1.Introduction
Two kinds of verb formation are found in Hindi[1]: simple verbs and complex verbs. Examples of the former are calnaa(walk) and caRhnaa (climb) etc and the examples of the latter are aarambh karnaa (start) bhang karnaa (end) etc. Figure 1 shows the types of Hindi verbs in a tree diagram.
Figure1. Types of Hindi Verbs
Complex verbs, also known as Complex Predicates (CP), are formed in two ways in Hindi. The first type is known as Conjunct Verbs. Here, the first member is either a noun or an adjective. Examples 1 and 2 are the instances of this process. The conjunct verbs are marked in boldface.
1. usne sabhaa aarambh kii
Pron-erg N N V
(s)he-ergconvention start do.
‘(S)he started the convention’
2. usne bacce ko shaant kiya
Pron-erg Nacc Adj V
(s)he-erg child acc pacify do
‘(S)he pacified the baby’
In Example 1 we see a noun+verb sequence, and in 2 that of an adjective+verb.
The second type of complex verb construction found in Hindi is a verb+verb combination. This type of verb is known as a Compound Verb (CV). The first or the main verb is in its stem or some other non-finite form while the second verb is in a morphologically finite form bearing the relevant inflections such as person, number, gender, tense, aspect and modality. In example 3 and 4 CVs are marked in boldface.
3. usne saaraa khaanaa khaa lia
Pron-erg adj N V1 V2
(S)He -erg all food eat take
‘He ate all the food’
4. vah ghar calaa gaya
Pron NV1V2
he home wentpast
‘He went home’.
Here, we undertake a detailed study of the CPs in Hindi. We have chosen a special type of conjunct verb, i.e., Noun+Verb (henceforth N+V) and compound verb (henceforth CV) as the domain of our investigation.
The paper is divided into following sections. Section 2 states the motivation of the study, section 3 deals with the N+V combinations in different languages and then the tests to identify them in Hindi. Section 4 discusses the compound verbs in Hindi, section 5 talks about Verb Knowledge Base and the way to deal with the CPs therein. Section 6 concludes the paper and gives future direction.
2. Motivation
The motivation of this work comes from the building of the Hindi Wordnet (HWN), a lexical database for Hindi [1]. Here, a concept is denoted by a synonym set (synset). Thus, {ghar, kamraa} in one synset gives the sense of room and {ghar, makaan} togetherrepresents the concept of house. These synsets are then linked with each other through the well known semantic relationshypernymy, hyponymy, meronymy, holonymy, antonymy, entailment, troponymy and form a net like structure.
In the process of building the HWN we have to deal with CPs and the question of how they are to be stored in the database. Since, there are large numbers of CPs in Hindi we need to have a uniform way to deal with these cases.
It has been found that the nouns denoting action always come with a verb in Hindi. It is not possible to construct a sentence with those nouns without combining them with a verb. For example, advice in English has two different constructions:
5. He advised me on road safety.
6. He gave me advice on road safety.
give advice in English is similar to Hindi paraamarsh denaa. In Hindi, it is possible to have a structure like 6 but not 5. Thus, the question arises whether to store paraamarsh or paraamarsh denaa in HWN. There are two possible ways of solving this:
- to keep paraamarsh in the lexicon and then to link it with the verb with which it co-occurs
This solution is the simplest and syntactically too in terms of the argument structure of the associated verb and the agreement properties there is no problem. But, there is a semantic mismatch. Consider chalaang maarnaa (dive). There are two senses of maarnaa in Hindi: beat and kill. None of these senses match with the maarnaa in chalaang maarnaa (dive). This approach appears to miss one property of these constructions.
- to keep paraamarsh denaaas a single entry in the lexicon and then to specify its syntactic and semantic features
The problem with this analysis is this we cannot keep all such constructions readily in the lexicon. In many instances the noun is really an object and there is no need to store it as a lexical unit along with the verb. So, it is necessary to separate the conjunct verbs from other N+V sequences.
