We have defined a formal ontology for cell-signaling events that allows us to describe these cellular pathways at various levels of abstraction. Using this formal representation, ROSPath (reactive oxygen species-mediated signaling pathway) database system has been implemented and made available on the web (rospath.ewha.ac.kr). ROSPath is a database system for reactive oxygen species (ROS)-mediated cell signaling pathways and signaling processes in molecular detail, which facilitates a comprehensive understanding of the regulatory mechanisms in signaling pathways. ROSPath includes growth factor-, stress-, and cytokine-induced signaling pathways containing about 500 unique proteins (mostly mammalian) and their related protein states, protein complexes, protein complex states, signaling interactions, signaling steps, and pathways. It is a web-based structured repository of information on the signaling pathways of interest and provides a means for managing data produced by large-scale and high-throughput techniques such as proteomics. Also, software tools are provided for querying, displaying, and analyzing pathways, thus furnishing an integrated web environment for visualizing and manipulating ROS-mediated cell-signaling events.

A genre or a style is another view of documents different from a subject or a topic. The style is also a criterion to classify the documents. There have been several studies on detecting a style of textual documents. However, only a few of them dealt with web documents. In this paper we suggest sets of features to detect styles of web documents. Web documents are different from textual documents in that Dey contain URL and HTML tags within the pages. We introduce the features specific to web documents, which are extracted from URL and HTML tags. Experimental results enable us to evaluate their characteristics and performances

In this paper, we propose a modified unsupervised linear alignment algorithm for building an aligned corpus. The original algorithm inserts null characters into both of two aligned strings (source string and target string), because the two strings are different from each other in length. This can cause some difficulties like the search space explosion for applications using the aligned corpus with null characters and no possibility of applying to several machine learning algorithms. To alleviate these difficulties, we modify the algorithm not to contain null characters in the aligned source strings. We have shown the usability of our approach by applying it to different areas such as Korean-English back-trans literation, English grapheme-phoneme conversion, and Korean morphological analysis.

We propose a format of a binary phrase structure grammar with composite labels. The grammar adopts binary rules so that the dependency between two sub-trees can be represented in the label of the tree. The label of a tree is composed of two attributes, each of which is extracted from each sub-tree, so that it can represent the compositional information of the tree. The composite label is generated from part-of-speech tags using an automatic labeling algorithm. Since the proposed rule description scheme is binary and uses only part-of-speech information, it can readily be used in dependency grammar and be applied to other languages as well. In the best-1 context-free cross validation on 31,080 tree-tagged corpus, the labeled precision is 79.30%, which outperforms phrase structure grammar and dependency grammar by 5% and by 4%, respectively. It shows that the proposed rule description scheme is effective for parsing Korean.

In this paper, we present a Korean partial parser based on rules, which is used for running applications such as a grammar checker and a machine translation. Basically partial parsers construct one or more morphemes and/or words into one syntactical unit, but not complete syntactic trees, and accomplish some additional operations for syntactical parsing. The system described in this paper adopts a set of about 140 manually-written rules for partial parsing. Each rule consists of conditional statements and action statement that defines which one is head node and also describes an additional action to do if necessary. To observe that this approach can improve the efficiency of overall processing, we make simple experiments. The experimental results have shown that the average number of edges generated in processing without the partial parser is about 2 times more than that with the partial parser.

Korean delimits words by white-space like English, but words In Korean Is a little different in structure from those in English. Words in English generally consist of one word, but those in Korean are composed of one word and/or morpheme or more. Because of this difference, a word between white-spaces is called an Eojeol in Korean. We propose a method for segmenting and classifying Korean words and/or morphemes based on syllables using an instance-based learning. In this paper, elements of feature sets for the instance-based learning are one previous syllable, one current syllable, two next syllables, a final consonant of the current syllable, and two previous categories. Our method shows more than 97% of the F-measure of word segmentation using ETRI corpus and KAIST corpus.

Since a natural language has inherently structural ambiguities, one of the difficulties of parsing is resolving the structural ambiguities. Recently, a probabilistic approach to tackle this disambiguation problem has received considerable attention because it has some attractions such as automatic learning, wide-coverage, and robustness. In this paper, we focus on Korean probabilistic parsing model using head co-occurrence. We are apt to meet the data sparseness problem when we`re using head co-occurrence because it is lexical. Therefore, how to handle this problem is more important than others. To lighten the problem, we have used the restricted and simplified phrase-structure grammar and back-off model as smoothing. The proposed model has showed that the accuracy is about 84%.

In this paper, we introduce a method to represent phrase structure grammars for building a large annotated corpus of Korean syntactic trees. Korean is different from English in word order and word compositions. As a result of our study, it turned out that the differences are significant enough to induce meaningful changes in the tree annotation scheme for Korean with respect to the schemes for English. A tree annotation scheme defines the grammar formalism to be assumed, categories to be used, and rules to determine correct parses for unsettled issues in parse construction. Korean is partially free in word order and the essential components such as subjects and objects of a sentence can be omitted with greater freedom than in English. We propose a restricted representation of phrase structure grammar to handle the characteristics of Korean more efficiently. The proposed representation is shown by means of an extensive experiment to gain improvements in parsing time as well as grammar size. We also describe the system named Teb that is a software environment set up with a goal to build a tree annotated corpus of Korean containing more than one million units.

An automatic document summarization is one of the essential techniques to display on small devices such as mobile phones and other handheld devices. Most researches in automatic document summarization have focused on extraction of sentences. Sentences extracted as a summary are so long that even a summary is not easy to be displayed in a small device. Therefore, compressing sentences is practically helpful for displaying in a small device. In this paper, we present a pilot system that can automatically compress a Korean sentence using the knowledge extracted from news articles and their headlines. A compressed sentence generated by our system resembles a headline of news articles, so it can be one of the briefest forms preserving the core meaning of an original sentence. Our compressing system has shown to be promising through a preliminary experiment.

A genre or a style is another view of documents different from a subject or a topic. The genre is also a criterion to classify the documents. There have been several studies on detecting a genre of textual documents. However, only a few of them dealt with web documents. In this paper we suggest sets of features to detect genres of web documents. Web documents are different from textual documents in that they contain URL and HTML tags within the pages. We introduce the features specific to web documents, which are extracted from URL and HTML tags. Experimental results enable us to evaluate their characteristics and performances.

In this paper, we propose a robust parser which can parse extragrammatical sentences. This parser can recover them using only syntactic information. It can be easily modified and extended because it utilize only syntactic information., Comment: 6 pages LaTeX, uses eaclap.sty, to appear in EACL-95.

Recent pervasiveness of mobile computing becomes an attractive motivation for EFL learners. Korea provides an excellent environment for using mobile devices such as smartphones and tablet PCs, with easy access to wireless networks. This offers the learners a chance to practice writing on-the-move. This paper introduces a design of a mobile-based tool to assist the learners of beginning to intermediate levels in improving English writing skills. It provides two types of learning mode: 1) the learners select grammatical categories for which the tool provides a brief description and exercise questions, and 2) they choose one of the suggested Korean sentences and write a corresponding English sentence to be evaluated and provided with instructional feedback.