Earticle

효과적인 분절을 통한 양질의 입력 자질 구성은 언어모델의 문장 이해력을 향상하기 위한 필수적인 단계이다. 입력 자질의 품질 제고는 세부 태스크의 성능과 직결된다. 본 논문은 단어와 문장 분류 관점에서 한국어의 언어적 특징을 효과적으로 반영하는 분절 전략을 비교 연구한다. 분절 유형은 언어학적 단위에 따라 어절, 형태소, 음절, 자모 네 가지로 분류하며, RoBERTa 모델 구조를 활용하여 사전학습을 진행한다. 각 세부 태스크를 분류 단위에 따라 문장 분류 그룹과 단어 분류 그룹으로 구분 지어 실험함으로써, 그룹 내 경향성 및 그룹 간 차이에 대한 분석을 진행한다. 실험 결과에 따르면, 문장 분류에서는 자모 단위의 언어학적 분절 전략을 적용한 모델이 타 분절 전략 대비 최대 NSMC: +0.62%, KorNLI: +2.38%, KorSTS: +2.41% 높은 성능을, 단어 분류에서는 음절 단위의 분절 전략이 최대 NER: +0.7%, SRL: +0.61% 높은 성능을 보임으로써, 각 분류 그룹에서의 효과성을 보여준다.

The construction of high-quality input features through effective segmentation is essential for increasing the sentence comprehension of a language model. Improving the quality of them directly affects the performance of the downstream task. This paper comparatively studies the segmentation that effectively reflects the linguistic characteristics of Korean regarding word and sentence classification. The segmentation types are defined in four categories: eojeol, morpheme, syllable and subchar, and pre-training is carried out using the RoBERTa model structure. By dividing tasks into a sentence group and a word group, we analyze the tendency within a group and the difference between the groups. By the model with subchar-level segmentation showing higher performance than other strategies by maximal NSMC: +0.62%, KorNLI: +2.38%, KorSTS: +2.41% in sentence classification, and the model with syllable-level showing higher performance at maximum NER: +0.7%, SRL: +0.61% in word classification, the experimental results confirm the effectiveness of those schemes.

Deep learning techniques such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) show superior performance in text classification than traditional approaches such as Support Vector Machines (SVMs) and Naïve Bayesian approaches. When using CNNs for text classification tasks, word embedding or character embedding is a step to transform words or characters to fixed size vectors before feeding them into convolutional layers. In this paper, we propose a parallel word-level and character-level embedding approach in CNNs for text classification. The proposed approach can capture word-level and character-level patterns concurrently in CNNs. To show the usefulness of proposed approach, we perform experiments with two English and three Korean text datasets. The experimental results show that character-level embedding works better in Korean and word-level embedding performs well in English. Also the experimental results reveal that the proposed approach provides better performance than traditional CNNs with word-level embedding or character-level embedding in both Korean and English documents. From more detail investigation, we find that the proposed approach tends to perform better when there is relatively small amount of data comparing to the traditional embedding approaches.

≪Xiandai Hanyu Cidian≫ is one of the dictionaries published in China. Although the dictionary is well written, we can still find some problems in it. It dsecribes that “Shiyi” is an adjective. A lot of dictionaries in China follow it. I argue that the dictionary’s explandtion of the term “Shiyi” is not entirely correct. In Modern Chinese adjective “Shiyi” can be followed by a fun-ction word “Yu”, and many grammarians argue that “Yu……” is the complement to the predecessor word “Shiyi”. But some scholars say that “Shiyi + Yu” is a structure which is equal to the role of a transitive verb. I think their argument is much better than that of the former. The reason is that the preposition “Yu” already adheres to the adjective “Shiyi” in Chinese. Another phenomenon worth paying attention to is that “Shiyiyu” sometimes can drop “Yu” in its structure, and result in “Shiyi” having its object directly. The transitive verb “Shiyi” and the structure of the “Shiyiyu” can take both verbal objects and noun structure objects. This means that “Shiyi” can also have the role of a transitive verb. In my opinion, the explanation of the term “shiyi” should be translated as follows: ｢【shiyi】Shi Yi Adjective heshi; xiangyi: xingdong～｜bianyuan shanqu, jiaotong∼｡ Transitive Verb shihe; dui……heshi/xiangyi; zai(zuomoufangmian)……heshi/xiangyi: shatu∼ zhong huasheng｡｣