The V+V constructions are also difficult to deal with and describe since there are other homotactic sequences found in the language. Consider,
7. raam kitaab parh rahaa hai
ram book V1 V2 is
‘Ram is reading the book’
8. raam ne kitaab parh daalii
ram erg book V1 V2
‘Ram read the book’ [finish]
rahaa‘is’ in 7 marks the progressive aspect where as daalaa ‘put’in 8 says something more than completeness, i.e., about the attitude (somehow finished). Thus, the verb in 7 will not be stored in the lexicon where as 8 may be. This is due to the shade of meaning that daalnaa ‘put’ adds.
3. Related work
A large number of Hindi verbs are N+V combinations. This phenomenon is seen cross linguistically and is treated as Noun Incorporation (NI) in the literature. The major question to answer with this type of verb construction concerns the status of the noun, i.e., whether the noun is incorporated into the verb complex or if the noun is an overt argument of the verb. Consider the two combinations caae lenaa‘to take tea’and jamhaaii lenaa ‘to yawn’. In the former case caae ‘tea’ is an overt object of the verb whereas in the latter jamhaaii ‘yawn’ is incorporated.
Mohanan (1995) was the first to explicitly claim that Hindi has incorporation [2]. In doing so, she drew on morpho-syntactic as well as semantic evidence. Peter Hook dealt with the incorporation in Hindi and Urdu [3]. According to him one criterion to distinguish IN from unincorporated nouns in these languages is the gender agreement. Hook argues that if a noun is incorporated in a verb then it strips of its identity as a direct object noun and is "incorporated" into the verb as an integral part of it. As a part of the verb it cannot be a direct object and cannot affect the verb's inflection. Hook’s claim regarding the agreement pattern of incorporated noun in Hindi is not visible uniformly. There are a large number of counter examples where this pattern is not seen. Dayal (2003) also notices this fact and concludes that incorporated nouns are syntactically visible and can function like regular arguments- atleast for the purpose of agreement [5]. Mohanan too has identified this phenomenon and remarked “In complex predicates, if the nominal predicate is logically transitive, and its second participant is a logical object, this object is selected as the grammatical object of the complex predicate.”[6]
3.NI in Hindi
Based on the complexities exhibited by incorporated nouns, following tests are used to identify an incorporated noun from the case dropped object nouns. The tests used for accomplishing this task include:
- Adding the accusative Case Marker
- Constituency Test
- Movement Test
ii. Conjunct Response test
iii. Coordination test
- Adding of the Adjectives
3.1. Adding the accusative Case Marker
To identify the syntactic status of the object noun, accusative (acc) case is assigned to it. The aim ofthis test is to see whether the sentence is acceptable after overtly case marking the object with the accusative case. The object noun will allow accusative marking which an incorporated noun will not or ought not to.
Primary objects in Hindi are either acc with the marker ko or are marked nominative (nom), i.e., bare nominal. A widely accepted generalization with regard to objects in Hindi is that the canonical case for animate objects is acc and for inanimate objects it is nom. This is a phenomenon found in many South Asian languages [7Kac]. The inanimate object can bear acc case if it is definite.
In Hindi, a large number of conjunct verbs are found with inanimate nouns. In this test these inanimate nouns are marked with a definite marker to make it possible for them to be assigned acc case. These were given to native speakers to judge their acceptability or correctness. Consider the following examples
9. a. raam ne caae lii
ram erg tea take
‘Ram took tea’
b. raam ne us caae ko liyaa jo khulii thii
ram erg that tea acc take which open was
‘Ram took that tea which was not covered ’
10. a. raam ne jamhaaii lii
ram erg yawn take
‘Ram yawned’
b.* raam ne us jamhaaii ko liyaa
ram erg that yawn acc take
‘Ram took that yawn’
11. a. vah mere kaam me rucii letii hai
(s)he my work in interest take is
‘She takes interest in my work’
b. *usne mere kaam me us rucii ko lii jo
(s)he my work in that interest acc take which....
