Semantic role

from Wikipedia, the free encyclopedia

Semantic roles - also called thematic roles - are a concept of linguistics with which the meaning properties are to be recorded, which are assigned to the additions of a predicate because they are interpreted in connection with this predicate. So it's a concept that combines semantics and grammar. The most common case is verb additions: if a verb denotes an event, then B. Subject and object of the verb Participants of this event, who take on different "roles".

For example, in the sentence The cat ate the mouse, the subject "the cat" can be characterized as the participant who caused the event (a so-called agent ) and the object "the mouse" as the participant who experiences a change in the event ( Patiens or named topic).

The concept of semantic roles is significant through attempts to explain grammatical properties of parts of sentences from such meaning properties, e.g. B. the assignment of supplements to the syntactic functions subject and object, or variation in the assignment of case etc. Due to the recourse to meaning properties, it is expected that this can be used to formulate conditions that represent universals, i.e. are independent of the grammar of a single language . Semantic roles play a central role in many functionalist, language-comparative theories of grammar, but on the other hand have also been criticized for the fact that they cannot be precisely defined and that they do not result in a consistent system that can fully predict grammatical patterns.

History of the term

The concept of semantic roles has been gaining attention since the late 1960s. The linguistic discussion at this time was largely determined by the reception and examination of the syntax theory of Noam Chomsky . With Aspects of the Theory of Syntax, Chomsky had presented a version of Generative Transformation Grammar in 1965 that took certain semantic aspects into account, but still gave syntax the central role in language description. In the period after 1965, many linguists tried to move semantics more into the center of linguistic theory. In addition to the so-called generative semantics ( George Lakoff , James D. McCawley (1938–1999)), the case grammar approach was seen as an important attempt to integrate semantics into a generative description of language.

The most influential contribution to this early phase is the essay The Case for Case , written by Charles J. Fillmore in 1968 . He differentiated between grammatical "surface case" ( nominative , accusative , genitive , dative ) and, on the other hand, semantically characterized " deep case " ( agentive , instrumental , dative, factive, locative and objective ). These deep casus characterize the semantic relationships of the various noun phrases that are mainly given by a verb.

In the sentence (1) 

(1) John öffnet die Tür mit dem Schlüssel.

the surface case of key is the dative;

in the sentence (2) 

(2) Der Schlüssel öffnet die Tür.

is key in the nominative.

According to Fillmore's concept of deep case, 'key' is instrumental in both cases in terms of its semantic role. John is in (1) agentiv. In addition to Fillmore's work, the approaches of Jeffrey S. Gruber (1967) and Ray Jackendoff (1972) also play a role in this initial phase. You do not speak of deep casus , but of case roles .

The approaches were widely received and integrated into a whole range of grammar models. In the particularly influential version of Chomsky's generative grammar from 1981, the “GB” model ( Lectures on government and binding ), the concept was then also indirectly reflected in the term “ theta role ”, which is used for this purpose to record the identity of a noun phrase in the course of a transformational derivation and to distinguish between interpretable and non-interpretable expressions (such as an expletive ). However, Chomsky did not commit to a specific semantic content of his theta roles, but used them as a purely grammatical mechanism. Therefore, Chomsky's concept of the “theta role” must be distinguished from that of the “semantic role”.

While grammar theories have been using semantic roles to varying degrees since then, semantic research began relatively late to take up and analyze the concept of semantic roles. In the work of David Dowty, a definition was attempted that traces the concept of semantic role back to conclusions from a predicate meaning (see below). Dowty was very skeptical about the feasibility of the traditional concept of semantic roles and replaced it with a model that reduces the set of semantic roles to two prototypical categories, i.e. H. Proto-agent or patiens.

Basics: predicates and arguments

Semantic roles (or: "thematic roles") are assigned to the arguments of a predicate . In a logical representation, the arguments can be noted as an ordered sequence of variables for a predicate, e.g. B. the transitive verb "kill" has two arguments that can be represented in the order corresponding to the sequence "subject - object" in the grammar. 

töten (x,y)
Im Beispiel: Die Katze tötete die Maus
entsteht die Belegung: x = "die Katze", y = "die Maus"

An orderly sequence is required, since swapping the assignment in the above example would result in a completely different meaning. But the question arises as to what makes such a fixed order come about. The term of the semantic role now means that content descriptions can be given for x and y, which then also determine the order of the arguments.

