Semantics 1, SoSe 2019 (Sailer)

From Lexical Resource Semantics
Jump to: navigation, search

General information

Course description

Semantics is the study of the (literal) meaning of words and sentences. The meaning of a sentence is usually predictable from the words in the sentence and its syntactic structure. Yet, this relationship between form and meaning is not a simple one-to-one mapping. Instead, it is rich in ambiguities, pleonastic marking and elements without any identifiable meaning contribution. We will work on an account that is founded on classical tools of semantic research but still directly addresses these empirical challenges. After the class, the participants will be able to identify - and partly analyze - interesting semantic phenomena in naturally occurring texts. They will have acquired a basic working knowledge in formal logic, which they will be able to apply in the description of meaning

Time and place

  • Tuesday 08:15-9.45
  • Starting: 16.10.2018
  • Room: IG 251 (IG-Farben-Haus)

Olat course

Direct link: https://olat-ce.server.uni-frankfurt.de/olat/auth/RepositoryEntry/5912854558

Password: Please send an e-mail to the lecturer (sailer@em.uni-frankfurt.de)

Modules

  • Lehramt Englisch (L2/5, L3): FW 2A, FW 2B
  • BA English Studies: 3.4(1)
  • BA Empirische Sprachwissenschaft: K 6.1

Contact

Manfred Sailer
e-mail: sailer@em.uni-frankfurt.de
office: IG 3.214
office hours: contact via e-mail!
www: http://user.uni-frankfurt.de/~sailer/index.htm

Course requirements

L2 and L5

  • regular attendance
  • pass all assignment sheets
  • Modulprüfung (optional): 90 min written exam (2 CP): 23.7.2019, 11am-1pm

L3

  • regular attendance
  • pass all assignment sheets
  • Modulprüfung (optional):
    • 20 min. oral exam
    • not possible: kleine Hausabeit

MSc Wirtschaftspädagogik

  • regular attendance
  • do all assignment sheets
  • Modulprüfung (1CP): Commented literary scenario


BA English Studies

  • regular attendance
  • pass all assignment sheets
  • literary scenario:
Part 1: Extract 15 ambiguous sentences from the text such that all types of ambiguity covered in class are represented provide unambiguous paraphrases of the readings determine the type of ambiguity
Part 2:
Define a formal model consisting of 3 characters from your text, which contains 2 properties, 1 2-place relation
Formulate 2 atomic formulae and compute their truth value.
Formulate 4 complex formulae with at least 1 logical connective in each and compute their truth value.
Formulate 1 complex formula with at least 2 logical connectives in

it and compute its truth value.

BA Empirische Sprachwissenschaft

K 6.1

  • regular attendance
  • Modulprüfung (obligatory): 90min. written exam: 23.7.2019, 11am-1pm

En 4.1

not possible: You have done this course as part of K6.1, so you can directly do constraint-based Semantics 2.

DH 6.1

not possible: You have done this course as part of K6.1, so you can directly do constraint-based Semantics 2.

Erasmus 6 CP

  • regular attendance
  • pass the assignment sheets
  • 90min. written exam: 23.7.2019, 11am-1pm
  • small literary scenario:
Part 1: Extract 4 ambiguous sentences from the text such that different types of ambiguity covered in class are represented provide unambiguous paraphrases of the readings determine the type of ambiguity
Part 2:
Define a formal model consisting of 3 characters from your text, which contains 2 properties, 1 2-place relation
Formulate 2 atomic formulae and compute their truth value.
Formulate 2 complex formulae with at least 1 logical connective in each and compute their truth value.
Formulate 1 complex formula with at least 2 logical connectives in it and compute its truth value.

The grade will be determined by the result of the written exam.

Meeting 14: Mock exam

The mock exam is available as a pdf file: mock-exam-sose19.pdf See the tasks with solutions at: Mock exam, SoSe 2019 (Sailer)

Meeting 12

Meeting 11

Basic combinatorics: Canonical examples

(the following exercises are adapted from the textbook material to Chapter 5.