자연어 처리를 이용한 문서 분류 분야에서도 전통적인 방법에서 벗어나 단어 임베딩을 활용한 합성곱 신경망과 순환 신경망 등 심층 신경망을 이용한 다양한 연구가 진행되고 있다. 본 논문에서는 Word2Vec과 두 개의 계층으로 구성 된 양방향 장단기 기억 네트워크를 이용한 특허 문서의 IPC(International Patents Classification) 자동 분류 모델을 제안한다. IPC는 세계지식재산권기구에서 제정한 국제적으로 통일된 특허 분류 기준이며, 각 국가의 공인된 기관에서 수작업으로 분류하고 있다. IPC 자동 분류를 위하여 입력 시퀀스에 Word2Vec을 이용한 단어 임베딩가 중치를 사용한다. 그리고 가중치가 부여된 시퀀스를 두 개의 계층을 갖는 깊은 구조의 양방향 장단기 기억 네트워크 신경망에 입력하여 IPC를 분류한다. 실험 결과 특허 문서의 분류 정확도가 합성곱 신경망 보다는 약 7% 향상되었 으며, 순환 신경망을 단일로 이용하는 것 보다는 약 5% 향상된 것을 확인할 수 있었다. 또한 전통적인 방법인 나이 브 베이시안, 로지스틱 분류 및 서포트 벡터 머신보다는 5~12% 이상 우수한 성능을 나타내었다.

There are various studies using Deep Neural Network such as CNN(Convolutional Neural Network) and RNN(Recurrent Neural Network) that utilize word embedding in document classification using natural language processing out of traditional methods. In this paper, we propose the IPC(International Patents Classification) automatic classification model of patent documents using two layers BLSTM (Bidirectional Long Short Term memory) network. The IPC is an internationally uniform standard for patent classification established by the World Intellectual Property Organization and is categorized by hand in authorized agencies in each country. For the IPC automatic classification, we use word embedding weight with Word2Vec in the input sequences. And they are classified by entering a weighted sequences into a deep neural network with two layers BLSTM. The experimental results showed that the accuracy of classification is improved by about 7% than that of CNN, and about 5% than that of single layer LSTM that is a field of RNN. Also it showed more than 5~12% higher performance than traditional methods such as Naive Bayes, Logistic and Support Vector Machine classification.

One of the most important processesin a machine learning-based natural language processing module is to represent words by inputting the module. This can be accomplished by representing words in one-hot form with a large vector size without applying the concept of semantic similarity between words, or by word representation (word embedding) with vectors to represent lexical similarity. This has attracted keen research interest by improving the performance of several natural language processing modelssuch as syntactic parsing and sentiment analysis (also known as opinion mining). In this study, classification performance of Word2Vec, canonical correlation analysis (CCA), and GloVeare tested on a corpus that established using the titles and abstractsof 204,674biomedical articles published in PubMed. Categories include disease name, disease symptom, and ovarian cancer marker.Ovarian cancer markers were used as bio-markers.The classification performance of each word representation model for each category is visualized by mapping the results in two-dimensional word representations using t-distributed stochastic neighbor embedding (t-SNE).

This paper mainly conducted two series of experiments. One is investigation between language dependent and language independent features. Bi-grams in Korean experiments and uni-grams in Chinese contributed most as basic features. And another one is utilization of Korean WordNet to improve the performance of Korean document classification. Korean WordNet is a Korean Lexical Semantic Network. Language independent features seem can lead better performance and stable. The performance of Korean text classification was improved by using Korean WordNet.