Approaches to the definition of semantic roles

(Individual) semantic roles

An attempt to give a precise logical definition of semantic roles can be found for the first time in works by David Dowty , which appeared around 1990. Here the "thematic role of an argument" is defined as the set of conclusions that can be drawn from the word meaning of a predicate about one of its arguments. The term "conclusion" means that the event description given by the verb meaning counts only for the definition of role properties, not typical assumptions about the situation or different perspectives that the representation of the event in the sentence or context opens up.

Dowty gives the following examples:

  • x murdered y /vs./ x kills y

From both verbs it follows:

y changes in the event (from alive to dead)
x is the cause of the change

However, only from "murder" follows

x intends the event as such (since "killing" can be intentional or unintentional in individual cases, there is no compelling inference to intent).
  • x persuaded y /vs./ x convinced y

has a similar contrast. From both verbs it follows:

y changes its view of an object in the event
x is the cause of this change

However, it only follows from persuasion

x intended the event as such (since "convince" can also happen when y takes something from what x says that changes his convictions without x having to have intended this, so the word meaning of "convince" does not guarantee any inference Intentionality of x).

Lists of conclusions can be drawn up for every single verb and for every single one of its arguments. Dowty refers to these as "individual thematic roles" (i.e., for individual argument positions).

Semantic role types

Individual roles are already suitable for distinguishing the arguments of a single verb from one another. Chomsky's mechanism of theta roles that a particular verb assigns to its syntactic arguments can also be seen as indexing an argument with an individual role. The main linguistically interesting question, however, is which classes of individual roles can be formed so that large groups of verbs and certain grammatical functions can be covered in general. In the example above, e.g. B. to see that the first argument of the verbs "persuade, convince, murder, kill " can generally be described as "causer", even if there are differences with regard to the conclusion "intentional causation". There is thus the possibility of defining a more general term "agent" which only contains the causative property. Roles such as "agents" are more precisely referred to as "role types" by Dowty, but they correspond to what is commonly understood as "thematic role".

If the observation that all arguments with the property "agent" appear as the first argument of a verb can be generalized, then this means that rules can be drawn up for the assignment of semantic roles to grammatical functions such as subject and object: for example that a "Agent" must always be assigned to the subject position (as far as the meaning of the verb is concerned, and before grammatical rules apply that block the appearance of an agent as a subject in infinitives or passive clauses, for example ). Such assignments are called "linking rules" (roles are "linked" with syntactical positions) or rules of argument projection (i.e. projection from semantics into grammar).

In order to build a general theory of the linking of arguments into the syntax on semantic role types such as "agent" etc., it is now necessary to define a limited set of role types that cover all the argument positions of all verbs and to bring them into a hierarchy which can be read off which of several arguments of a verb is entitled to the subject position first and which becomes the object. This is attempted in various grammar models. In the Role and Reference Grammar , a predominantly functionalist syntax model that relies heavily on meaning properties, z. For example, the following list is suggested: It is agreed that the names for the semantic roles are capitalized.

  • AGENT
Subjects of verbs like give, kill, dance
  • EXPERIENCER
Subtype: COGNIZER (subjects of verbs such as think, believe )
Subtype: PERCEIVER (subjects of verbs like see, feel )
Subtype: EMOTER (subjects of verbs like love, hate )
  • RECIPIENT
Dative objects of verbs like giving something to someone
  • STIMULUS
Accusative objects of verbs like see, hear, like
  • THEME
Accusative objects of verbs like "give, move (trans.)"
Subjects of verbs such as are (in a place) increase
  • PATIENS
Accusative objects of verbs like zerbrechen (trans.), Kill
Subjects of verbs like zerbrechen (intrans.), Die

Linking rules that can be formulated on the basis of this list can then read (for transitive verbs with 2 arguments):

  • The argument whose semantic role is relatively higher up the hierarchy becomes the subject.
  • The argument, whose semantic role is relatively below, becomes a direct object.

Such lists of role types, widely used in the literature, have been criticized for methodological reasons (e.g. by Dowty), because there is no well-defined procedure to decide which summaries should be made and because it is not clear that any such list is to be found each can completely cover all verb meanings with all their arguments. Other authors do indeed make other classifications (in a German work on sentence semantics that takes up thematic roles, P. von Polenz assigns 19 different roles, for example). Roles that are often used in other constellations but are missing in the above variant are e.g. B. Instruments or inanimate causers (such as natural forces; or, in another sense, logical facts in verbs that express logical relationships), these are sometimes added to the agent role or set separately from it. On the other hand, very many role lists do not specifically differentiate between "topic" and "patient" (the classic definition of the topic according to J. Gruber is that it is a participant who is localized or changes his location).