‘(s)he takes that interest in my work which ....’
In 9 the direct object of the verb lenaa (take)is caae (tea). 9a is without an acc marker and 9b is marked acc and both of them are acceptable. In 10 and 11jamhaaii (yawn) and ruci (interest) appears in the direct object position. 10a and 11a are without the acc marker and acceptable but 10b and 11b are marked unacceptable when they bear the acc marker.
3.2 Constituency Test
Three tests are discussed here. The first test is the movement test. It tries to see whether the incorporated noun can move from its preverbal position, i.e., the canonical position. Second, the constituent response test is applied to see whether a given N+V combination behaves as a constituent or not. Third, the coordination test examines whether an incorporated noun can also be coordinated or conjoined with other nouns in the direct object position.
i. Movement of the Noun:This test is applied to see whether the incorporated nouns in Hindi can be moved from its canonical position. If it is possible to move objects in Hindi to non-canonical (for example, postverbal) positions is it also possible with the incorporated nouns? If the N+V combination forms a single constituent then the incorporated noun will resist movement from its position near the verb.
Two types of sentences are formed with the same N+V combination. In the first sentence the noun is adjacent to the verb and in the second sentence it is moved away from the verb. These two constructions were judged by the native speakers. Consider the following examples. The noun that is moved is in boldface.
12a.usne subaha uthkar caae lii
Pron morning wake tea take past.
having woken up in the morning (s)he took tea’
b. caae usne subaha uthkar lii
tea Pron morning wake take past.
‘having woken up in the morning (s)he took tea’
13akahaani ne dukhaant ruup liaa
story erg tragic shape take past.
‘the story took a form of tragic ending’
b.*ruup kahaanii ne dukhaant liaa
shape story erg tragic take past.
‘the story took a form of tragic ending’
14a.usne pratiyogita me bhaag liaa
Pron competition in part take.
‘(S)he took part in the competition’.
b.*bhaag usne pratiyogita me liaa
part Pron competition in take.
‘(S)he took part in the competition’.
All the b counter parts in the above examples are instances where the noun has been moved from its canonical position. In 12b caae ‘tea’ is the actual object of the verb lenaa ‘take’. Both 12a and b are marked acceptable.13a and 14a are declared acceptable by the speakers but 13b and 14b in which the nouns are moved from their canonical position are marked unacceptable. This argues for the fact that even though Hindi has a relatively free word order it resists movement when it comes to NI. Two or more elements that form a constituent cannot be moved independent of each other. Thus, ruup lenaa‘take a shape’ andbhaag lenaa‘take part’are considered as cases of NI.
ii. Constituent Response Test:This test is applied to see whether a given N+V combination behaves as one constituent or not. Consider the following N+V sequences jamhaaii lenaa‘yawn take’, chalaang maarnaa ‘jump beat’.
15. raam ne jamhaaii lii
ram erg yawn take
‘Ram yawned’
16. raam ne chalaang marii
ram erg jump beat
‘Ram dived’
In order to get a meaningful response about the action denoted by 15 and 16 we framed questions in the following manner:
17a. usne kyaa kiyaa (What did he do?)
17b. and not What did he take?
18a. usne kyaa kiyaa (What did he do?)
18b. and not usne kyaa maaraa (What did he beat?)
Now consider, 19
19. vah bazaar se phal laayaa hai
he market from fruit brought is
‘he brought fruits from the market’
To get a meaningful response about the action in 19 the question may be formed in the following ways:
20a. vah kya laayaa (What did he bring?)
20b. or usne kyaa kiyaa(What did he do?)
Questions identify or replace single constituents and the fact that one must question N+V suggests that they form a single constituent. Not so in the case of phal laanaa‘bring fruit’
iii. Coordination Test:This test is to verify if the noun that is assumed to be an incorporated noun can be conjoined with another noun, i.e., does it form an independent constituent. Coordination is possible only between identical syntactic constituents.
