Lexical semantics is about how we can describe the meaning of words -- what does a word mean on its own, and how does the word's meaning change when we combine it with others. Word meaning is central to the area of Natural Language Processing; all modern NLP applications require knowledge of lexical semantics, and rely on the outcome of previous lexical semantics processing.
Word meaning might appear as self-explanatory, but it is not. As computational linguists, our goal is often to determine meaning differences automatically, and to apply them in many applications. This course shows how we can profit enormously from the relevant theoretical and practical approaches in Linguistics and Computational Linguistics. This course carries together such relevant approaches, and provides an overview of the area of lexical semantics. It treats lexical properties of words such as their subcategorisation, their exact word sense, and others. The questions in lexical semantics treated here are well-known, but are rarely taught in the form of a specialised course that aims to be theoretical thorough but also enable students to build better-informed and semantically more precise NLP applications.
The style of the class is lectures mixed with students' own annotation of sample texts. Students perform homework annotations in most classes to experience by themselves the phenomena treated each week.
At the end of the course students should be able to
(a) explain what "lexical semantics" is and its relationship to natural language processing
(b) explain several phenomena in lexical semantics treated in detail and appreciate their difficulty through their own annotation of textual material (homework)
(c) explain and understand computational approaches to each of the lexical semantics areas considered in detail
(d) be able to design and implement a sample application based on the methods above.
Specialist skills will be developed. Communication skills will be developed as well, because the course language is English and because student presentations are part of the assessment. Critical skills will also be developed, as the textbook reading is supplemented by research articles.
word sense, word sense disambiguation, lexical semantics, subcategorisation frame, sentiment detection, figurative language, entailment, pragmatics
✔ 専門力 | ✔ 教養力 | コミュニケーション力 | 展開力(探究力又は設定力) | 展開力(実践力又は解決力) |
Specified in the class
授業計画 | 課題 | |
---|---|---|
第1回 | Overview | 講義において指定する |
第2回 | Word senses -- the phenomena | |
第3回 | Word sense disambiguation algorithms (symbolic) | |
第4回 | Word sense disambiguation algorithms (ML) and evaluation | |
第5回 | WordNet, Lexical Chains and Learning of WN relations | |
第6回 | Lexical semantics of verbs | |
第7回 | Subcategorisation frame acquisition and selectional restrictions | |
第8回 | Lexical semantics of adjectives | |
第9回 | Adjective-based sentiment detection | |
第10回 | Lexical semantics of nouns and Noun-Noun relations | |
第11回 | Figurative Language 1 -- Metonymy | |
第12回 | Figurative Language 2 -- Metaphor | |
第13回 | Entailment and the RTE task | |
第14回 | Introduction to Pragmatics; Presuppositions | |
第15回 | Wrap-up and Outlook |
Background reading (papers) will be announced ahead of each lecture on the website.
Jurafsky and Martin, Speech and Language Processing, 2nd edition, 2008.
10 % Contribution to Class discussion.
20 % Annotation Homework.
70 % Literature-based Essay or Report on Student-Designed Experiment.
Programming Ability.
Good English ability (in order to appreciate subtle details)
Completion of Course "Natural Language Processing" is preferable.