It may not be desirable, even though effective, to study the compound verbs only which have already become glossaries or have strong potential to become words. The purpose of the current study is to reinvestigate various criteria which have been utilized to classify extant compound verbs, newly formed compound verbs, and their word structures, based on analyses of triplet or quartet compound verbs and their innate structures. The analyses showed that while there were differences in expression forms between extant compound verbs and newly created compound verbs in terms of the structures of process-result, there were no differences in the patterns found in the creation process. It was also revealed that triplet and/or quartet compound verbs were also characterized as the combination of both the structures of processes-results and those of non process-result, not just either of them. The current study intended to analyze various linguistic phenomena and approached from the perspective of synchronic linguistics, not to discuss whether triplet and/or quartet compound verbs could be settled as forms of phrases or as grammatical phenomena.

데이터 마이닝의 문서분류 기술에서 발전된 오피니언 마이닝은 이제 국외뿐만 아니라 국내 산업에서 중요한 관심분야로 자리잡아가고 있다. 오피니언 마이닝의 핵심은 문서에서 감정 단어를 추출하여 긍정/부정 여부를 얼마나 정확하게 판별하느냐를 평가하는 것이다. 국내에서도 이에 관련된 많은 연구가 이루어 졌으나 아직 실용적으로 적용 할 만큼의 분류 정확도를 보이지 않고 있다. 한국어의 경우 비문법적 표현, 감정단어의 다양성 등으로 인해 문서의 극성을 판별하기가 쉽지 않기 때문이다. 본 논문에서는 문법적 요소를 최대한 배제하고 단어패턴의 빈도만을 고려한 새로운 오피니언 문서 분류기법을 제안한다. 제안된 방법에서는 문서를 단어들의 리스트로 추상화한 후, 패턴들의 빈 도를 이용하여 기계학습 알고리즘을 적용한다. 이후에 적절한 스코어 함수를 적용하여 문서의 극성을 판별한다. 또한 제안된 기법의 정확도를 평가하기 위해서 실험결과를 제시한다.

An opinion mining technique which was developed from document classification in area of data mining now becomes a common interest in domestic as well as international industries. The core of opinion mining is to decide precisely whether an opinion document is a positive or negative one. Although many related approaches have been previously proposed, a classification accuracy was not satisfiable enough to applying them in practical applications. A opinion documents written in Korean are not easy to determine a polarity automatically because they often include various and ungrammatical words in expressing subjective opinions. Proposed in this paper is a new approach of classification of opinion documents, which considers only a frequency of word patterns and excludes the grammatical factors as much as possible. In proposed method, we express a document into a bag of words and then apply a learning algorithm using a frequency of word patterns, and finally decide the polarity of the document using a score function. Additionally, we also present the experiment results for evaluating the accuracy of the proposed method.

In document classification, target of analysis is not document itself but words appeared in the document. Word information, therefore, is a significant factor in document classification. In this study, we are dealing with the classification of Korean document based on words and feature vectors. First, we present the performance of document classification using nouns and keywords. Second, we compare to the results for the size of feature vectors.

Most of the fixed word combinations were recognized as a kind of idiom in phraseology. They consist of several words, but their meaning cannot be inferred from, or is not the sum of the meanings of its component words. There are many types of fixed word combinations, for example, collocations, idioms, conventional expressions, routine formulas, proverbs etc. There are quite many various definitions of fixedness and idiomaticity and it is difficult to list them consistently. Many fixed word combinations are more or less fixed in form, but there are also a great number of idioms that are variable. Furthermore, there are different degrees of idiomaticity. In this paper, we focused on the essential features, stability or fixedness and idiomaticity. A fixedness is considered based of different flexibility of the syntactic distribution in corpus. An idiomaticity is reflected linguistic meaning type and is considered the correlation between whole meaning and extralinguistic reality and whether to present information in dictionary, etc. According to this feature, we mainly analyzed word-combinations which is consisted of two linguistic signs. Through this study, fixed combinations are divided into three types and each type have two sub-types, so it was classified into a total of six types. This study is expected to be able to understand and explain systematically the principle and meaning of the fixed word combinations, which are difficult to treat lexicographically and pedagogically.