Overview of common semantic role types

The question of what a final list of roles (types) should look like has never been finally decided in the literature. Role designations used in the literature are summarized below, without such a list being exhaustive:

semantic role meaning example
ADDRESSEE to whom is the communicative action directed Paul told everything to the friend .
AGENT who controls the action Paula saws the wood.
BENEFICIAL whoever benefits, benefits from the action Paula bought her boyfriend a dog.
EXPERIENS (perception carriers ) who perceives a sensory impression, feeling Paul annoys himself ; The failure annoys him .
FACTITIVE what is made by the agent, what causes the agent's action Paula wrote Paul a letter .
INSTRUMENT what is used by the agent in the act Paula drove in the nails with a hammer .
COMITATIVE who accompanies another participant in the action Paul helped Paula with the nailing.
LOCUS where is the action located, where does it take place Paul works in Munich .
PATIENS who or what is influenced by the action, affected Paul drives the truck .
RECIPIENT whoever receives receives a transferred object in the act (recipient) Paul gave the truck keys to Paula .
THEME what is the cognitive, communicative or emotional situation about Paula was happy about the truck .
ORIGIN where does the movement of an action begin Paula ran out of the house .
TARGET where is the movement in the action directed Paula traveled to Hamburg .

Proto roles

In his very influential 1991 essay (see bibliography), David Dowty was very skeptical about the possibility of a system of role types that assigns each syntactic argument exactly one thematic role from a narrowly limited list:

In order for such systems to work in an account in which the roles Agent, Theme, Goal, etc., are given explicit semantic content, the meanings of all natural-language predicates must turn out to be of a very particular sort: for every Verb in the language, what the verb semantically entails about each of its arguments must permit us to assign the argument, clearly and definitely, to some official thematic role or other-it cannot be permitted to hover over two roles, or to 'fall in the cracks' between roles - and what the meaning entails about every argument must always be distinct enough that two arguments clearly do not fall under the same role definition. This is a very strong empirical claim about natural-language predicates, and, as soon as we try to be precise about exactly what agent, patient, etc., 'mean', it is all too subject to difficulties and apparent counterexamples.
(In order for such systems to work, in an approach in which the roles of agent, subject, goal etc. are given an explicit semantic interpretation, it would have to be shown that the meanings of all natural language predicates are of a very special kind: for every verb in the language it would have to be possible to clearly and definitely assign what the verb allows to infer about each of its arguments to some official thematic role - it could not be allowed that a conclusion remains in the balance once about two roles or "in the cracks" between two roles fall. And what inferences the verb meaning creates about each argument must always be so distinct that no two arguments [ie of the same verb] never fall under the same role definition. That is a very strong empirical claim about the predicates of natural languages, and, as soon as we try to specify what agents, patients, etc. exactly mean, it is all too difficult d exposed to apparent counterexamples.)

Dowty proposes to avoid the problems of a systematics of semantic roles by not striving for a complete coverage of all argument places of all predicates by a uniform description grid, but the formulation of semantic roles is only geared to the problem of which factors of importance for the arrangement of subject and object are responsible in transitive constructions. There are only two poles between which all conclusions can be arranged. Dowty calls them "proto-agents" and "proto-patients", and regards each of these two as prototypically organized categories; H. a family of features that form a connection with one another due to a variety of similarity relationships, but for which there need not be a common feature throughout. Dowty identifies a host of semantic features that make an argument subject-worthy, the so-called proto-agent properties:

  • Voluntary participation in an event or state
  • Sensation or perception
  • Causing an event or a change in the other participant
  • Movement relative to the other participant
  • possibly also: existence independent of the event

As semantic properties that predestine an argument for the place of the object in the transitive construction (proto-patient properties), he identifies:

  • The participant experiences a change
  • The participant is an incremental topic
  • affected by a causal effect by another participant
  • stationary relative to the other participant
  • no existence of the participant independent of the event, or nonexistence.

There is no common denominator in each of the categories, nor should participant roles clearly correspond to one such characteristic. The linking rule proposed by Dowty states that that participant is coded as a subject who has relatively more proto-agent or relatively fewer proto-patient properties. An important part of Dowty's model is the indication that there are cases where an argument is determined to be the subject only by exactly one of the diverse proto-agent properties (although, as I said, there is no uniform definition under which all cases fit) , e.g. B .:

  • Willful participation alone: "Hans ignored Maria", "Hans refused to eat for two days"
  • Perception alone: "Hans knows / fears Maria"
  • Causation alone: "loneliness causes depression"
  • Relative movement alone: "The probe passed the planet"
  • Independent existence alone: "Hans needs a new car" (here the object does not have to exist)

