The Subjectivity of Some Probability Adverbs
Speaker: Prof Dingxu Shi, The Hong Kong Polytechnic University
Abstract: In Chinese, words like youshi (seldom), wangwang (usually) and zong (generally) are commonly considered frequency adverbs, on a par with words like ouer (rarely), changchang (often) and yizhi (always). It was discovered after analyzing large corpora data that the former are probability adverbs and only the later are frequency adverbs. Frequency and probability are the results of different calculations which require different items. A frequency adverb like the changchang (often) typically requires one set of things only, since the frequency for a given activity will remain constant no matter which time frame is adopted for the calculation. While in the case of probability adverbs, two sets are required for calculating probability. This is why using wangwang (usually) is not acceptable without a proper context. More importantly, youshi (seldom), wangwang (usually) and zong (generally) are subjective adverbs that express the speaker’s evaluation of the proposition. Such an evaluation is always against a baseline. The wangwang (usually) means the speaker thinks that the probability is close to the baseline, which should be the statistical average of 0.5. The youshi (probably) may be based on a statistic baseline but it could simply express the speaker’s subjective opinion. The sentence “Ni zenme zong bu jie wode dianhua?” (How dare you not answer my call in general?) is a girl’s whining against her boyfriend and her baseline for not answering her phone call is 0, a purely subjective decision.
Biography: Dingxu Shi received his MA in Applied Linguistics from University of Pittsburg and his MA and PhD in Linguistics from University of Southern California. He is now a Chair Professor in Chinese Linguistics at the Department of Chinese and Bilingual Studies of the Hong Kong Polytechnic University. His main research interest is on syntax, the interface of syntax and semantics, language contact and language change and Chinese pedagogical grammar. He has published a number of books in publishers such as Cambridge University Press and Peking University Press, and over one hundred papers in journals like Language, Natural Language and Linguistic Theory, Journal of Chinese Linguistics, Zhongguo Yuwen, Foreign Language Teaching and Research and Contemporary Linguistics.
Continuous Space Models and the Mathematics of Semantics
Speaker: Dr Rafael E. Banchs, Institute for Infocomm Research, Singapore
Abstract: In this talk, we present and discuss the most recent advances on the use of continuous space models in natural language processing applications. More specifically, some of the fundamental concepts of distributional semantics and vector space models are presented, followed by a brief discussion on linear and non-linear dimensionality reduction techniques and their implications to the parallel distributed approach to semantic cognition. Finally, some interesting examples about the use of continuous space models in both monolingual and cross-language natural language processing applications are presented.
Biography: Rafael E. Banchs is a Research Scientist at the Institute for Infocomm Research, in Singapore, where he leads the Dialogue Technology Lab of the Human Language Technology Department. He obtained a PhD degree in Electrical Engineering from The University of Texas at Austin in 1998. Since then he has worked on a variety of problems ranging from signal processing, electromagnetic modelling and parameter estimation to natural language processing, across different industries and applications. In 2004, he was awarded a five-year “Ramon Y Cajal” fellowship from the Spanish Ministry of Science and Technology to work on natural language processing related problem. He published over 30 journal papers. He held positions and taught in a number of universities and gave lectures and talks at many international conferences and workshops. His current area of research is focused on the construction and use of semantic representations to support different natural language processing applications, including machine translation, information retrieval, natural language understanding and chat-oriented dialogue.
Activation of the Weak Grammatical Feature [+L] and Its Activator
Speaker: Prof Jie Xu, Macau University
Abstract: This paper argues that most object/human-denoting nominals, whether N, NP or pronoun, have a weak grammatical feature [+L] (‘locative’), and that the feature needs to be activated properly in order to induce its syntactic effects. The feature is activated through different mechanisms in different languages. Along this line of approach, the widely observed construction of ‘nominal + place word’ in the modern Chinese language such as zhuozi pang ‘table side’ and Lao Wang zher ‘Lao Wang here’ is re-cast and re-captured as a consequence of the weak feature activation.
Biography: Jie Xu obtained his PhD in Linguistics from University of Maryland at College Park in 1993. He taught Chinese language and Linguistics at the National University of Singapore and Central China Normal University from 1993 to 2008 before joining the faculty of University of Macau. He is a professor of Chinese Linguistics in the Department of Chinese Language and Literature, Faculty of Arts and Humanities, University of Macau. His primary academic interests are in syntax, semantics, language acquisition, and language education. His major publications include Ten Essays on Chinese Descriptive Grammar, Grammatical Principles and Grammatical Phenomena, Sentence Head and Sentence Structure, Language Planning and Language Education, and over fifty journal articles in his fields. For this academic year, he visits the Nanyang Technological University for academic collaboration.
