2.1.1. Adjective-Adverbs

In the case of adjective‑adverbs (also called bare adverbs, adverbialized adjectives or short adverbs), a word belonging to the word‑class of the adjective takes adverbial functions (for the theoretical background, see Hummel 2017), such as the modification of the verb (manner adverb) as shown in the following examples: Spanish claro in hablar claro 'to speak clearly', French haut in voler trop haut 'to fly too high', Portuguese rápido in aprender rápido 'to learn fast', Italian bello in scrivere bello 'to write beautifully' or Romanian drept in a merge drept 'to go straight/right'. These adverbs are annotated with the category morphosyntactic structure: adjectival. Only adjective‑adverbs —that is, adjectives with adverbial functions (see section 2.2.1 for the syntactic functions of the adverbs)— are tagged and annotated, whereas adjectives with adjectival syntax are not tagged in the aaif‑database (e.g. examples such as Spanish alto in el arbol es alto 'the tree is high' or French droit in l'angle droit 'the right angle').

For the lemmatization of the examples, the uninflected standard form in present‑day language is used. It is identical to the masculine singular form of the base adjective, as in the case of the above‑mentioned forms: Spanish claro, French haut, Portuguese rápido, Italian bello, Romanian drept. Regarding the rich diversity in varieties of southern Italy (Napolitano, Siciliano, Salentino and Calabrese, see subcorpus It_A_aaif), the corresponding standard Italian lemma is annotated in order to enable a simpler lemma search, e.g. dialectal biello in Calabrian texts is lemmatized as bello. In case there is no such corresponding standard Italian lemma, a reconstructed vernacular Latin form is marked with an asterisk and written in capital letters, for example, the lemma *maiaticum for the Neapolitan adjective‑adverb majàteco 'solidly' in the following example:

(1) parlava Napoletano chiantuto e majàteco (La Posilecheata, 1684, Napoli, It_A_aaif)
'he was speaking Neapolitan robustly and solidly'.

The category inflection refers to the number and gender of the adjectival form. Since most adjective‑adverbs appear in their masculine singular form (as in the above‑mentioned examples), they are considered an invariable and uninflected word‑class. Despite the normative preference for invariable/uninflected adverbs, especially older examples in historical corpora show inflected uses of adverbs, as in this Spanish example:

(2) que se vayan luego derechos a Marbella (Sp_AP_SH3, Epistolario del Conde de Tendilla, 1504‑1506)
'they should go straight to Marbella'

In this example, the adjective‑adverb derechos (lemma: derecho 'straight') agrees with the subject; therefore, the inflectional category of this example is annotated as inflection: masculine plural. Consequently, Spanish derecha is annotated as inflection: feminine singular and derechas as inflection: feminine plural. Since the masculine singular form of the adjective‑adverb is identical to the unmarked, uninflected form, in examples such as (3) it is not possible to distinguish between an uninflected or inflected use of the adjective‑adverb derecho:

(3) que él _{[masc. sg.]} se vaya derecho _{[masc. sg.]} a Marbella

The only way to clearly identify uninflected/unmarked adverbs are cases such as (4). In this example, the subject would cause plural agreement (the form derechos would agree with the subject), whereas the adverb derecho is clearly uninflected:

(4) que se vayan_[plural] derecho_{[masc. sg.]} a Marbella

However, in most cases the examples are simply ambiguous, as in (3). Therefore, the tag inflection: uninflected is considered the default‑tag which necessarily also covers the inflectional category masculine singular, whereas the other inflectional categories can be tagged separately (masculine plural / feminine singular / feminine plural and —for languages like Romanian— neuter singular and neuter plural).

The lemmatization enables searching for all inflected variants of one adjective‑adverb. For example, the simple query for the Spanish lemma derecho (without specifying the category "inflection") delivers all inflected and uninflected examples, derecho, derechos, derecha, derechas, as well as diminutive forms such as derechito, derechitas, etc.

The specific inflectional categories (masculine/feminine/neuter and singular/plural) are annotated in all corpora except in the French ones. For French, the tags for the inflectional categories (besides the default category uninflected = masculine singular) are based on the distinction between inflection: audible inflection (for example, sain > saine in vivre saine) and inflection: inaudible inflection (for example, the inflection of seul > seule is only relevant for written French, as in the example je m'en vais seule, but not in spoken French).