1. Sentence: Pat snored.
Logical form: snore(pat)
Which parts of the logical form are contributed by which word?

pat ¦ snore ¦ __ ( __ )
Pat
snored

2. Sentence: Pat likes Chris.
Logical form: like(pat,chris)
Which parts of the logical form are contributed by which word?

pat ¦ chris ¦ like ¦ __ ( __ , __ )
Pat
likes
Chris

Your score is 0 / 0


Possible EX-CONT values

Given the following PARTS lists, what are possible EX-CONT values (if we do not assume other restrictions)

1. PARTS < pat, alex,like, __ (__,__) >

Check your answer

like(pat,alex)
like(alex,pat)


2. PARTS < alex,snore, __(__), ¬(__) >

Check your answer

¬(snore(alex))


3. PARTS < alex,alex,snore >

Check your answer

There is no possible EX-CONT value because the three elements on the PARTS list cannot be combined.


3. PARTS < alex,alex,snore, __ (__) >

Check your answer

snore(alex)

4. PARTS < alex,alex,snore, __ (__), __ ∧ __ >

Check your answer

snore(alex) ∧ snore(alex)

Meeting 10

Videos

The following video (21') introduces the way we will write down lexical entries in our course.

The next video shows the HPSG analysis of sentences (38').


Lexical entries as Attribute-Value Matrix

Provide the required information on the lexical properties of the underlined words in the following sentences.
Note:

  • Put a minus ("-") if a slot should not receive any filling
  • Use det, noun, prep or verb for the HEAD values.

1. Alex read a book yesterday.

PHON < >
HEAD
SUBJ < >
SPR < >
COMPS < >

2. Alex talked to a friend.

PHON < >
HEAD
SUBJ < >
SPR < >
COMPS < >

3. Pat liked this new documentary on African wild life.

PHON < >
HEAD
SUBJ < >
SPR < >
COMPS < >

4. Alex talked to a friend.

PHON < >
HEAD
SUBJ < >
SPR < >
COMPS < >

Your score is 0 / 0


Feel free to send feedback on this exercise to Manfred Sailer.

Analysis of simple sentences

In the following examples, we assume a feature SYNSEM (for "syntax and semantics") whose value comprises the features HEAD and VAL. We also assume that we find SYNSEM-values on the valence lists.

1. Indicate the missing values of the VAL and the HEAD features using tags ([1], ...) or "-" for empty lists.

Alex snored.
syntactic structure: Tree-AlexSnored.jpeg
Words:                                                                                                   Phrase:
Alex                                                             snored                                    S: Alex snored.
HEAD [4]noun                                  HEAD [5]verb                                    HEAD
SUBJ < >                                  SUBJ < >                                    SUBJ < >
SPR   < >                                  SPR < >                                     SPR < >
COMPS < >                              COMPS < >                               COMPS < >

Your score is 0 / 0


1. Indicate the missing values of the VAL and the HEAD features using tags ([1], ...) or "-" for empty lists.

Fido chased a mouse.
syntactic structure: Tree-FidoChasedAMouse.jpeg
Words:
Fido                                                             chased                                    a                                                              mouse
HEAD [8]noun                                  HEAD [9]verb                                    HEAD [10] det                                   HEAD [11] noun
SUBJ < >                                  SUBJ < >                                    SUBJ < >                                   SUBJ < >
SPR   < >                                  SPR < >                                     SPR < >                                     SPR < >
COMPS < >                              COMPS < >                               COMPS < >                                COMPS < >
Phrases:                                                                                                  
NP: a mouse                               VP: chased a mouse                                S: Fido chased a mouse.
HEAD                                      HEAD                                        HEAD                               
SUBJ < >                                  SUBJ < >                                    SUBJ < >
SPR   < >                                  SPR < >                                     SPR < >
COMPS < >                              COMPS < >                               COMPS < >

Your score is 0 / 0


1. Indicate the missing values of the VAL and the HEAD features using tags ([1], ...) or "-" for empty lists. Don't use spaces.

Pat gave Alex a ride.
syntactic structure: Tree-PatGaveAlexARide.jpeg
Words:
Pat                                                             gave                                         Alex                                                              a                                      ride
HEAD [9]noun                                  HEAD [10]verb                                   HEAD [11] noun                                HEAD [12] det                                HEAD [13] noun
SUBJ < >                                  SUBJ < >                                    SUBJ < >                                   SUBJ < >                                 SUBJ < >
SPR   < >                                  SPR < >                                     SPR < >                                     SPR < >                                   SPR < >
COMPS < >                              COMPS < >                        COMPS < >                                COMPS < >                            COMPS < >
Phrases:                                                                                                  
NP: a ride                               VP: gave Alex a ride                                S: Pat gave Alex a ride.
HEAD                                      HEAD                                        HEAD                               
SUBJ < >                                  SUBJ < >                                    SUBJ < >
SPR   < >                                  SPR < >                                     SPR < >
COMPS < >                              COMPS < >                               COMPS < >

Your score is 0 / 0


Feel free to send feedback on this exercise to Manfred Sailer.