Sentiment analysis employs classification models to discern people's opinions automatically. Recent strides made with Large Language Models (LLMs) have significantly enhanced the accuracy of binary-level sentiment classification, particularly through zero-shot and few-shot learning approaches. However, when it comes to fine-level sentiment classification, LLMs face challenges because they are not specifically trained for this downstream task. In contrast, other classification models utilize word embedding, a vector representation of words, as input data. Contemporary word embedding algorithms create these embeddings by considering the surrounding context of each word. Nonetheless, these word embeddings often fail to capture the nuances of intensity differences between words. For example, words like 'more' and 'less' have embeddings closely positioned in the semantic space, despite representing distinct intensity levels. These intensity words such as 'much', 'more', and 'less' are frequently used to convey the strength of opinions. Their intensity distinctions are crucial in fine-level sentiment classification. This paper introduces an innovative intensity-aware feed-forward neural network, equipped with a novel referential loss function designed to capture these intensity differences between words. The proposed model effectively separates words of varying intensities while bringing together words sharing the same intensity in the semantic space. To assess the effectiveness of this refined word embedding in sentiment analysis tasks, diverse fine-level sentiment datasets are employed. The results demonstrate that the refined word embedding surpasses original embeddings and popular LLMs for fine-level sentiment analysis.

본 논문에서는 중 고등 수준 단문형 영어 작문시험의 자동채점 시스템을 위한 사전 미등록어 분류기 구현에 대해 다룬다. 영어 자동채점 과정에서 발생하는 사전 미등록어의 유형을 정의하고 각 유형에 대한 검출 방법에 대해 논의하였다. 또한 영작문 답안에서 나타날 수 있는 비속어의 유형을 정의하고 검출 방법에 대해 연구하였다. 영작문 자동평가 시스템의 모듈로서 비속어 검출 기능이 포함된 미등록어 분류기를 구현하였다. 미등록어 분류와 비속어 검출 방법에 대한 성능을 실제 시험 데이터에 적용하여 그 성능을 평가하였다.

In this paper, we deal with implementation issues of an unknown word classifier for middle-school level English writing test. We define the type of unknown words occurred in English text and discuss the detection process for unknown words. Also, we define the type of swear words occurred in students's English writings, and suggest how to handle this type of words. We implement an unknown word classifier with a swear detection module for developing an automatic English writing scoring system. By experiments with actual test data, we evaluate the accuracy of the unknown word classifier as well as the swear detection module.

《現代漢語詞典》에서 “便利”를 형용사와 동사의 겸류사로서 이 중 동사는 사동용법으로 “使……便利”의 의미를 가지는 것으로 본다.《現代漢語詞典》이 중국에서 영향력이 있는 까닭에 다른 사전들도 모두 이를 따르고 있다. “便利”의 형용사에 대해서 연구자는 이견이 없지만 동사 “便利”가 단지 사동 용법만 갖고 있다는 점에 대해서는 동의하지 않는다. 연구에 따르면 타동사 “便利”는 그 목적어로 명사성 결구， 동사성 결구， 主謂小句를 갖는데， 목적어가 명사성 결구일 시는 단지 사동용법으로 “使便利”의 의미만 가지지만， 목적어가 동사성 결구일 경우는 岐義句로서 “便利”가 사동용법으로 “使……便利”의 뜻을 갖을 수도 있고 非사동용법으로 쓰일 때는 “便於”， “比較容易(做某事)”의 뜻을 갖고 있으며 이는 언어환경에 따라서 판단해야 한다고 본다. 이런 용법의 출현은 현대 중국어에서 “형용사/동사+於”가 하나의 어법단위로서 술어를 구성함과 관계가 있다. 이런 격식은 심지어 타동사 뒤에도 “於”가 출현하는 결과를 갖고 왔으면 “便利於”가 그 경우이다.

