Skip to main content

Yu-ming Hsieh

Followers

8

Following

1

Co-author

1

Public Views

Interests

Uploads

Papers by Yu-ming Hsieh

Ambiguity Resolution for Vt-N Structures in Chinese

The syntactic ambiguity of a transitive verb (Vt) followed by a noun (N) has long been a problem ... more The syntactic ambiguity of a transitive verb (Vt) followed by a noun (N) has long been a problem in Chinese parsing. In this paper, we propose a classifier to resolve the ambiguity of Vt-N structures. The design of the classifier is based on three important guidelines, namely, adopting linguistically motivated features, using all available resources, and easy integration into a parsing model. The linguistically motivated features include semantic relations, context, and morphological structures; and the available resources are treebank, thesaurus, affix database, and large corpora. We also propose two learning approaches that resolve the problem of data sparseness by autoparsing and extracting relative knowledge from large-scale unlabeled data. Our experiment results show that the Vt-N classifier outperforms the current PCFG parser. Furthermore, it can be easily and effectively integrated into the PCFG parser and general statistical parsing models. Evaluation of the learning approaches indicates that world knowledge facilitates Vt-N disambiguation through data selection and error correction.

語法規律的抽取及普遍化與精確化的研究 (Grammar Extraction, Generalization and Specialization) [In Chinese]

International Conference on Computational Linguistics, Sep 1, 2004

Ta jiao Li-si jian qiu. "He asked Lisi to pick up the ball."

Resolving the Representational Problems of Polarity and Interaction between Process and State Verbs

Event classification is one of the crucial tasks in lexical semantic representation. Traditionall... more Event classification is one of the crucial tasks in lexical semantic representation. Traditionally, researchers have regarded process and state as two top-level events and discriminated between them by semantic and syntactic characteristics. In this paper, we add cause-result relativity as an auxiliary criterion to discriminate between process and state by structuring about 40,000 Chinese verbs to the two correspondent event hierarchies in E-HowNet. All verbs are classified according to their semantic similarity with the corresponding conceptual types of ontology. As a result, we discover deficiencies of the dichotomy approach and point out that any discrete event classification system is insufficient to make a clear-cut classification for synonyms with slightly different semantic focuses. We then propose a solution to remedy the deficiencies of the dichotomy approach. For the process or state type mismatched verbs, their inherited semantic properties will be adjusted according to their PoS and semantic expressions to preserve their true semantic and syntactic information. Furthermore, cause-result relations will be linked between corresponding processes and states to bridge the gaps of the dichotomy approach.

基於相依詞向量的剖析結果重估與排序(N-best Parse Rescoring Based on Dependency-Based Word Embeddings)

Rescoring approaches for parsing aim to re-rank and change the order of parse trees produced by a... more Rescoring approaches for parsing aim to re-rank and change the order of parse trees produced by a general parser for a given sentence. The re-ranking quality depends on the precision of the rescoring function. However it is a challenge to design an appropriate function to determine the qualities of parse trees. No matter which method is used, Treebank is a widely used resource in parsing task. Most approaches utilize complex features to re-estimate the tree structures of a given sentence [1, 2, 3]. Unfortunately, sizes of treebanks are generally small and insufficient, which results in a common problem of data sparseness. Learning knowledge from analyzing large-scaled unlabeled data is compulsory and proved useful in the previous works [4, 5, 6]. How to extract useful information from unannotated large scale corpus has been a research issue. Word embeddings have become increasingly popular lately, proving to be valuable as a source of features in a broad range of NLP tasks [7, 8, 9]...

N-best Rescoring for Parsing Based on Dependency-Based Word Embeddings

Int. J. Comput. Linguistics Chin. Lang. Process., 2016

Rescoring approaches for parsing aims to re-rank and change the order of parse trees produced by ... more Rescoring approaches for parsing aims to re-rank and change the order of parse trees produced by a general parser for a given sentence. The re-ranking performance depends on whether or not the rescoring function is able to precisely estimate the quality of parse trees by using more complex features from the whole parse tree. However it is a challenge to design an appropriate rescoring function since complex features usually face the severe problem of data sparseness. And it is also difficult to obtain sufficient information requisite in re-estimatation of tree structures because existing annotated Treebanks are generally small-sized. To address the issue, in this paper, we utilize a large amount of auto-parsed trees to learn the syntactic and sememtic information. And we propose a simple but effective score function in order to integrate the scores provided by the baseline parser and dependency association scores based on dependency-based word embeddings, learned from auto-parsed tr...

