They are kind of/sort of similar: a parallel corpus-based analysis of English KIND OF and SORT OF and their Lithuanian correspondences

The present paper reports on the English type nouns kind of and sort of and their Lithuanian correspondences in a contrastive perspective. This paper aims to describe the quantitative and qualitative distribution of the English kind of andsort of, to determine their translational correspondences in Lithuanian as well as to reveal how Lithuanian correspondences correlate with the functions (textual and interpersonal) that kind of and sort of perform in original and translated fiction texts. The research method is a quantitative and qualitative contrastive analysis based on data extracted from the self-compiled bidirectional corpus ParaCorpEN→LT→EN comprising fiction texts. The results show that kind of and sort of are prone to be used NP-internally; however, even in this construction they can feature as DMs. Kind of and sort of function as unambiguous DMs when they completely lose their nominality, i.e. are used NP-externally. The functional and semantic potential of the type nouns is fully reflected by their TCs. Very rarelykind of and sort of denoting a type are translated congruently into a Lithuanian type noun; they usually correspond to demonstrative pronouns. As discourse markers, kind of and sort of are realised by different Lithuanian correspondences which may help establish the common ground between the speaker and the hearer or refer to the previous context, may indicate epistemic imprecision, approximation or downtone a proposition. The high number of zero correspondence shows that the Lithuanian type nouns have not advanced on the grammaticalization path the way the English type nouns have and due to the multifunctionalilty, non-propositionality and context-dependence there is no one-to-one correspondence of the markers under scrutiny.


Introduction
In Present-day English, there are three main nouns expressing the general meaning of 'type': type of, kind of and sort of. Traditionally, the type noun constructions are defined as nominal expressions dealing with (sub)categorization (Brems 2011, 2), or in other words, as nominal phrases used to ascribe a lexical item to a particular group or type sharing exact or similar features, i.e. dealing with taxonomic relationships, as in the example "a (special) sort of rose" (Traugott 2008, 226). The definition, however, applies only to the lexical meaning of these nouns. It must be admitted that, throughout the years, kind of and sort of, via the grammaticalization process, have gone through the pragmatic-semantic process of subjectification, which, according to Traugott (1995, 32), is: a gradient phenomenon, whereby forms and constructions that at first express primarily concrete, lexical, and objective meanings come through repeated use in local syntactic contexts to serve increasingly abstract, pragmatic, interpersonal, and speaker-based functions.
Thus, the term type noun in this paper encompasses not only the primary lexical meaning, but is also used to account for other pragmatically charged uses of the items in question. 3 terms of (inter)subjectivity (Traugott 2010), and to investigate how Lithuanian correspondences correlate with the function that kind of and sort of perform in original and translated fiction texts: 2 (1) EN-orig: Usually Hatsumomo liked to drink a special kind of sake called amakuchiwhich was very light and sweet.