2.1.2. Derived adverbs

Some subcorpora of the database include examples of derived adverbs, although they are not the main focus of the aaif‑database. Derived adverbs are formed by a derivational suffix like ‑mente which is added to an adjectival base. Hence, contrarily to adjective‑adverbs, where an adjective is directly converted into an adverb, derived adverbs show a specific adverbial suffix which clearly marks them as belonging to the word‑class adverb. The tag morphosyntactic structure: derived -mente covers all variants of this suffix in several Romance languages: French ‑ment and Spanish / Portuguese / Italian ‑mente, as well as historical variants such as Spanish ‑mientre, ‑miente or ‑mentre and regional varieties such as Calabrian and Sicilian ‑menti. Romanian only has a few borrowed mente‑adverbs, which are not documented in the aaif‑database.

The lemma of derived adverbs is the adjectival lemma of the derivational base: Spanish ciertamente 'certainly' is lemmatized as cierto, Portuguese exatamente 'exactly' as exato. Since the systematic derivational base of mente‑adverbs is the feminine singular form of the adjective (Spanish cierta > ciertamente and Portuguese exata > exatamente), we consider that a fossilized form of the stem, but not an inflectional process on the whole adverb. Therefore, mente‑adverbs are by default categorized as inflection: uninflected. We would only consider mente‑adverbs to be inflected if the whole adverb takes, for example, a plural morpheme: e.g. informal Spanish los ríos meramentes flotables, which is considered incorrect but still attested (not in the database).

Besides ‑mente, other language‑specific derivational suffixes are the Romanian suffixes ‑eşte (e.g. crestineşte 'in a Christian manner'), ‑iş (e.g. făţiş 'openly, directly') and ‑ul; for each derivational suffix there is a specific tag: morphosyntactic structure: derived -eşte, / derived -iş, derived -ul.

The annotation tool also includes tags for the Italian suffix ‑oni/‑one (used for actions that involve a body part touching the ground: ginocchione/i 'on one's knees', bocconi 'on one's mouth') and the English suffix ‑ly. Those two options as well as derived -ul are not shown in the search interface, since there have not (yet) been any annotated examples.

2.1.3. Nouns used as adverbials

This category applies to nouns that are used with adverbial functions. For example, in Spanish pasarlo bomba 'to have a very good time', the noun bomba 'bomb' is used as a manner adverb meaning (metaphorically) 'very well'. Nouns also appear in adverbial locutions (see section 2.1.5), e.g. Portuguese com certeza 'certainly' (literally 'with certainty') and com clareza 'clearly' (literally 'with clarity'), as well as in reduplicated locutions, e.g. Neapolitan ardere miccia miccia (see section 2.1.4). These nouns are tagged as adverbs and further annotated with the tag morphosyntactic structure: noun.

The lemma of these adverbials is the basic (singular) form of the noun. Therefore, contrarily to mente‑adverbs, adjective‑adverbs and prepositional phrases of the type "preposition + adjective" —which are all lemmatized with an adjectival lemma— nouns are lemmatized as nouns. The category inflection (feminine singular / masculine plural, etc.) is also tagged; note that the inflectional categories in the annotation tool were established for adjectives, thus gender (masculine, feminine, neuter) appears as an inflectional category even though it is inherent and nouns are only inflected for number (singular, plural).

This category is not systematically annotated in the database, since nouns are not the main target of the aaif‑project. For example, in the Portuguese database Pt_APM_DeG, only the prepositional phrases com certeza and com clareza are tagged because they form paradigms with the adverbs certo, ao certo, certamente (common adjectival lemma: certo) and claro, claramente, às claras, em claro (common adjectival lemma: claro). On the contrary, other prepositional adverbials such as sem interrupções 'without interruptions' are not tagged in this subcorpus, since they share no lexemic root with any adjective‑adverb.

Only nouns that are used adverbially (via direct conversion or via reduplication) or that form part of a prepositional phrase are tagged as nouns. In cases where the derivational basis of a derived adverb (see previous section) is a noun (instead of an adjective), the adverb has been classified as derived. For example, the Romanian adverb făţiş 'openly, directly' is formed by the noun faţă 'face, visage' and the derivational suffix ‑iş, and has been classified as a derived adverb with the morphosyntactic structure : derived -iş.

