|| Previous Lecture || Next Lecture || Assignments ||

CTL5403  SEMANTICS and DISCOURSE

Haihua PAN (Dr.), CTL, CityU., HK

Office: Rm B7610; Phone: 2788 8795; Email:  cthpan@cityu.edu.hk

Meaning Components

1.  Lexical Relations in CA

• Words are not the smallest semantic units but are built up of smaller components of meanings which are combined differently to form different words;
• The smaller components are called semantic components or primitives, and the analysis employing these primitives is often called Componential Analysis (CA);
• Three reasons for CA: (a) allow an economic characterization of lexical relations and sentence relations; (b) have linguistic import outside semantics, e.g. syntactic and morphological processes; and (c) form part of our psychological architecture, and provide a unique view of conceptual structure.
• Hyponymy: A lexical item P can be defined as a hyponym of Q if all the features of Q are contained in the feature specification of P. For example, spinster [female, adult, human, unmarried] is a hyponym of woman [female, adult, human].
• Antonymy (or incompatibility): Lexical items P, Q, R, ... are incompatible if they share a set of features but differ from each other by one or more contrasting features. For example, bachelor [male, adult, human, unmarried], spinster [female, adult, human, unmarried], and wife [female, adult, human, married] are incompatible with one another.
• Componential analysis often makes use of binary features and redundancy rules.
• Binary features: [±human], [±adult], [±female], [±married], etc.
• Redundancy rules: human ---> animate, adult ---> animate, animate ---> concrete, married ---> adult, etc.

2.  Katz's Semantic Theory

• Katz and Fodor (1963), Katz and Postal (1964) and Katz (1972).
• Semantic rules have to be recursive for the same reason as syntactic rules.
• The relationship between a sentence and its meaning is not arbitrary and unitary, and meaning is componential. The way words are combined into phrases and phrases into sentences determines the meaning of the sentences.
• The semantic component parallels the syntactic component. It gives (a) specifications of meanings of lexical items (dictionary); (b) rules showing how the meanings of lexical items build up into the meanings of phrases and so on up to sentences (projection rules); and (c) it does (a) and (b) in a universally applicable metalanguage.
• The typical dictionary item consists of three parts: (a) grammatical information in curly brackets; (b) semantic markers in parentheses (features mentioned earlier); and (c) distinguishers in square brackets which is the idiosyncratic semantic information that identifies the lexical item. (See example 9.12 on p. 235)
• The projection rules are responsible for showing how the meaning of words combines into larger structures. They interface with a syntactic component. They can give us something like compositionality.
• Selection restriction: represented in the dictionary as angle brackets. (See examples 9.13 and 9.14 on p. 236)
• The theory is decompositional.
• Example 9.16

3.  Grammatical Rules and Semantic Components

• Some linguists claim that with semantic components, one can describe grammatical processes correctly. i.e., that certain units of meaning are shared by different lexical items.
• Four English transitive verbs cut, break, touch, hit behave differently with respect to the following constructions: middle, conative, and body part ascension (see examples on pp. 238 and 239). These four verbs are differentiated as follows: touch is a pure verb of contact, hit is a verb of contact by motion, cut is a verb of change of state by moving something into contact with the entity that changes state, and break is a pure verb of change of state. (Levin 1993).

        We can have the following use features like change, motion, contact, and cause:

        cut:         cause, change, contact, motion

        break:    cause, change

        touch:    contact

        hit:          contact. motion

The information above corresponds to Katz's markers not distinguishers which specify the meaning proper of the words in question, and only the former (markers) participate in grammatical processes, as claimed in Pinker (1989), and only the former are shared by different words.

• Thematic roles and linking rules: different semantic features contained in a verb may correspond to different syntactic behaviors of the verb. The construction involved is locative alternation.

        Vera sprayed paint onto the wall.

        Vera sprayed the wall with paint.

• Levin and Rappaport (1991) suggest to set up semantic verb classes: (a) verbs which share the semantic structure "X causes Y to go away from Z", e.g. remove, take; (b) verbs that share the same semantic structure "X causes Y to go away from Z" but include specification of the means of removal, either (i) the manner of removal, e.g. wipe, rub, scrub, etc, or (ii) the instrument of removal, e.g. brush, hose, mop; and (c) verbs which have the semantic structure "X causes Z to change by removing Y", i.e. change of state verbs which focus on the resultant state, e.g. clear, empty, drain.