Meeting 9

No special exercises.

Meeting 8

Video

Watch the following video (33') on the basic step in a syntactic analysis as we need it in our course.

The next video (14') introduces the way we talk about syntactic trees. Please watch it.


Meeting 7

Video

Watch the following video on logical determiners:

Exercises

After having watched the video, work on the following tasks.

Task 1 Identify the determiners in the following sentence.

(a) Juliet talked to some stranger at the party.

(b) Every Capulet is an enemy to some Montague.

(c) Many people in Verona are not happy about the Capulet-Montague feud.

Check your solutions here:

(a) some

(b) every, some

(c) many


Task 2 Identify the formula that corresponds to the translation of the sentence.

1. Some Montague who was at the party fell in love with Juliet.

x (montague1(x) : (at-party1(x) ∧ fall-in-love-with2(x,juliet)))
In restricted quantifier notation, the "complete" semantic representation of the noun phrase (NP) appears in the restrictor (-> square brackets).
x ((montague1(x) ∧ at-party1(x)) : fall-in-love-with2(x,juliet))
x (montague1(x) : (at-party1(x) ∧ fall-in-love-with2(x,juliet))
In restricted quantifier notation, the semantic representation of the noun phrase (NP) appears in the restrictor.
x ((montague1(x) ∧ fall-in-love-with2(x,juliet)) : at-party1(x))
In restricted quantifier notation, the semantic representation of the noun phrase (NP) appears in the restrictor, that of the VP in the scope.

Your score is 0 / 0


Task 3 The sentence: Some Tybalt loved some Montague. is translated into the formula
∃ y (montague1(y) : love2(tybalt,y).

1. Mark all the cells in the table that stand for a true statement.

montague1(y) zwisch love2(tybalt,y)zwisch
Romeo
Mercutio
Juliet
Tybalt
Laurence
Paris

Your score is 0 / 0


Given this table, is the overall formula true or false? (Give a reason for your answer.)

Check your solutions here:

The formula is false, because there is no individual in our model for which both the restrictor and the scope are true.


Task 4 Variable assignment function
Start with the following variable assigment function g: g(u) = Romeo, g(v) = Juliet, g(w) = Romeo, g(x) = Laurence, g(y) = Mercutio, g(z) = Juliet

Provide the changed variable assignment function g[v/Paris].

Check your solutions here:

g[v/Paris](u) = g(u) = Romeo
g[v/Paris](v) = Paris
g[v/Paris](w) = g(w) = Romeo
g[v/Paris](x) = g(x) = Laurence
g[v/Paris](y) = g(y) = Mercutio
g[v/Paris](z) = g(z) = Juliet

Meeting 5

Computing the truth value of complex formulae

Connectives

The following video presents the step-by-step computation of the truth value of two formulae with connectives. The example uses a model based on Shakespeare's play Macbeth. The two formulae are:

  • ¬ king(lady-macbeth)
  • king(duncan) ∨ king(lady-macbeth)

The next video shows how the truth value of a more complex formula can be computed. The example contains two connectives:

kill(malcom,lady-macbeth) ∨ ¬thane(macbeth)

The video shows two different methods: top down and bottom up.

The next video shows how the truth value of a more complex formula can be computed. The example contains two connectives:

kill(malcom,lady-macbeth) ∨ ¬thane(macbeth)

The video shows two different methods: top down and bottom up.


Truth tables

Truth tables are also useful to compute the truth value of complex formulae. This is shown in the following podcast, created by Lisa Günthner.



Back to the course page.

Meeting 3

Computing the truth value of atomic formulae

The following video presents the step-by-step computation of the truth value of two atomic formulae. The example uses a model based on Shakespeare's play Macbeth. The two formulae are:

  • kill2(macbeth,duncan)
  • kill2(lady-macbeth,macbeth)

Back to the course page.