2.1.4. Reduplicated adverbs

The category reduplication refers to the lexicalized or spontaneous consecutive repetition (doubling) of the same adverb. In some dialects of Southern Italy, the reduplication of adjective‑adverbs is especially productive. For example, Neapolitan sano sano means 'entirely':

(5) ma la diceno tutta sana sana (It_A_aaif, Aminta, 1725)
'but they tell all [these things] entirely'

Reduplicated adverbs are tagged only once: e.g. only one word‑form sana has been tagged with the same categories as a single‑word adverb, that is: lemmatized as sano and annotated as an adverb with the morphosyntactic structure : adjectival and its specific inflectional features (inflection: feminine singular). Additionally, sana has been tagged with the separate category reduplicated : true.

Likewise, one instance of paso 'slowly' was annotated as an adjective‑adverb and tagged as reduplicated in the following Spanish example:

(6) ¡Paso, paso, Pármeno, no saltes! (Sp_AP_SH3, La Celestina, 1499)
'Slowly, slowly, Pármeno, do not jump'.

As this example shows, adjective‑adverbs can also be productively reduplicated in order to boost their meaning, without having a specific lexicalized or idiomatic meaning. Any type of reduplication is tagged in the aaif‑database, regardless of its meaning.

In the same vein, reduplications of a noun may be used adverbially, e.g. the Neapolitan noun miccia 'fuse':

(7) ardere miccia miccia (It_A_aaif, La vita e morte de lo sciatamone 'mpetrato, 1750)
'to burn weak'

In the case of miccia miccia, one word‑form miccia is annotated as an adverb with the tag morphosyntactic structure : noun and inflection: feminine singular. Additionally, miccia has been tagged with the category reduplicated : true. In this case, miccia miccia (literally 'fuse') has a specific lexicalized idiomatic meaning.

Hence, in the case of reduplicated adverbials, only one form is annotated with the same tags as a single‑word adverb (be it an adjective‑adverb or a noun) and the feature of reduplication is not found under the category "morphosyntactic structure", but as a separate category: the tag reduplicated : true. Therefore, the default tag for all other examples is reduplicated : false.

2.1.5. Prepositional phrases (adverbial locutions)

Some periphrastic adverbials are formed by multi‑word sequences introduced by a preposition. The following combinations can be tagged with the annotation tool, although not all possibilities are attested in the database:

Formation	Examples
Preposition + Adjective	Portuguese de novo 'again'
Preposition + Preposition + Adjective	Spanish por de pronto 'provisionally'
Preposition + Article + Adjective	Spanish por lo seguro 'surely'
Preposition + Possessive + Adjective	Spanish a mis solas 'on my own'
Preposition + Noun	Portuguese com certeza 'certainly'
Preposition + Article + Noun	Portuguese na verdade 'in truth'
Preposition + Adjective + Article	Romanian cu dereptul 'rightly, correctly'
Preposition + Derived Adverb	Romanian de făţiş 'openly, directly'

For each word‑class (part of speech) there is a separate tag; the tag for the word‑class adverb is thus combined with a separate word‑class tag for prepositions and any articles or possessives. Therefore, in these adverbial locutions (prepositional phrases), each part of the multi‑word sequence is annotated with the corresponding word-class category.

For the lemma of these Romanian adverbials, the underlying adjectival (or nominal) base is annotated, i.e. cu dereptul 'rightly, correctly' is lemmatized as drept (hence, the same etymology as Spanish derecho, see examples above).

For example, the adjective novo 'new' in the Portuguese locution de novo 'again' is annotated with the tag for the word‑class adverb. Hence, novo is lemmatized and categorized as morphosyntactic structure: adjectival. Additionally, novo is tagged with the category part of prepositional phrase: true (hence, the default for all other examples is part of prepositional phrase: false). Likewise, the noun certeza 'certainty' in the prepositional phrase com certeza 'certainly' is categorized as an adverb with the morphosyntactic structure: noun.

The Romanian derived adverb făţiş (faţă 'face' + derivational suffix -iş) occurs both alone and in the prepositional phrase de făţiş 'openly, directly'; in both cases it is tagged with morphosyntactic structure: derived -iş. Whereas examples of de făţiş are tagged as part of prepositional phrase: true, examples of the simple adverb făţiş are tagged as part of prepositional phrase: false.