From type nouns to discourse markers
Scholars seem to disagree on the exact number of the constructions the type nouns can appear in. However, traditionally there is a unanimous agreement that kind of and sort of may be used noun-phrase (NP) internally (i.e. nominal use) or noun-phrase externally, i.e. they may be exploited to modify adjectives, verbs and adverbs, prepositional phrases or function as completely independent units (i.e. free adjuncts) modifying whole propositions.
When kind of and sort of are used NP-internally, the first construction they appear in is the binominal one (Denison 2005;Traugott 2008), also known as lexical-head use (Davidse et al. 2008, 147), in which the type noun has a generic reference, is used in hyponymy statements, and denotes a specific subclass or, as in example (1) above and examples (4) and (5) Here the type noun is the head and the of-phrase is its modifier (Brems, Davidse 2010). 3 The binominal kind of and sort of may "either introduce a (potential) discourse topic or refer back to an existing discourse topic" (Keizer 2007, 155-156). The type noun can be both singular and plural in number and all determiners are freely available in the pre-modification field of the phrase (Traugott 2008, 234;Denison 2005). They are not restricted to affirmative environments but can also appear in questions.
The second construction of the nominal kind of and sort of is the qualifying construction (Denison 2005), or degree modifier (Traugott 2008, 226-227), which "is the only one that is unanimously recognized besides the head use" (Davidse et al. 2008, 156). The qualifying construction, in contrast to the binominal construction, does not carry a generic reference; here the type noun functions as a discourse marker and hedge and "conveys that the description is only approximate" (Davidse et al. 2008, 157) or, according to Traugott (2008, 227), "[p]ragmatically such phrases cast doubt on the accuracy of the description", e.g.: (6) EN-orig: "I think that was a sort of joke," Mameha said.
In this construction the type nouns are modifiers having the meaning of more or less or somewhat and "NP2 is perspectivized as in the foreground" (Traugott (2008, 226-27). Here the type noun normally occurs in its singular form and is preceded by an indefinite determiner or no determiner at all.
(8) EN-orig: I sort of like you.
(9) EN-orig: She sort of accidentally made it sound like I was trying to kill myself or something.
(10) EN-orig: That's kind of how they got into this mess in the first place.
The non-NP internal construction fulfils numerous "discourse functions such as self-repair, hedging strong opinions, establishing common ground between interlocutors, politeness, and so forth" (Dehé, Stathi 2016, 918). (12) EN-orig: Unfortunately it was the kind of job that only keeps hands busy.
construction is quite special, because, due to lexicalization, the type nouns in plural appear in fixed phrases with all and carry the meaning of 'many/much': (13) EN-orig: Fache had posited all kinds of explanations tonight to explain Sophie's odd behavior <...>.
In addition to quantifying uses with all, negative totality quantifiers as no or (not) any also fall under this category.
Finally, the NP-internal kind of and sort of can function as modifiers. The main characteristic of the modifier use of the type nouns is the use of an adjective or a noun that is more related to N2 than the type noun itself, for instance: It's a cool quirky kind of song. (Davidse et al. 2008, 147) Here kind of and sort of have a metalinguistic value, indicating that the lexical items preceding them "have to be interpreted as ad hoc, often very creative, classifiers" (Brems, Davidse 2010, 188). They also act pragmatically, signalling that "the description is only approximative" (ibid.).
In this paper a distinction is made between the propositional use of kind of and sort of expressing 'a type of' and their discourse marker use. All the cases of sort of and kind of where they do not mean 'a type of' are treated as having a discourse qualifying value, i.e. functioning as discourse markers (Aijmer 2002, 178). Discourse markers are perceived as having "by definition a discourse function, which entails indexing the utterances to the surrounding discourse, both in terms of structuring the ongoing discourse and in terms of signalling to the addressee how he/she should interpret the speaker's stance" (Wichmann et al. 2010, 107).

Data and Methods
The present paper deals with a quantitative and qualitative contrastive corpus-based analysis drawing on the data extracted from a self-compiled bidirectional parallel corpus -  (Biber et al. 1999, 869). Secondly, such a corpus design is advantageous as it offers different directions of linguistic comparison and can be used both as a parallel corpus and a comparable corpus (Johansson 2007, 11). The corpus consists of about 5 million running tokens (see Table 1). In order to generate concordance lines with the type nouns kind of and sort of as well as their translations, the multilingual concordancer ParaConc (Barlow 1995) was used. The corpus output files in the plain text format were imported into an Excel spreadsheet for further analysis. The search was bidirectional: first, kind of and sort of were searched as nodes in the EN-orig  LT-trans direction; second, they were entered as nodes in the EN-trans  LT-orig search direction. Table 2 shows the raw frequency of the items under investigation: It is important to note that the cases where kind of was used as an adjective in the structure kind of somebody to + verb were discarded from the analysis as a first step before any calculation. Based on the raw frequency count, it was decided to take 200 randomized hits from the EN-orig texts and 100 randomized hits from the EN-trans texts of each kind of and sort of from both sub-corpora for further functional analysis. All in all, there were 600 cases of the use of kind of and sort of analysed. The randomization procedure was performed using the RAND function available in the Microsoft Excel spreadsheet -each occurrence of the node in question was assigned with a random number from 0 to 1. Then the numbers were ranked from the lowest to the highest and the first 200 and 100 hits were selected for the analysis.
Though the data extraction and randomization were automatic, the qualitative analysis of the concordances was carried out manually. Firstly, the uses of the type nouns were divided into two groups depending on whether they function NP-internally or NP-externally. The NPinternal cases were grouped into the ones expressing 'type of' and the ones functioning as DMs.
All of the non-internal cases were categorised according to the item they modify or act as a stand-alone adjunct. Finally, their translational correspondences as well as different functions were considered.
On the basis of the empirical data derived from a parallel corpus we can establish translation paradigms (Johansson 2007, 23) or semantic mirrors (Dyvik 2004, 311) which contain translational correspondences. First, we can divide TCs according to the direction of translation. Then having analysed the TCs of a linguistic unit in question in terms of expression, we can single out zero and overt correspondences. The latter fall into two groups: congruent (direct) and non-congruent (divergent) correspondences. Congruent correspondence is the case when a linguistic item is translated into another language keeping the same part of speech (form and meaning); there is no system shift (Johansson, 2007, 24), for example: (15) EN-orig: There was this new kind of cancer that was getting young men.
Divergent correspondences, on the other hand, are more interesting for contrastive studies, as they "serve[…] as a means of uncovering differences where they may be unexpected" (Johansson 2007, 25). These are the cases when, from the perspective of meaning, a greater or lesser part of meaning of the SL text is compensated in the TL by various means, usually not listed in dictionaries as traditional prototypical equivalents (Usonienė 2006, 101), for instance: (16) EN-orig: He was determined to make this some kind of double date, apparently.
Zero correspondences are cases where there is no clearly identifiable semantic counterpart in the TL or SL text (Johansson 2007, 26). Sometimes translators decide to omit a word or phrase in translation (omission), as in (17), or add some new shades of meaning in TL texts that were absent in the original (addition), as in (18) The Lithuanian TCs of kind of and sort of were categorised according to the three abovementioned types.