Meeting 2

Our literary scenario

Literary scenario for this course: The Hunchback of Notre Dame, wikipedia summary at https://en.wikipedia.org/wiki/The_Hunchback_of_Notre_Dame_(1996_film)


Why it is too difficult to go directly from language to the world

The following architecture is extremely useful when talking about semantics:

  1. A natural language expressions: Harry meets Hagrid.
  2. ... is mapped to some expression from a formal language (here: predicate logic): meet2(harry,hagrid)
  3. This logical expression is then interpreted with respect to our scenario/world: The formula meet2(harry,hagrid) is true, because, in our scenario, Harry meets Hagrid.

The following properties of natural language make it useful to use the intermediate step of a logical language:

  1. The same expression can have different meanings (ambiguity).
  2. Different expressions can have the same meaning (synonyms, paraphrases)

Find examples for the above-mentioned properties (ambiguity, synonymy, paraphrases).

Check your answers

1. one form, two meaingns: Ambiguity: (see earlier in this meeting and the slides of last week's meeting)

1.a Ambiguous words: date (fruit or point in time); bank (financial institute or bank of a river)

1.b. Ambiguous sentences: Sycorax and Prospero were stranded on the island with their children.

2. two forms, one meaning:

2.a Synonymous words: couch - sofa; instant - moment

2.b Paraphrases:

  • active-passive pairs: Prospero set Ariel free. - Ariel was set free by Prospero.
  • cleft sentences: Prospero set Ariel free. - It was Prospero who set Ariel free.
  • different ways to express a possessor: Sycorax was the first inhabitant of the island. and Sycorax was the island's first inhabitant.

Towards a formal model

First steps

The following material is an adapted form of material created by student participants of the project e-Learning Resources for Semantics (e-LRS). Involved participants: Lisa, Marthe, Elisabeth, Isabelle.

You can think of building a formal model like being the producer of a film who has to collect everything that should be included in the film.

Here is a very simple story from which we can derive an example model.

1. Mark those elements that we need in a model.

relations
Yes. We use relations to express what is true between various individuals. For example the relation grandmother-of.
individuals
Yes. In the video, we have three individuals, Red Riding Hood, Grandmother, and Wolf.
nouns
nouns are a syntactic category and as such part of the language, not of the "world".
properties
Yes. The video mentions some properties such as having a red hood.
relatives
(this is a nonsense alternative)

Your score is 0 / 0


1. What is the status of the following entities in the video on Little Red Riding Hood?

individualpropertyrelation
Red Riding Hood
lives in the forest
Grandmother
is afternoon snack for
has a red hood
has a big mouth
is grandmother of

Your score is 0 / 0


The universe and name symbols

Task: Assume three individuals from our Game of Thrones-scenario.

Formally we collect the individuals of our model in a so-called universe (U). For the fairy-tale story, we can define the universe as follows:

U = {Redridinghood, Grandmother, Wolf}

Do a similar definition for your own scenario.


We can introduce name symbols for some of our individuals. For example: redridinghood, grandmother, wolf.

We link the name symbols to the individuals in our modal. To do this, we introduce the interpretation function. We will written the interpretation function as as I.
This function can be defined in the following way:

I(grandmother) = Grandmother
I(redridinghood) = Red Riding Hood
I(wolf) = Wolf


Properties and predicate symbols

A property is a specification that either holds of an individual or not. In the little story, having a big mouth is a property of the Wolf, but of noone else in the story. Being female holds of both Little Red Riding Hood and the Grandmother.

We can think of a property as the set of individuals that have this property. Under this view, the property of being female would be the set {Redridinghood, Grandmother}.

Alternatively it is convenient to think of properties as 1-place relations. Under this view, the property of being female would be a set of lists of length 1. This is what the property of being female then looks like: { <Redridinghood>, <Grandmother> }

Task: Using your Game of Thrones universe, define two properties in the format of 1-place relations.

Just like before, we want to have symbols that we can use in the logical language. For our example, let's take the predicate symbols female_1 and has-big-mouth_1. There interpretation is defined as follows:

I(female_1) = { < x > | x is female } = { <Redridinghood>, <Grandmother> }.

Task: For each of your properties, invent an appropriate predicate symbol. Define its interpretation.

Meeting 1

Literary scenario for this course: The Hunchback of Notre Dame, wikipedia summary at https://en.wikipedia.org/wiki/The_Hunchback_of_Notre_Dame_(1996_film)