In China， the dictionary 《Xiandai Hanyu Cidian》is one of the most popular dictionaries in China. It has a significant influence on other dictionaries published in China. But it often fails to correctly describe some word classifications and the meanings. One example is “Bianli”. The dictionary describes the word “Bianli” as both adjective and verbs. It does not categorize between intransitive verbs and transitive verbs. Actually， the word “Bianli” has a double classification of both adjective and transitive verbs. I agree with its definition of the meaning of the adjective “Bianli”， but I have a different opinion regarding the meaning of the transitive veb “Bianli”. The dictionary defines “Bianli” as “Shi Bianli” when it is uesd as a transitive verb. I think it also has the meaning of “Bijiao Rongyi(Zuo Moushi)”. Except for 《Xiandai Hanyu Cidian》， many other dictionaries exclude definition of “Bijiao Rongyi(Zuo Moushi)”. I think they must modify their definition of “Bianli” and include a definition as a transitive verb

'不宜' 에 대한 《현대한어사전》의 관점은 동사이고 "不適宜" 의미를 가진다고 본다. 그러나 저자는 이에 동조하지 않는다. 저자의 연구에 다르면 "不宜"는 타동사, 조동사, 형용사의 사성을 갖는다고 본다. 저자의 관점에 따르면 "不宜"는 아래와 같이 정리된다. 「【不宜】bùyí ①形容詞 不合適;不適合;不適宜:少兒∼┃對他的腸胃如此∼┃努力剔除他們認爲∼的表達方式 ②及物動詞 不適合;不適宜;不適合於;不適宜於:這塊地∼種植水稻 ③助動詞 不應當:不可誇大也∼縮小┃解決思想問題要耐心細致,∼操之過急」

한국어 정보처리 시스템은 유형별로 다양한 형태의 형태소 분석 정보를 필요로 하는데 이를 위하여 한국어의 단어 유형을 분류하고 형태소 분석 결과를 효율적으로 저장하는 자료구조를 제안한다. 형태소 분석에 필요한 단어 유형은 일반적인 유형과 단순화된 유형으로 구분하여 비교하였으며, 이를 기반으로 형태소 분석을 위한 새로운 단어 구성 전이도를 제시하였다. 형태소 분석 결과를 저장하는 자료구조는 HAM에서 사용되고 있는 자료구조를 기반으로 응용시스템에서 필요로 하는 정보를 쉽게 사용할 수 있도록 보완하고 저장 공간의 효율성을 개선하였다.

《現代漢語詞典》,《學生現代漢語詞典》,《現代漢語常用詞用法詞典》,《實用漢語用法詞典》,《現代漢語八百詞》,《漢語敎與學詞典》등 사전에서 “是否”를 부사로 규정한다. “是否”가 사전상에 부사로 명시되었다고 하여, “是否”가 결코 부사라고만 할 수는 없다. 필자는 현대중국어에서 “是否” 뒤에 체언성 성분이 오는 경우를 발견했다. 이 경우 “是否”를 술어로 보고 그 뒤에 출현하는 체언성 성분을 목적어로 본다. 즉, 이러한 경우 “是否”를 타동사로 본다. “是否” 뒤에 체언성 성분이 올 때, 판단 동사 “是”가 체언성 성분을 목적어로 갖는 것과 관계가 깊으며, 이때는 그 기능과 의미가 “是不是”와 동일하다. 필자는 이 경우 “是否”를 술어로 보고 그 뒤에 오는 성분을 목적어로 본다. “是否”는 체언성 성분 외에 술어성 성분도 목적어로 가질 수 있으므로 필자는 이를 體賓謂 (체빈/위빈)동사로도 본다. 體賓謂 (체빈/위빈)동사 “是否”는 주어와 목적어 간의 관계를 나타내므로 관계 동사로 볼 수도 있다. 또한 “是否”가 동사로 활용될 때 반드시 목적어를 대동하니 이를 粘賓관계동사로도 볼 수 있다. 그러나 동사 “是否”와 “是不是”의 용법이 완전히 동일한 것은 아니므로 이에 대해서는 반드시 세밀하게 연구할 필요가 있다.