4.  Components and Conflation Patterns

• Talmy's work (1975, 1983, 1985)

        Figure: an object moving or located with respect to another object (the Ground)

        Motion: the presence per se of motion or location in the event

        Path: the course followed or the site occupied by the Figure object with respect to the Ground object

        Manner: the type of motion

        For example: Charlotte swam away from the crocodile. 

Charlotte is the Figure; the Ground is the crocodile; the Path is from; and the verb encodes the Manner of motion: swam.

In the following example, The banana hung from the tree. the banana is the Figure; the tree is the Ground; from is the Path; and Manner is again expressed in the verb hung.

Languages differ in the way how the different components are encoded.  In English, Manner is encoded in the verb like running, while the direction or Path is encoded in external prepositional phrase.

5.  Jackendoff's Conceptual Structure

• Conceptual semantics (Jackendoff 1987, 1990, 1992): a decompositional theory.
• Mentalist Postulate: Meaning in natural language is an information structure that is mentally  encoded by human beings.
• Sentence meaning is constructed from word meaning
• The semantic component CAUSE:

        George killed the dragon. The dragon died.

        There is an entailment relation between the two sentences above. The first entails the second.

        x killed y entails y died

        Similar entailments:

        x lifted y entails y rose

        x gave z to y entails y received z

        x persuaded y that P entails y came to believe that P

        x cause E to occur entails E occur

• That is, there is a semantic element CAUSE that occurs in many lexical items, and as a result, produces many entailment relations.
• Semantic decomposition can be used to investigate the mapping between semantics and grammatical processes
• The semantic components: event, state, material thing (object), path, place, property.
• The basic conceptual situations are event and state.

        For example, Bill went into the house.

        [s [np Bill] [vp [v went]  [pp [p into] [np the house]]]]

        [event GO ([thing BILL], [path TO ([place IN ([thing HOUSE])])])]

• The motion event has: the motion itself GO, which is further composed of two categories: the entity or THING, moving, and the trajectory, or PATH. This Path may have a destination or PLACE, where the motion ends.

        reads as "Bill traverses a path that terminates at the interior of the house."

        Another example, The car is in the garage.

        [s [np The car] [vp [v is] [pp [p in] [np the garage]]]]

        [state BE ([thing CAR], [place IN ([thing GARAGE])])]

• Localist semantic fields: the above sentence describes a state of being in a spatial location. Jackendoff uses BE to represent the state of being. He differentiates four BE's: BEloc, BEtemp, BEposs, and BEident. (See p. 253)
• Complex event: CAUSE and INCH (inchoative).

        Examples: The pool emptied. 

                          [s [np The pool] [vp [v emptied]]]

                          [Event INCH ([State BEident ([thing POOL], [Place AT ([Property EMPTY])])])]

         John emptied the pool.

                         [s [np John ] [vp [v emptied] [np the pool]]]

                         [Event CAUSE ([thing JOHN],  [Event INCH ([State BEident ([thing POOL], [Place AT ([Property EMPTY])])])])]

• Formation Rules:

        [EVENT] ---> [Event GO ([THING], [PATH])]

        [STATE] ---> [State BE ([THING], [PLACE])]

        [PATH]   ---> [TO ([PLACE])]

        [PLACE] ---> [IN ([THING])]

        [PLACE] ---> [AT ([TIME])]

        [PLACE] ---> [AT ([PROPERTY])]

        [PLACE] ---> [AT ([THING])]

        [EVENT] ---> [Event INCH ([STATE])]

        [EVEVT] ---> [Event CAUSE ([THING], [EVENT])]

• Nominal categories:

Material entity can be divided into four groups: individuals, groups, substance, and aggregates. They are differentiated by two features [± bounded] and [± internal structure], simply [± b] and [± i].

individuals: [+b, -i]   (count nouns)

groups:        [+b, -i]   (collective nouns)

substance:    [-b, -i]   (mass nouns)

aggregates:  [-b, +i]   (plural nouns)

• [± bounded] can also be used to describe verbs.
• Plurals can change the features of nouns (see p.257).

        brick [+b, -i] ------> bricks [-b, +i]

• Problems with components of meaning

i. identification of semantic primitives which are simply a variation of, or equivalent to, the necessary and sufficient conditions approach to word meaning that was discussed in Chapter 2 is very difficult: knowing how to validate any proposed set of primitives, and when to stop identifying them, i.e. knowing what are the right features and how many is enough.

ii. the use of metalanguages. the devices provided are ad hoc and unsystematic: at best another arbitrary language; at worst, a kind of garbled version of English, French, etc. of the writer.