Manifestation of Certainty in Semantics: Case Study in Mandarin Chinese
Speaker: Prof Jiun-Shiung Wu, National Chung Cheng University, Taiwan
Abstract: This paper examines three modal adverbi-als in Mandarin Chinese: yídìng, kěndìng and dǔdìng. These three lexical entries can all express epistemic necessity or in-tensification. However, denoting intensi-fication, kěndìng and dǔdìng have addi-tional semantic requirements. First, they both require that there be at least one al-ternative to the proposition they present. Second, the speaker uses kěndìng to as-certain the truth of a proposition it takes, although all the alternatives are potentially true. Third, dǔdìng is used to assert the certainty that only the proposition it takes is true. Concerning certainty, two cases are demonstrated here. For yídìng, certainty is used implicitly, because certainty manifests itself through the speaker’s attitude. However, for kěndìng and dǔdìng, certainty is revealed explicitly, since (part of) the semantics of these two lexical items is certainty.
Biography: Jiun-Shiung Wu is a professor of Linguistics in Institute of Linguistics, National Chung Cheng University, Taiwan. He received his Ph.D. degree in Linguistics from University of Texas at Austin, Texas, U.S.A. in May, 2003. His research interests include formal semantics, formal pragmatics, computational semantics, discourse analysis, interface between syntax and semantics, etc. In the recent years, he focuses on aspect, tense, temporal relations, discourse structure, anaphor resolution, modality, etc. He has published more than a dozen papers in prestigious journals, including Journal of East Asian Linguistics, Journal of Chinese Linguistics, Language and Linguistics, Taiwan Journal of Linguistics, Cahiers de Linguistique – Asie Orientale, International Journal on Asian Language Processing, Journal of Chinese Language Teaching, etc. He attends at least one conference and presents a paper every year since 2003. He has several book chapters as well. Furthermore, he published a monograph entitled Temporal and Atemporal Relations in Mandarin in 2007, and a textbook entitled Computational Linguistics for Humanities Students: A First Course [in Chinese] in 2015. In terms of academic service, he served as the Chair of Department of Foreign Languages, National Chiayi University, from August, 2008 to January, 2010. He also served as a board member at Linguistic Society of Taiwan since 2008 to 2014. He was the President of Linguistic Society of Taiwan from 2010 to 2012.
Integrating Character Representations into Chinese Word Embedding
Speaker: Prof Peng Jin, Leshan Normal University, China
Abstract: We propose a novel word representation for Chinese based on a state-of-the-art word embedding approach. Our main contribution is to integrate distributional representations of Chinese characters into the word embedding. Recent related work on European languages has demonstrated that information from inflectional morphology can reduce the problem of sparse data and improve word representations. Chinese has very little inflectional morphology, but there is potential for incorporating character-level information. Chinese characters are drawn from a fixed set – with just under four thousand in common usage – but a major problem with using characters is their ambiguity. In order to address this problem, we disambiguate the characters according to groupings in a semantic hierarchy. Coupling our character embeddings with word embeddings, we observe improved performance on the tasks of finding synonyms and rating word similarity compared to a model using word embeddings alone, especially for low frequency words.
Biography: Peng Jin, is an professor and Dean of School of Computer Science at Leshan Normal University. He is the founding director of Internet natural language intelligent processing lab authorized by the Education of Department of Sichuan Province. He received his B.Sc., M.Sc. and Ph.D. degrees in computer science from Zhengzhou Textile Institute, Nanjing University of Science and Technology, Peking University, China, in 1999, 2002 and 2009, respectively. From Nov. 2007 to Apr. 2008, sponsored by CSC, he studied as a visiting research fellow in department of informatics, Sussex University, UK. From Aug. 2014 to Feb. 2015, he visited Sussex University for the second time. His research focuses on natural language processing, especially the statistical methods for lexical semantic and language resource. In these areas, he has published more than 30 papers in well-known international journals or conferences and inventor of one Chinese patent. He has lead two NSFC projects and several collaborative projects with industry. He served as organization chair of CIS2013 and YSSNLP2014, Program Committee members of NLP&CC'13, NLPOE'09-14, etc.