Lushuxiang argues that when “Shifou” has the meaning of “shibushi”, it is an adverb. Many grammarians agree with this assessment. The reason why they define “Shifou” as an adverb is deeply related with their opinion of the term “Shifou”. They argue that it＇s role is only being adverbial in Modern Chinese. However, l have found that the term “Shifou” in many cases can be changed into the structure of “Shibushi”. Furthermore, unlike adverbs, “Shifou” can not be removed from many sentences. lt also questions the relationship between the subjects and objects. Therefore, l argue that “Shifou” is also a verb. “Shibushi” can take both predicate objects and nominal objects. On the other hand, when “Shifou” functions as a transitive verb, it also takes predicate objects and nominal objects. But the nominal objects they take are not exactly the same. Through my studies, l have found that “Shifou”’objects are usually restricted in abstract noun structures. Why is that? That is the problem l intend to solve shortly.

Xiandai Hanyu Cidian is one of the dictionaries that is popular among Chinese scholars. Although the dictionary is well written, we can still find some problems in it, and one of them is about “Xiangdeng”. It describes “Xiangdeng” as a verb. At present, lots of dictionaries published in China also follow Xiandai Hanyu Cidian, such as Xuesheng Xiandai Hanyu Cidian, Dangdai Hanyu Cidian, Xiandai Hanyu Guifan Cidian, Xiandai Hanyu Tujie Cidian, HSK Ciyu Yongfa Xiangjie, Hanyu Jiao Yu Xue Cidian, and Xiandai Hanyu Fenlei Cidian. But I do not agree with them. According to my research, “Xiangdeng” is not related with actions or behaviors, so it can not be an intransitive verb. “Xiangdeng” can not be a transitive verb, because it can not take objects. If we analyse the term “Xiangdeng”, “Xiang” has the character of adverbs, and “Deng” of adjectives, then the combine of “Xiang” and “Deng” can be analysed as an adjective. Actually, there are many examples similar to this in Modern Chinese. Generally speaking, adjectives can be modified by degree adverbs like “Hen”, “Zui” and so forth. But not all the adjectives can be modified by these sorts of adverbs, for example “Xiangtong”, “Yiyang”. The term “Xiangdeng” is one of them. And when the “Xiangdeng” is followed by “Yu”, the function of “Xiangdengyu” is similar to a transitive verb. Therefore the structure of “Xiangdengyu” can take objects. In my opinion, the explanation of the term “Xiangdeng” should be translated as follows: 「【Xiangdeng】 xi?ngd?ng Adjective (shumu, fsnliang, chengdu and so force) bici yiyang: mianji??Zhe liang jian fangzi de mianji?。」

“傷心”의 사성에 대해 ≪現代漢語詞典≫ 등 사전에서 이를 형용사로 보지만 ≪應用漢語詞典≫ 등에서는 이를 동사로 본다. 形福義와 湯廷池 등은 동사를 먼저 자동사와 타동사로 나누고 있다. 자동사와 타동사의 가장 큰 차이는 목적어를 갖을 수 있느냐 없느냐에 있다. 반면 형용사와 자동사의 차이는 동작, 행위성을 갖느냐 않느냐에 있다. 이런 인식하에 “傷心”을 검토하면 형용사일 수가 없는데, 이는 “傷心” 이 심리적인 동작, 행위성을 갖고 있기 때문이다. “傷心” 은 이합사 중 술빈식 이합사로 일반적으로 그 뒤에 다시 목적어를 갖지 않는다. 이는 “傷心” 의 품사가 자동사임을 의미한다. 술빈식 이합사에서 일부는 쓰임새가 높고 결합도가 강해서, 목적어를 갖을 수가 있다. 그 중의 하나가 “傷心”인데, 그 뒤에 원인 목적어를 갖는다. 결론적으로 말하면, “傷心” 은 자동사와 타동사의 경류사이다.