中文語料庫構建及管理系統設計 (Design of Management System for Chinese Corpus Construction) [In Chinese]

Ä^ë)jtÖÛ 8wÄëjtÖÛ Tjojdb Dpsqvt ëyÑ zx°õ!¨í>ÖjtÖÛí$ 2::5 úÝÉeFfÚ]ÙÃë)c ð'±Þco*ÄªÚ¯ëÞ°¨ÌªòH'1o5||í... more Ä^ë)jtÖÛ 8wÄëjtÖÛ Tjojdb Dpsqvt ëyÑ zx°õ!¨í>ÖjtÖÛí$ 2::5 úÝÉeFfÚ]ÙÃë)c ð'±Þco*ÄªÚ¯ëÞ°¨ÌªòH'1o5||í2::8 úÉ" ë)ÖÛ 4/1 w¹íiÍ$Û7ÌªíÑq±ÿÞ 3114 ú¿si ¦ÛÌªí íä[YîÞ|RÇ]ëÌÞ•|o*Ä ¯¬½îËëÞ¨íûÃ'dí ÙëÑqØ¸5NC#x&]jtÖÛëoá58Ê7±;

Semantic relation identification for consecutive predicative constituents in Chinese

Lingua Sinica, 2017

In this paper, we propose a general methodology for designing semantic role/ relation system. Bas... more In this paper, we propose a general methodology for designing semantic role/ relation system. Based on this methodology, we establish a succinct semantic relation system for consecutive predicative constituents for Chinese, which includes serial verb construction, discourse construction, and other constructions describing serial events. This semantic relation system has 13 middle-level classes and 24 fine-grained sub-classes in contrast to conventional complex classification schemes and meets the uniqueness and completeness criteria of semantic relation identification. We conduct experiments on our system by training four annotators in 1 h to label 200 sentences extracted from Sinica Treebank and HIT-CDTB. With the help of our predesigned feature-based decision tree and a connective markers checklist, the annotators attain a 73% consistency with the reference standard annotation and substantial agreement by Cohen's kappa coefficient for middle-level labeling. By analyzing the labeling error types, we slightly revise our classification scheme and propose six methods to improve the classification and labeling system, hoping to achieve even better agreement in the future.

Lexical Representation and Classification of Eventive Verbs - Polarity and Interaction between Process and State

Event classification is one of the most crucial tasks in lexical semantic representation. Traditi... more Event classification is one of the most crucial tasks in lexical semantic representation. Traditionally, researchers regarded process and state as two top level events and discriminated them by semantic and syntactic characteristics. In this paper, we add cause-result relativity as an auxiliary criterion to discriminate between process and state by structuring about 40,000 Chinese verbs to the two correspondent event hierarchies in E-HowNet. All verbs are classified according to their semantic similarity with the corresponding conceptual types of the ontology. As a consequence, we discover deficiencies of the dichotomy approach and point out that any discrete event classification system is insufficient to make a clear cut classification for synonyms with slightly different semantic focuses. We then propose a solution to remedy the deficiencies of the dichotomy approach. For the process or state type mismatched verbs, their inherited semantic properties will be adjusted according to ...

Improve Parsing Performance by Self-Learning

International Journal of Computational Linguistics & Chinese Language Processing, Volume 12, Number 2, June 2007, Jun 1, 2007

There are many methods to improve performances of statistical parsers. Among them, resolving stru... more There are many methods to improve performances of statistical parsers. Among them, resolving structural ambiguities is a major task. In our approach, the parser produces a set of n-best trees based on a feature-extended PCFG grammar and then selects the best tree structure based on association strengths of dependency word-pairs. However, there is no sufficiently large Treebank producing reliable statistical distributions of all word-pairs. This paper aims to provide a self-learning method to resolve the problems. The word association strengths were automatically extracted and learned by parsing a giga-word corpus. Although the automatically learned word associations were not perfect, the built structure evaluation model improved the bracketed f-score from 83.09% to 86.59%. We believe that the above iterative learning processes can improve parsing performances automatically by learning word-dependence knowledge continuously from web.