Therefore, the adjectives, nouns or adverbs that form part of a prepositional phrase are tagged and lemmatized as a single‑word adverb (be it an adjective‑adverb, a noun or a derived adverb, see previous sections), also regarding their possible inflection.

Consequently, prepositional phrases can be queried in the search interface by selecting the category Adverb: part of prepositional phrase: true. Additionally, the prepositions that form part of a prepositional phrase are lemmatized and can be found via lemma search. Therefore, it is possible, for example, to search prepositions cross‑linguistically in several Romance languages, e.g. the lemma search for the preposition de in various subcorpora simultaneously delivers examples such as Portuguese de novo 'again', de leve 'lightly', Spanish de cierto 'certainly', de seguro 'surely', and Romanian de făţiş 'openly', de curând 'quickly'. In the case of the subcorpus It_A_aaif, the lemmatization of the prepositions is based on the standard Italian form, therefore the dialectal form de veru 'really' (attested in Salento) can be found via lemma search for di.

In prepositional phrases which are formed by more than one preposition, each preposition is tagged and lemmatized separately. Therefore, examples such as Spanish por de pronto 'provisionally' show three tagged word‑classes: two prepositions and one adjective‑adverb. Since both prepositions are lemmatized, examples like the following can be found both via lemma search for por and lemma search for de:

(8) Supongamos, por de pronto un sustantivo, por ejemplo: "herencia". (Sp_AP_CORDIAM, Documentos comentativos: El Coronel, Uruguay, 1880)
'Let’s assume, provisionally, a noun, for example: "inheritance"'

In some cases, the preposition is contracted with the adjective: Romanian deplin 'completely, fully' is an contracted form: preposition de + adjective plin. In these cases, both the preposition and the adjective are tagged and lemmatized. The contracted preposition is additionally tagged with the category contracted: true. Hence, the default tag for all prepositions that are not fused is contracted: false.

Occasionally, prepositional adverbials are formed with two contracted prepositions. The Romanian adverbial îndelung 'during a long time' is formed as follows: contracted preposition în + contracted preposition de + adjective lung'long'.

In some cases, the periphrastic adverb may also include an article which is tagged as a separate word‑class and annotated with the categories gender (masculine / feminine / neuter / undefined) and number (singular / plural / undefined). Articles are not lemmatized. Therefore, prepositional phrases like Spanish a la fija, a lo seguro contain three tagged word‑classes (preposition + article + adverb), which can be queried in the search mask independently or in different combinations.

In some Romance languages, articles may be contracted with the preposition, e.g. Portuguese ao (preposition a + masculine singular article o) and às (preposition a + feminine plural article as). Therefore, prepositional adverbials like ao certo 'certainly' and às claras 'clearly' are tagged as follows:

ao certo
preposition	lemma a contracted: false
article	gender: masculinenumber: singularcontracted: true
adverb	lemma certo morphosyntactic structure: adjectivalinflection: uninflectedpart of prepositional phrase: true

às claras
preposition	lemma a contracted: false
article	gender: femininenumber: pluralcontracted: true
adverb	lemma claro morphosyntactic structure: adjectivalinflection: feminine pluralpart of prepositional phrase: true

In these examples, the prepositions themselves are tagged as contracted: false, since the category contracted: true applies to prepositions contracted to the adjective‑adverb (see above). Simply for practical reasons and in order to differentiate the patterns, the contraction of a + o and a + as, in turn, is tagged by annotating the article as contracted.

In Romanian, the article is postponed and contracted to the noun or nominalized adjective. For example, the morphemes of the Romanian adverb de amănuntul 'detailed' are: de a‑mănunt‑ul. Hence, this adverbial is based on the adjective mănunt 'small, detailed' (< lat. minutus) and further consists of two prepositions (de and contracted a) and of the postponed and contracted article -ul. It is tagged with four tags:

de amănuntul
Preposition	lemma de contracted: false
Preposition	lemma a contracted: true
Adverb	lemma mănunt morphosyntactic structure: adjective part of prepositional phrase: true
Article	gender: masculine number: singular contracted: true

In those cases where the periphrastic adverb includes a possessive (instead of an article), it is also annotated with a tag for this part of speech. Possessives are not lemmatized. The categories for possessives are gender (masculine / feminine / neuter / undefined), number (singular / plural / undefined), and person (1st / 2nd / 3rd). For example, the Spanish prepositional phrase a mis solas 'on my own' has been tagged as follows:

a mis solas:
preposition	lemma a contracted: false
possessive	number: plural gender: undefined person: 1st
adverb	lemma solo morphosyntactic structure: adjectival inflection: feminine plural part of prepositional phrase: true

In sum, the annotation tool serves to annotate each word of a prepositional phrase separately and therefore allows different (or the same) word‑class tags to be freely combined.