Many dictionaries which are published in China like Xiandai Hanyu Cidian, Xiandai Hanyu Guifan Cidian, Xiandai Hanyu Fenlei Cidian, Dangdai Hanyu Cidian, Xuesheng Xiandai Hanyu Cidian and others. define “shangxin” is an adjective, many other dictionaries like HSK Ciyu Yongfa Xianjjie, Hanyu Jiao yu Xue Cidian, Zhongchao Cidian, Shiyong Hanyu Yongfa Cidian and others. argue that “shangxin” is a verb. When “shangxin” has the same function and meaning in the sentences, Ii can not be both of an adjective and a verb. therefore l argue that the word classification of shangxin” is in need of study. “Shangxin” in modern chinese belongs to “liheci”. The so called “liheci” is king of verbs, not adjectives. According to my study, verbs can be divideded into intransitive verb and transitive verb, and “liheci” is basically an intransitive verb. So l argue that “shangxin” is not an adjective , but an intransitive verb. In modern chinese, some “liheci” can also have objects, for examples “danxin”, and “shangxin”. So l can say that “shangxin” has double classification of intransitive and transitive verb. When “shangxin” is an intransitive verb, its meaning is “neixin beishang, tongku”, and when it is a transitive verb , it has the meaning of “yin.......neixin beishang, tongku”.

關於“有害”詞性，我們得出如下結論: 1)《中朝》、《進明》、《敎與學》收錄了“有害”一詞，但它們對“有害”的詞性有不同的觀點。《中朝》認爲“有害”是形容詞，而《進明》、《敎與學》認爲是動詞。 2) “有+抽象名詞”結構在現代漢語中具有形容詞性意義。其中單音節“有”與單音節“害”凝固成一個述賓式形容詞, 其意義是“有害處”。 3) “有害”能受程度副詞的修飾，而不能帶賓語。但“有害”後面還能出現“於”， “有害於”是“有害(形容詞)+於(虛詞)”結構充當述語而帶賓語的。其功能相當於一個黏賓及物動詞。 4) 在“有害於”中, 特別當賓語很短時, 虛詞“於”往往容易丟失。久而久之，“有害”獲得了及物動詞的詞性，而直接帶賓語。這種及物動詞“有害”裏面實際上隱含着“於”, 在意思和功能上相當於及物動詞性結構“有害於”,其釋義爲“對……有害處”, 屬於靜態性及物動詞。

Dictionaries like 《Xiandai Hanyu Cidian》, 《Xuesheng Xiandai Hanyu Cidian》 do not include the term "Youhai", on the contrary other dictionaries like 《Jinming Zhonghan Dacidian》, 《Zhongchao Cidian》 take on the word "Youhai". 《Jinming Zhonghan Dacidian》 argues that "Youhai" is a verb, but 《Zhongchao Cidian》 describes that "Youhai" is an adjective. So I think that it is worthing to study the word classification of the "Youhai". "Youhai" is consisted of the morphemes of "You" and "Hai". So I need to consult with the similar constructure "Youyi". According to a lot of dictionaries published in China, "Youyi" is an adjective. And in Modern Chinese the structure of the "You+abstract noun" has the meaning of an adjective, though it is the structure of "Verb+Noun", so "Youyi", "Youli", "Youxiao" and so on are adjectives. So I argue that "Youhai" is also an adjective. The adjective "Youhai" can be followed by the function word "Yu", and the structure of "Youhaiyu" roles as a predicate and takes an object. But now the morpheme "Yu" of the "Youhaiyu" can be easily dropped, and the result, "Youhai" can take objects directly. So I argue that in this case, "Youhai" has the word classification of a transitive verb.