Introduction to CKIP Chinese Spelling Check System for SIGHAN Bakeoff 2013 Evaluation

In order to accomplish the tasks of identifying incorrect characters and error correction, we dev... more In order to accomplish the tasks of identifying incorrect characters and error correction, we developed two error detection systems with different dictionaries. First system, called CKIP-WS, adopted the CKIP word segmentation system which based on CKIP dictionary as its core detection procedure; another system, called G1-WS, used Google 1T uni-gram data to extract pairs of potential error word and correction candidates as dictionary. Both detection systems use the confusion character set provided by the bakeoff organizer to reduce the suggested correction candidates. A simple maximizing tri-gram frequency model based on Google 1T tri-gram was designed to validate and select the correct answers. The CKIP group of Academia Sinica participated in both Sub-Task1 (Error Detection) and Sub-Task2 (Error Correction) in 2013 SIGHAN bakeoff. The evaluation results show that the performances of our systems are pretty good on both tasks.

Resolving Ambiguities of Chinese Conjunctive Structures by Divide-and-conquer Approaches

This paper presents a method to enhance a Chinese parser in parsing conjunctive structures. Long ... more This paper presents a method to enhance a Chinese parser in parsing conjunctive structures. Long conjunctive structures cause long-distance dependencies and tremendous syntactic ambiguities. Pure syntactic approaches hardly can determine boundaries of conjunctive phrases properly. In this paper, we propose a divide-andconquer approach which overcomes the difficulty of data-sparseness of the training data and uses both syntactic symmetry and semantic reasonableness to evaluate ambiguous conjunctive structures. In comparing with the performances of the PCFG parser without using the divide-andconquer approach, the precision of the conjunctive boundary detection is improved from 53.47% to 83.17%, and the bracketing f-score of sentences with conjunctive structures is raised up about 11 %.

Correcting Chinese Spelling Errors with Word Lattice Decoding

ACM Transactions on Asian and Low-Resource Language Information Processing, 2015

Chinese spell checkers are more difficult to develop because of two language features: 1) there a... more Chinese spell checkers are more difficult to develop because of two language features: 1) there are no word boundaries, and a character may function as a word or a word morpheme; and 2) the Chinese character set contains more than ten thousand characters. The former makes it difficult for a spell checker to detect spelling errors, and the latter makes it difficult for a spell checker to construct error models. We develop a word lattice decoding model for a Chinese spell checker that addresses these difficulties. The model performs word segmentation and error correction simultaneously, thereby solving the word boundary problem. The model corrects nonword errors as well as real-word errors. In order to better estimate the error distribution of large character sets for error models, we also propose a methodology to extract spelling error samples automatically from the Google web 1T corpus. Due to the large quantity of data in the Google web 1T corpus, many spelling error samples can be...

SETM*-MaxK: An Efficient SET-Based Approach to Find the Largest Itemset

Lecture Notes in Computer Science, 2002

In this paper, we propose the SETM*-MaxK algorithm to find the largest itemset based on a high-le... more In this paper, we propose the SETM*-MaxK algorithm to find the largest itemset based on a high-level set-based approach, where a large itemset is a set of items appearing in a sufficient number of transactions. The advantage of the set-based approach, like the SETM algorithm, is simple and stable over the range of parameter values. In the SETM*-MaxK algorithm, we efficiently find the L k based on Lw, where L k denotes the set of large k-itemsets with minimum support, L k = ∅, L k+1 = ∅ and w = 2 log 2 k −1 , instead of step by step. From our simulation, we show that the proposed SETM*-MaxK algorithm requires shorter time to achieve its goal than the SETM algorithm.

Translating Chinese Unknown Words by Automatically Acquired Templates

In this paper, we present a translation template model to translate Chinese unknown words. The mo... more In this paper, we present a translation template model to translate Chinese unknown words. The model exploits translation templates, which are extracted automatically from a word-aligned parallel corpus, to translate unknown words. The translation templates are designed in accordance with the structure of unknown words. When an unknown word is detected during translation, the model applies translation templates to the word to get a set of matched templates, and then translates the word into a set of suggested translations. Our experiment results demonstrate that the translations suggested by the unknown word translation template model significantly improve the performance of the Moses machine translation system.

