Thematic and Grammatical Hierarchies in Persian Language

There is no one-to-one correspondence between semantic roles and grammatical relations. Based on the concept of implicational hierarchy, some linguists have attempted to identify the relationship between grammatical relations and thematic roles. In this descriptive research, based on the Persian Proposition bank, Semantic Role Corpus in the Persian Language, and Persian Syntactic Dependency Treebank, the semantic hierarchy for different grammatical relations like subject, direct object, indirect object, subject complement and object complement have been explored in the one hand and also the grammatical hierarchy for different thematic roles like agent, experiencer, cause, theme, stimulus, attribute have been considered on the other hand. The syntactic and semantic corpora above consist of approximately 30,000 sentences and about half a million tokens. Observations showed that in the Persian language, among different kinds of semantic roles, the first candidate to be the subject is the agent followed by the theme, experiencer and cause. For the object position, the theme outranks the result which in turn outranks the topic and the recipient. In the direct object position, the most frequent role is the location followed by theme, recipient and destination; and finally, in the subject and object complement position, the attribute is the most frequent role among the other thematic roles in the sentences. On the grammatical hierarchy for thematic roles, subject outranks the object, as the agent and the experiencer; subject outranks the adjunct as the cause; triple preposition outranks the subject as the stimulus; object outranks the subject as the theme, and finally, the subject complement outranks the subject complement as the attribute. Apart from this dominant tendency, this research showed that there are other remarkable correlations between syntactic and semantic arguments; for example, the agent in addition to subject position can be placed in the direct object, indirect object, and even in the adjunct position. The appearance of experiencer in object position and stimulus in subject position is the other remarkable result.



1. Introduction

Grammatical relations do not necessarily correspond to the specific semantic roles. For example, although the subject usually tends to be agent, data-driven studies show that the other thematic roles like theme, experiencer or cause tend to be in the subject position too. In this study, we sought to answer these two questions; what the relationship between the hierarchy of semantic roles and grammatical relations is? And what the thematic and grammatical hierarchies in the Persian language are? This study hypothesizes that Persian native speakers usually place agent, patient, recipient, and attribute respectively into the subject, object, prepositional object, object complement, and subject complement position.



2. Literature Review

Based on the concept of grammatical hierarchy, Fillmore (1968), Jackandoff (1972), Comrie (1981), Givon (1984), Kroft (1990), Dowty 1991, Saeed (2003), and some other linguists have shown the relationship between semantic roles and grammatical relations in different ways. Fillmore (1968) states that not all semantic roles appear equally in the subject position. It seems that there is a tendency for the actor to be the subject than the instrument. Among the instrument and the receptive, the instrument more likely appears in the subject position. Saeed (2003: 155) constructs the subject hierarchy as below:

AGENT>RECIPIENT/BENEFACTIVE>THEME/PATIENT>INSTRUMENT>LOCATION

In the Persian language, some studies have separately concentrated on semantic roles, grammatical relations, or grammatical hierarchies. For example, Farrokhi Rad (2007), Mirzaei & Moloodi (2014), and Alizadeh & Rezghi (2015) have studied thematic roles in the Persian language. Bahrami (2014) and Bahrami (2017) have explored the animacy hierarchy and definiteness hierarchy. Gholipour (2018) has studied some thematic roles in identical syntactic valency structures of Persian verbs.

3. Methodology

To achieve the hierarchy of grammatical relations and semantic roles in Persian language and to examine the hypotheses of this research, we used the Persian proposition Bank (Mirzaei & Moloodi: 2016), Semantic Role Corpus in Persian Language (Mirzaei and Moloudi, 2015), and Persian syntactic dependency treebank (Rasooli et al., 2013). These corpora, which have the same content, consist of approximately 30,000 sentences. The two first were annotated whit the predicate-argument information and the third with syntactic annotations in addition to morpho-syntactic features.

In the present study, in order to investigate the correspondences between the semantic roles of the verbs and grammatical relations, at first, the semantic correlations of grammatical relations including subject, object, the prepositional object, subject complement, and object complement have been considered. Then the syntactic correlations of semantic roles including agent, experiencer, cause, theme, stimulus, and attribute have been explored.



4. Results

As shown in Table 1, findings indicated that among different semantic roles, the first candidate that tends to be the subject is the agent followed by the theme, experiencer and cause. For the object position, the theme outranks the result which in turn outranks the topic and the recipient. In the direct object position, the most frequent role is the location followed by theme, recipient and destination; and finally, in the subject and object complement position, the attribute is the most frequent role among the other thematic roles in the sentences.



Table 1

grammatical relations and equivalent semantic roles




first thematic role & frequency
second thematic role & frequency
third thematic role & frequency
forth thematic role & frequency




Subject
agent

14440
theme

9985
experiencer

3145
cause

1409


Object
theme

1343
result

1093
topic

564
recipient

372


Prepositional object
location

1680
theme

1463
recipient

1084
destination

644


Subject complement
attribute

3875
the other roles

< 100




Object complement
attribute

692
the other roles

< 100







On the grammatical hierarchy for thematic roles, as shown in Table 2, the subject, object, adjunct, and prepositional object respectively tend to be the agent. This hierarchy is different for the cause, in that subject tends to be in the first position and then adjunct, triple preposition, and subject complement. For the experiencer, stimulus, theme and attribute, the grammatical hierarchy respectively is as below:

experiencer: subject> object> triple preposition> subject complement

stimulus: triple preposition> subject> complement clause> object

theme: object > subject> triple preposition> complement clause

attribute: subject complement> object complement> subject> complement clause



Table 2

semantic roles and equivalent grammatical relations




agent
subject

14440
object

214
adjunct

221
prepositional object

62


cause
subject

1409
adjunct

99
* triple preposition

44
subject complement

27


experiencer


subject

3145
object

348
triple preposition

92
subject complement

29


stimulus
triple preposition

615
subject

218
complement clause

110
Object

63


theme
object

13430
subject

9985
triple preposition

6006
complement clause

3205


attribute
subject complement

3875
object complement

692
subject

152
complement clause

74



^* object complement, prepositional complement of noun and prepositional complement of adjective are introduced as the triple preposition.



5. Conclusion

According to Dowty 1991, the best theory for describing the argument selection will be one that involves just two roles named Proto-Agent and Proto-Patient. Proto-Agent includes agent, causer, experiencer as the roles that they have volitional involvement in the event or state, causes an event or change of state in another participant, moves relative to the position of another participant, internally causes the event or state.

The concept of Proto-Patient also includes that role which undergoes a change of state, is causally affected by another participant, is stationary relative to the movement of another participant, or is subject of an external cause

This study showed that the subject as a relational meaning in the Persian language is equal to the concept of Proto-Agent as a thematic role. The object also is nearly equal to the concept of Proto-Patient as a thematic role.

Additionally in this study, the semantic hierarchy for different grammatical relations like subject, direct object, indirect object, subject complement, and object complement have been explored on the one hand and also the syntactic hierarchy for different thematic roles like agent, experiencer, cause, theme, stimulus, attribute have been considered on the other hand.