References to language processing

In psycholinguistic works on the subject, the status of semantic roles is often associated with their influence on the structure. The main question here is whether semantic roles as non-syntactic information units in the first pass of the understanding process ("first-pass parse") determine the syntactic structuring, which - if the answer is positive - leads to the assumption of an interactive and non-modular model of language processing forces (see, inter alia, Friederici et al., 1996; McRae et al., 1997).

literature

  • Anthony R. Davis: Thematic roles . In: Claudia Maienborn, Klaus von Heusinger, Paul Portner (Eds.): Semantics: an international handbook of natural language meaning . Vol. 1. Berlin 2011, pp. 399-420.
  • David Dowty: Thematic proto-roles and argument selection . In: Language 67-3, (1991), pp. 547-619.
  • Eva Engels, Sten Vikner: Sentences , cases and semantic roles: an introduction. Aarhus University, Denmark. Tidsskrift for Sprogforskning, Årgang 4, No. 1-2 (2006), pp. 17-37
  • Charles Fillmore : The case for case. In: Emmon Bach, Robert T. Harms (Ed.): Universals in linguistic theory. Holt, Rinehart & Winston, London 1972, pp. 1-88
  • Angela D. Friederici , Anja Hahne, Axel Mecklinger: Temporal structure of syntactic parsing. Early and late event-related potential effects. In: Journal of Experimental Psychology: Learning, Memory and Cognition. 22-5, (1996), pp. 1219-1248.
  • Jeffry S. Gruber: Studies in Lexical Relations. Bloomington, 1967
  • Jeffry S. Gruber: Thematic relations in syntax. In: Mark R. Baltin, Chris Collins (Eds.): The handbook of contemporary syntactic theory. Blackwell, Oxford 2000, ISBN 0-631-20507-1 , pp. 257-298.
  • Ray S. Jackendoff: Semantics and cognition. Cambridge Mass. 1972
  • Beatrice Primus: Semantic Roles. Winter, Heidelberg 2012, ISBN 978-3-8253-5977-5
  • Beatrice Primus: Participant roles . In: Nick Riemer (Ed.): The Routledge Handbook of Semantics . London 2016, pp. 403-418.
  • Ken McRae, Todd R. Ferretti, Liane Amyote: Thematic roles as verb-specific concepts. In: Language and cognitive processes. 12-2 / 3, (1997) 137-176.
  • Gisa Rauh: Tiefenkasus, thematic relations and theta roles. The development of a theory of semantic relations. Tübingen 1988, ISBN 3-87808-369-6
  • Robert Van Valin: Introduction to Syntax. Cambridge University Press, Cambridge 2008, ISBN 0-521-63566-7

Web links

Wiktionary: semantic role  - explanations of meanings, word origins, synonyms, translations

Individual evidence

  1. ↑ Beatrice Primus: Semantic roles. Winter, Heidelberg 2012, ISBN 978-3-8253-5977-5 , p. 3
  2. ↑ Thomas Stolz: Ergativ for the bloodiest beginners. University of Bremen, pp. 1–12
  3. ↑ direct object in non-case terminology
  4. ↑ Indirect object in non-case terminology
  5. ↑ Grammatical Categories and Relations. University of Bremen
  6. ↑ Dowty (1991), pp. 552f. (where earlier work of the author is taken up)
  7. ↑ "From the semantic point of view, the most general notion of thematic role (type) is A SET OF ENTAILMENTS OF A GROUP OF PREDICATES WITH RESPECT TO ONE OF THE ARGUMENTS OF EACH. (Thus a thematic role type is a kind of second -order property, a property of multiplace predicates indexed by their argument positions.) "(Dowty (1991), p. 552)
  8. ^ Dowty (1991), p. 550
  9. ↑ Van Valin (2001), p. 29. The references to possible subjects with the same roles are not found in this list, but are added here
  10. ↑ Peter von Polenz : German sentence semantics. Basic concepts of reading between the lines. 3rd edition, Verlag Walter de Gruyter, Berlin 2008, ISBN 978-3-11-020366-0
  11. ↑ Gruber (1967)
  12. ^ Christian Lehmann: Semantic roles. May 7, 2015
  13. ↑ Beatrice Primus: Semantic roles. Winter, Heidelberg 2012, ISBN 978-3-8253-5977-5 .
  14. ↑ Thematic roles (theta roles). Linguistic Institute Ruhr University Bochum, May 10, 2006 ( Memento from June 11, 2007 in the Internet Archive )
  15. ^ Dowty (1991), p. 549
  16. ↑ Dowty 1991, p. 572
This version was added to the list of articles worth reading on August 31, 2005 .