Results
This section presents the quantitative and qualitative findings of the analysis. It starts with the analysis of the quantitative distribution of kind of and sort of (Section 4.1). The subsequent sub-sections further elaborate on both quantitative and qualitative findings.

Quantitative distribution of kind of and sort of
The normalized frequency (per 10,000 words) of kind of and sort of in both the original and translated English texts is presented in Table 3. The table gives all the relevant occurrences of kind of and sort of in the corpus. Though the frequencies seem to be more or less levelled out across the different sub-corpora, the only discrepancy that strikes the eye is the overuse of kind of in the EN-trans texts in comparison with sort of. The reason behind this might be the fact that the majority of the translators in ParaCorp EN→LT→EN are American English-based bilinguals, and some studies proved that kind of is much more frequent in American English than in British English (see Biber et al. 1999, 870;Quirk et al. 1985, 598).  Table 4 shows the parts of speech the type nouns have scope over.  As has already been mentioned, the prevailing use of kind of and sort of in the EN-orig and EN-trans texts was NP-internal. Table 5 and Table 6    As is obvious from the tables above, sort of conveyed pragmatic meanings more often than kind of in both sub-corpora. Its use as a discourse marker is almost twice more frequent than that of kind of (62 % vs. 37 % in the EN-orig texts and 60 % vs. 32 % in the EN-trans texts). The findings are in line with Aijmer (2002), who noted that "kind of is not used to the same extent as sort of at least on the British scene" (Aijmer 2002, 207). In her data kind of was used as a clear discourse marker only in 17.8 %. The reason for this is hard to find. One of the tentative observations might be the findings presented by Brems and Davidse (2010, 193), whose diachronic analysis shows a constant growth of the use of kind of in the binominal construction from the year of 1780 to the present day.

The NP-internal kind of and sort of as 'a type of'
The whole range of TCs of kind of and sort of denoting a type of something is presented in

LT-trans: Tokiai vaikinų kategorijai jis priklausė.
The majority of translational correspondences were divergent and included demonstrative pronouns such as tie/šie 'those/these' and toks/šitoks 'such', which are generally used to refer back to some information mentioned before or something that will be presented later: (20) LT-orig: Blogiausia, kad tokios neapykantos neišperka meilė.

EN-trans:
The worst of it is that love doesn't compensate for that kind of hatred.

LT-trans: Aš visada maniau, kad Veruka -tai tokia karpa, kuri atsiranda ant kojos pado!
The results show that wh-questions with kind of and sort of quite frequently denoted a type and their Lithuanian TCs mostly encompass koks or kas 'what/who', for example:

EN-trans: What kind of humility can there be these days when men and women have equal rights!
The findings seem to coincide with the ones obtained by Janebová and Martinková (2017), as their study unveiled that in such contexts the Czech equivalents of the type nouns "were phoric expressions (demonstrative pronouns, phoric expressions meaning "similar" or "same", and adjectives and adverbs referring to the common ground such as "typical", "precise/ly", or "definite"" (Janebová, Martinková 2017, 205).