Improving Chinese Parsing with Special-Case Probability Re-estimation

2013 International Conference on Asian Language Processing, 2013

Syntactic patterns which are hard to be expressed by binary dependent relations need special trea... more Syntactic patterns which are hard to be expressed by binary dependent relations need special treatments, since structure evaluations of such constructions are different from general parsing framework. Moreover, these different syntactic patterns (special cases) should be handled with distinct estimated model other than the general one. In this paper, we present a special-case probability re-estimation model (SCM), integrating the general model with an adoptable estimated model in special cases. The SCM model can estimate evaluation scores in specific syntactic constructions more accurately, and is able for adopting different features in different cases. Experiment results show that our proposed model has better performance than the state-of-the-art parser in Chinese.

DOMCAT: A Bilingual Concordancer for Domain-Specific Computer Assisted Translation

In this paper, we propose a web-based bilingual concordancer, DOMCAT 1, for domain-specific compu... more In this paper, we propose a web-based bilingual concordancer, DOMCAT 1, for domain-specific computer assisted translation. Given a multi-word expression as a query, the system involves retrieving sentence pairs from a bilingual corpus, identifying translation equivalents of the query in the sentence pairs (translation spotting) and ranking the retrieved sentence pairs according to the relevance between the query and the translation equivalents. To provide high-precision translation spotting for domain-specific translation ...

Improving PCFG Chinese Parsing with Context-Dependent Probability Re-estimation

Selecting the best structure from several ambiguous structures produced by a syntactic parser is ... more Selecting the best structure from several ambiguous structures produced by a syntactic parser is a challenging issue. The quality of the solution depends on the precision of the structure evaluation methods. In this paper, we propose a general model (context-dependent probability re-estimation model, CDM) to enhance the structure probabilities estimation. Compared with using rule probabilities only, the CDM has the advantage of an effective, flexible, and broader range of contexturefeature selection. We conduct ...

Chinese Treebanks and Grammar Extraction

Lecture Notes in Computer Science, 2005

Preparation of knowledge bank is a very difficult task. In this paper, we discuss the knowledge e... more Preparation of knowledge bank is a very difficult task. In this paper, we discuss the knowledge extraction from the manually examined Sinica Treebank. Categorical information, word-to-word relation, word collocations, new syntactic patterns and sentence structures are obtained. A searching system for Chinese sentence structure was developed in this study. By using pre-extracted data and SQL commands, the system replies the user's queries efficiently. We also analyze the extracted grammars to study the tradeoffs between the granularity of the grammar rules and their coverage as well as ambiguities. It provides the information of knowing how large a treebank is sufficient for the purpose of grammar extraction. Finally, we also analyze the tradeoffs between grammar coverage and ambiguity by parsing results from the grammar rules of different granularity.

Linguistically-Motivated Grammar Extraction, Generalization and Adaptation

Proceedings of the Second International Joint Conference on Natural Language Processing, 2005

In order to obtain a high precision and high coverage grammar, we proposed a model to measure gra... more In order to obtain a high precision and high coverage grammar, we proposed a model to measure grammar coverage and designed a PCFG parser to measure efficiency of the grammar. To generalize grammars, a grammar binarization method was proposed to increase the coverage of a probabilistic contextfree grammar. In the mean time linguistically-motivated feature constraints were added into grammar rules to maintain precision of the grammar. The generalized grammar increases grammar coverage from 93% to 99% and bracketing F-score from 87% to 91% in parsing Chinese sentences. To cope with error propagations due to word segmentation and part-of-speech tagging errors, we also proposed a grammar blending method to adapt to such errors. The blended grammar can reduce about 20~30% of parsing errors due to error assignment of pos made by a word segmentation system.

Improving PCFG Chinese Parsing with Context-Dependent Probability Re-estimation

by Keh-jiann Chen and Yu-ming Hsieh

Selecting the best structure from several ambiguous structures produced by a syntactic parser is ... more Selecting the best structure from several ambiguous structures produced by a syntactic parser is a challenging issue. The quality of the solution depends on the precision of the structure evaluation methods. In this paper, we propose a general model (context-dependent probability re-estimation model, CDM) to enhance the structure probabilities estimation. Compared with using rule probabilities only, the CDM has the advantage of an effective, flexible, and broader range of contexturefeature selection. We conduct ...