Hongying Zan

Computation of Word Similarity Based on the Information Content of Sememes and PageRank Algorithm

Based on sememe structure of HowNet and PageRank algorithm, this article proposes a method to compute word similarity. Using depth information of HowNet as information content of sememes and considering sememe hyponymy, this method builds a transfer matrix and computes sememe vector with PageRank algorithm to obtain sememe similarity. Thus, the word similarity can be calculated by the sememe similarity. This method is tested on several groups of typical Chinese words and word sense classification of nouns in Contemporary Chinese Semantic Dictionary (CSD). The results show that the word similarity computed in this way quite conforms with the facts. It also shows a more accurate result in word sense classification of nouns in the CSD, reaching 71.9% consistency with the judgment of human.

Inducing Chinese Selectional Preference Based on HowNet

Selectional preference (SP) is an important semantic knowledge. It can be used in various natural language processing tasks, including metaphor computing, lexicon building, syntactic structure disambiguation, word sense disambiguation, semantic role labeling, etc. However, handcrafted SP knowledge can not meet the requirement of large scale real text processing. Based on the noun taxonomy of How Net, this paper proposes a statistical and knowledge-based method to automatically induce Chinese SP. Preliminary experimental results show that the automatically acquired SP agree well with human judgment.

Research on Chinese Parsing Based on the Improved Compositional Vector Grammar

The basic task of syntactic parsing is to determine the syntactic structure of the sentence. Because the natural language is very complex, syntactic structure has a lot of ambiguities. Resolving ambiguity need to introduce a lot of information, and Compositional Vector Grammar (CVG) can well capture fine-grained syntactic and compositional-semantic information on phrases and words. In this paper, we first use a standard CVG model for Chinese parsing, and then we have made improvements on the CVG model. In order to introduce more information, for the word vector, we add the part of speech information; for the type of newborn node after binarization, add temporary node basic type; when computing node score, add the node type information. We also propose a solution for unknown word, replaced with structural vector. Our CVG parser improves the standard CVG parser by nearly 1% F1 on the development set of CTB8.0.

Proofreading and Revision of the Semantic Classes in the Contemporary Chinese Semantic Dictionary

The Contemporary Chinese Semantic Dictionary(CSD) is a semantic knowledge base which is developed on the basis of GKB: Grammatical Knowledge Base of Contemporary Chinese. It can support automatic semantic analysis in various natural language processing systems, such as machine translation systems. In order to improve the CSD and prepare it for a wider range of application, the NLP laboratory of Zhengzhou University investigates the current distribution of its semantic classes and revises the unreasonable aspects of its semantic classification according to the latest development of linguistic research, particularly that in lexical semantics. This paper introduces the rules and specifications for proofreading and revising the semantic classes in CSD, the tools developed for these tasks, the statistical analysis of the current distribution of semantic classes and Part-of-Speech in CSD, the problems found in this process, and our reflections on the rules of semantic classification and semantic taxonomy. Our research provides reference and tools for the construction of semantic dictionaries in the future.

Word Relevance Computation for Noun-Noun Metaphor Recognition

Metaphor in languages is an analogy-based meaning transfer phenomenon that impacts on question answering, machine translation and other tasks requiring deep semantic analysis. Noun-noun metaphor is a common type of metaphor and is studied in noun-noun semantic analysis. Lexical knowledge bases are important knowledge source for metaphor recognition and understanding. This paper proposes a method to compute word relevance for noun-noun metaphor recognition with the help of a lexical knowledge base, which shows good performance in the experiments. Furthermore, we investigate the representation of metaphor in lexical knowledge bases and its impact on the construction of lexical knowledge bases.

Study on the Annotation Framework of Chinese Logic Complement Semantics

The meaning expressed by elements of negation, degree, tense and aspect, modality and mood in a sentence attached to the basic predicate-centered proposition is called logic complement semantics, which is embodied as semantic constraints of logic semantic operators to the predicate. Logic complement semantics is the effective supplement to the basic logic meaning, and is important for deep understanding of sentence semantics. In this paper, a Chinese logic complement semantics annotation framework aimed for deep semantic comprehension is preliminarily practiced, which constructed a classification system including negation, degree, tense and aspect, and mood on the basis of existing research results, built the operator dictionary, established rules for annotation, and annotated logic complement semantics operators of a sentence which have been tagged with basic propositional arguments. Finally, the statistics of the annotation result are presented, and the problems in annotation process are analyzed.

Acquiring Selectional Preferences for Knowledge Base Construction

Selectional preference, or SP, is an important lexical knowledge that can be applied to many natural language processing tasks, like semantic error detection, metaphor detection, word sense disambiguation, syntactic parsing, semantic role labeling, and machine translation. This paper studies semantic class level SP acquisition for knowledge base construction. Firstly, the noun taxonomy of SKCC, a Semantic Knowledge-base of Contemporary Chinese, is adjusted for SP acquisition. Secondly, a MDL-based tree cut model is implemented. Thirdly, SP in SKCC is introduced as the source of gold standard test set to evaluate SP acquisition performance. Three kinds of predicate-argument relations are investigated in the experiments, including verb-object, verb-subject, and adjective-noun relations. For the verb-object relation, the top1 strict accuracy is 24.74% while the top3 relaxed accuracy reaches 75.26%.

Study on Modality Annotation Framework of Modern Chinese

Modality is the speaker’s subject idea processed and expressed for sentence objective express system. Modal meaning is important for deep understanding of sentence semantics. In this paper, a Chinese modality annotation framework aimed for deep semantic understanding is preliminarily practiced, which constructed a modal meaning classification system on the basis of existing research results, built the modal operator dictionary, established rules for annotation, and annotated modal operators of a sentence which have been tagged basic proposition arguments.

Analysis of Typical Annotation Problems in Bilingual Case Grammar Treebank Construction

In recent years, the study of machine translation has made great progress. However there are still many things to do for machine translation to reach the semantic level. In this paper, case grammar’s features that could well describe the semantic relationships in sentences were concluded. 23 thousand annotation errors that occurred in treebank construction were analyzed. 13 typical problems were summarized and the corresponding revolutions were proposed. The application of case grammar may contribute a new way of thinking for machine translation.

A Preliminary Contrastive Study on the Part-of-Speech Classifications of Two Lexicons

Having been debated and studied for more than one century, the part-of-speech classifications of contemporary Chinese words is still attracting considerable attention from many linguists today. In this study, we aim to compare the classification systems of two lexicons C Dictionary of Contemporary Chinese (Fifth Edition) and Grammatical Knowledge-Base of Contemporary Chinese, and to observe the similarities and differences between their part-of-speech classifications of words in a comprehensive way. This paper discusses our preliminary observations, especially on the comparison of prepositions in the two lexicons. We expect that this type of contrastive studies will contribute to a deeper understanding of the parts-of-speech in Chinese, especially to the part-of-speech classification of certain Chinese words, which has long been debated.