21. log caae aur namkin le rahe the
people tea and snack take prog past
‘people were taking tea and snacks’
22. *log niind aur jamhaaii le rahe the
people sleep and yawn take prog past
‘people were taking sleep and yawn’
23. *usne bacce ko ksamaa aur vidaa kii
(S)he children acc forgive and farewell do
‘(S)he forgave and bade farewell to the children’
The above test shows that the N+V combination has the status of one lexical or syntactic unit.
3.3 Adding of the Adjectives
True arguments may be modified by adjectives, determiners, numerals, sentences etc. The Noun Phrase (NP) can have very rich structure. Incorporated nouns do not permit such modifiers, i.e., they are bare nouns not NPs.
24. usne jor se dhakka maaraa
Pron ADV push beat
(S)he pushed heavily.
25. usne merii bahut madad kii
Pron me ADV help do
‘(S)he helped me a lot’
The modifier jor se ‘heavily’ in 24 modifies the whole sequence dhakkaa maarnaa ‘push-beat’ and in 25bahut ‘lot’ modifies madad karnaa ‘help-do’ and not karnaa ‘do’.
3.4. Features of NI
NI has certain semantic and syntactic properties as we saw in the previous sections. These properties of NI convince us to deal with them as a result of lexical compounding. These are listed here.
I. A possible trigger for NI is either the noun or the verb that is morphologically defective and cannot appear as independent word. In Eskimo languages there are certain verbs that are simply subcategorized to attach to a Noun. Polysynthetic languages have defective determiners. Baker (1996)[8] calls it defective because the determiners cannot govern the nouns and so the latter have to adjoin to the verb to get case. Similarly, in Hindi some nouns only appear in the N+V collocations.
II.As noted by Mithun (1984) [9] and others, incorporated inanimates tend to be more common than incorporated animates. The same is found in Hindi. There are rare instances of animate incorporation in the language.
III. It is noted that there are certain idiosyncratic features of the noun that is adjoined to the verb in a possible N+V combination. Thus, jamhaaii ‘yawn’ will be attached with the verb lenaa‘take’ and not with any other verb. The verb with which a noun attaches is arbitrarily selected and is possibly determined lexically. This is a strong reason for the debate that NI is lexical in Hindi and not syntactic.
IV. The meanings of incorporated structures are often non-compositional. jamhaaii lenaarefers to ‘yawn’ and not physically taking anything like caae lenaa ‘take tea’. This feature argues for the semantic lightness of the verb. The verbs to which the noun is adjoined in Hindi is semantically null, i.e, it has nothing to do with its original meaning seen in isolated contexts.
Here, we want to make a distinction between the syntactic incorporation and lexical incorporation. There are cases of syntactic incorporation in Hindi as well.Consider, 26 taken from Dayal (2003) [6].
26. anu bacchaa/bacche ko sambhaal rahii
Anu child child-acc is-looking after
‘Anu is looking after children(one or more)/the child’
We will not deal with these cases here but surely we make a difference between syntactic and lexical incorporation in Hindi. We will call the lexical noun incorporation as lexical compounding and the resulted verbs as conjunct verbs.
4. Compound Verb
There are many V+V sequences in Hindi. This is a phenomenon seen across the South-Asian languages. Butt [10] makes a difference between permissive, instructive and aspectual complex predicates and takes the stand that complex predicates formation in Urdu takes place at argument structure. Hook (1991) [11] placed the second verb, i.e., the light verb in an aspectual complex predicate on a par with auxiliaries in the language. Bashir [12] identified the compound verbs of Kalasha and gave a semantic analysis of them based on prepared and unprepared mind. Fedson [13] has dealt with the semantic structure of the complex predicates in Tamil and classified them into various groups like stative, perfective, indication of status, aspectual etc. Pandharipande [14] in her article on ‘Serial Verb Construction in Marathi’ pointed out that V1 and V2 are paired on the basis of their semantic compatibility, which is subject to syntactic constraints. Paul [15] in her thesis has dealt with the CVs of Bangla exclusively.