Kind of and sort of as discourse markers
The TCs of kind of and sort of functioning as discourse markers in the EN-orig texts are displayed in Table 8 below. The cases of their non-internal use (i.e. the adverbial constructions) have been added to the total. The translational profile of kind of and sort of as DMs in the EN-trans texts yielded somewhat similar results (see Table 9); however, the range of TCs is much narrower. The three cases of their non-internal use (i.e. the adverbial constructions) have been added to the total, too. The reason for a less divergent translational profile might be the fact that Lithuanian does not have fully grammaticalized equivalents for the English type nouns; other means of expression of approximation or mitigation may be in use, which did not fall under the scope of the paper. As seen from the tables, in both sub-corpora the pronominal strategy of TCs predominates. The (24) EN-orig: <...> I didnʼ t feel terror so much as a kind of vague queasiness.
In (23) and (24) the speaker is looking for the appropriate word to describe a range of obstacles or to name a feeling at the same time conveying his/her subjective attitude.
The adjectival TCs of kind of and sort of show that the speaker keeps the description at a fuzzy level, too. The most frequent TCs are savotiškas ʻpeculiarʼ and panašus ʻsimilarʼ , which express non-typicality, vagueness and imprecision: Interestingly, the analysis revealed that the phrase kind/sort of thing was quite common in the corpus. Beeching (2016, 158) calls this lexicalised chunk as "the general extender" and maintains that it "also serves, retroactively, to hedge or downplay any perceived conceitedness and thus serves as a face-threat mitigator". In my dataset it usually appeared as part of a sentence (27) or as a stand-alone unit (28). The TCs included phrases such as panašiai 'similarly', panašūs dalykai 'similar things' kažkas tokio 'something like' and maždaug taip 'approximately', which is also a means to convey vagueness and imprecision: (27) EN-orig: Not in a bad way, just to calm someone down, that kind of thing.
The greatest variety of TCs was attested in the intersubjective domain: tartum/tarsi/lyg (ir)/it 'as if', kažkaip 'somewhat', galima sakyti/pavadinti 'it can be said', kaip ir 'somehow', atrodo 'it seems', etc. The most frequent TCs were tartum/tarsi/lyg (ir)/it 'as if' which seem to be the closest equivalent to kind of/sort of as they may equally show inexactness; serve as discourse markers that weaken the illocutionary force and flag the speaker's uncertainty, as in (29)-(31): (29) EN-orig: "He never really got to the point." "You looked kind of mad," she fished.

EN-trans:
It has its own sort of self-rule.

LT-trans: Mano senelis Efraimas Blekas buvo tarsi paskutinis mūsų vadas <...>.
In the examples above the speaker tries to make his/her opinion sound safely vague and thus softens the effect of his/her assertion. Example (29) is also a case of the speaker's unwillingness to impose on the hearer, by using lyg ir 'as if' he/she tries to save the interlocutor's face.
Finally, there are TCs such as beveik ʻalmostʼ , gana ʻsomewhatʼ , and kažkaip ʻsomehowʼ , which downtone the proposition, but at the same time explicate the meaning of imprecision and approximation, for example: (32) EN-orig: Iʼ ve moved out. Sort of, anyway.
The two units in question may also signal a close relationship between the speakers. They "make inferences or assumptions about each other's knowledge" (Aijmer 2002, 202), for example: (34) EN-orig: "He works at Hogwarts." "Oh," said the boy, "I've heard of him. He's a sort of servant, isn't he?" "He's the gamekeeper," said Harry.
In example (34) the boy makes an inference that a person is a servant based on the hearsay and general knowledge and at the same time seeks a confirmation from his interlocutor.
The Lithuanian correspondence lyg ir 'as if' mitigates the question. In such contexts the Lithuanian correspondences also include galima sakyti 'it can be said': (35) EN-orig: "If we are dealing with newborns, he'll be helpful." "Jasper? Why?" Edward smiled darkly.
Even though Edward does not seem to have a clue why Jasper may be useful in dealing with newborns, the initiator of the dialogue appeals to common knowledge that he wants to share.
The implication is that in general Jasper is known to be a good specialist when it comes to dealing with young vampires. Here galima sakyti 'it can be said' could be substituted by žinok 'you know'.
The varied list of TCs may prove that Lithuanian does not have a single equivalent for the two multifunctional units discussed herein. Also, it is sometimes difficult to disentangle different functions performed by kind of and sort of since they overlap (Beeching 2016, 160).

Zero correspondence
Overall, zero correspondence is a unifying feature of DMs in cross-linguistic studies (Johansson 2007, 26). Due to their extreme multifunctionality, non-propositionality, contextdependence and non-referential (textual and interpersonal) function, they exhibit a wide array of different TCs and their exact cross-linguistic equivalent is hard to find. The cases of zero correspondence of kind of and sort of as DMs amount to 48 % in the LT-trans texts. Even though they had no particular source in the LT-orig texts, the two DMs were also inserted in the English translations quite frequently (41 %). The propositional type noun usage in the present study also showed a relatively high percentage of zero correspondence (24 %). Similar findings are presented in Janebová and Martinková (2017, 188): even more than a half of kind of and sort of uses have no Czech correspondences in their dataset.
One of the factors determining zero correspondence of the type nouns might be the redundancy factor and the language norm, e.g.: (36) EN-orig: His lips pulled back over his teeth and his eyes shone with an odd light -a wild, fierce kind of hope.

viltis.
Example (36) is typical of the pattern modifier + kind of/sort of + NP, which is especially inclined to have kind of/sort of omitted in translation, seemingly because Lithuanian does not require an additional mediator between the modifiers and the modified head.
In addition, the language norm might also be an explanation why kind of and sort of are inserted into the English translation; it seems that it is more natural to resort to mitigation strategies in English: Kind of in (38) is used to mitigate face threats. By opting for kind of obvious the speaker has an intention to save his/her or the interlocutor's face. Kind of here softens a strongly voiced opinion, whereas, on the contrary, visiškai ʻcompletelyʼ shows sheer confidence and marks a of the already emotionally charged word kankynė 'torture'.

Conclusion
The present paper reports on the multifunctional English type nouns kind of and sort of and their Lithuanian correspondences in a contrastive perspective. There have been a lot of monolingual studies that analysed the English type nouns in terms of their structure or function in different perspectives, but contrastive corpus-based studies comparing two or more languages are rather scarce. This paper aims to describe the quantitative and qualitative distribution of the English kind of and sort of, to determine their translational correspondences in Lithuanian, as well as to reveal their functional diversity. The research method is a quantitative and qualitative contrastive analysis based on data extracted from the self-compiled bidirectional corpus ParaCorp EN→LT→EN comprising fiction texts.
The results show kind of and sort of are prone to be used NP-internally; however, even in this construction they can feature as DMs. Kind of and sort of function as unambiguous DMs when they completely lose their nominality, i.e. are used NP-externally. Sort of conveyed pragmatic meanings more often than kind of in both sub-corpora, whereas kind of was more frequently used propositionally to indicate a type.
The functional and semantic potential of the type nouns is fully reflected by their TCs.
Very rarely kind of and sort of are translated congruently into a type noun (rūšis 'kind', kategorija 'category' and tipas 'type'); even though they denote a type and are used propositionally, they usually correspond to demonstrative pronouns such as tie/šie 'those/these' and toks/šitoks 'such' in Lithuanian.
The translational profile has revealed that certain functions of kind of and sort of as DMs, such as textual and interpersonal, may be realised by different Lithuanian TCs which may help establish the common ground between the speaker and the hearer or refer to the previous context, may indicate epistemic imprecision, approximation or may downtone a proposition (cf. Janebová, Martinková 2017, 209). The most common TCs included kažkas/kažkoks 'something like', kažin koks/kokia 'somewhat', tartum/tarsi/lyg (ir)/it 'as if', beveik ʻalmostʼ , gana ʻsomewhatʼ , kažkaip ʻsomehow', etc. Finally, the same TCs may appear in both functions, which may be indicative of the functional overlap.
The high number of zero correspondence has also revealed some differences between English and Lithuanian. First of all, it is obvious that due to the multifunctional nature and context-dependence of kind of and sort of their exact cross-linguistic equivalents are hard to find. Second, it signals that the Lithuanian type nouns have not advanced on the grammaticalization path the way the English type nouns have and there is no one-to-one correspondence of the forms. Moreover, it seems that the two languages do not put the same emphasis on the mitigation and politeness strategies. Discourse markers are a feature of vague language; their precise meaning is difficult to pin down, so they disappear in translation very frequently.
As this particular research was based on a quite limited number of fiction texts, the analysis of larger corpora encompassing different registers, for example, spoken or academic, would provide more interesting and insightful results. Further research could also include the plural forms kinds of and sorts of. Moreover, the most frequent Lithuanian TCs of kind of and sort of toks/šitoks 'such' and kažkas/kažkoks 'something like' should not escape further analysis.