2.2.1. Attribution target

The main function of the word‑class adverb has been defined as modification: adverbs modify another part of the sentence (a word, a phrase or even the sentence as a whole). The syntactic scope of the adverb —the segment it modifies— is tagged under the category attribution target. Depending on which types of adverbs have been compiled for a specific subcorpus (e.g. only manner adverbs or also other adverbial functions), more or fewer categories have been annotated and can be found via the search interface. Note that some examples may be ambiguous and can be interpreted differently; therefore, some categories may be vague and may have been used slightly differently (i.e. a broader or more general classification) by the different linguists who annotated the various subcorpora.

The following classifications were established for the aaif‑annotation tool:

• attribution target: verb

According to literature, the main function of adverbs is manner, that is, verb modification: Spanish proceder de ligero 'to proceed quickly', French dire tout bas 'to say very quietly', Portuguese participar ativamente 'to participate actively', Italian scrivere bello 'to write beautifully', or Romanian a zice chiar 'to speak clearly'. In these examples, the syntactic scope of the adverb is annotated as attribution target: verb. The verb form modified by the adverbial is tagged and lemmatized too (see the tag for the word‑class verb, section 3).

• attribution target: verb and subject

In certain sentences, the manner adverb modifies both the action described by the verb of the sentence and the subject involved, e.g. the following Spanish example:

(9) decía el Marquesito muy serio (Sp_AP_SH3, La Regenta, 1884‑1885)
'said the little marquis very serious(ly)'

The adjective‑adverb serio 'serious(ly)' describes both the action expressed by the verb decir 'to say' (the serious manner of speaking) and the attitude of the subject, el Marquesito, while speaking. Whereas secondary predicates modify exclusively the subject, albeit limited to the scope of the verb (e.g. Juan compró una casa desesperado. 'Desperate, John bought a house'), the adverbs of the category attribution target: verb and subject also modify the verb at the same time.

The inflection of the adjective‑adverb (see section 2.1.1.1) occasionally agrees in number and gender with the subject: in the French example elle vivait saine 'she lived healthy', the adjective‑adverb saine (fem. sg.) agrees with the subject elle (fem. sg.) and describes both a quality of the subject ('she is healthy') and a characteristic of the verb vivre ('she lives in a healthy manner'). In these kinds of examples, in addition to the verb, occasionally the subject the adverb refers to is also annotated in some corpora (see the tag for the word‑class subject, section 4). It should be added however that the distinction between "subject only", "verb and subject" and "secondary predicate" (not included in the database) is all but clear‑cut. Hence, the database provides a rough classification that should be subject to more detailed analyses in research using the data.

• attribution target: verb and object

Similarly, this tag applies to those manner adverbials that modify both the (transitive) verb of the sentence and the object involved, as in the following Spanish example:

(10) los martirios [...] los ha de pagar bien caros (Sp_AP_SH3, Los bandidos del Río frío, México, 1889‑1891)
's/he has to pay very dearly for the torments'.

In this example, the adjective‑adverb caro 'dear(ly), expensive' is inflected (masc. pl. form caros) and it agrees not with the (singular) subject of the sentence, but with los martirios (masc. pl.), i.e. the direct object of the verb pagar 'to pay'.

Likewise, reduplicated adverbs can also be inflected and may agree with the object of the sentence, as in the following Neapolitan example:

(11) ma la diceno tutta sana sana (It_A_aaif, Aminta, 1725)
'but they tell all [these things] entirely'

The reduplicated adjective‑adverb sana sana 'entirely' (literal translation: 'healthy‑healthy') concords with the direct object la (fem. sg.).

Likewise, in the French sentence

(12) On peut la [=la soupe] manger chaude, froide (2015, Fr_A_Web)
'you can eat it [=the soup] hot or cold',

the adjective‑adverbs chaud and froid are inflected (fem. sg.), agreeing with the direct object la [=la soupe]. Hence, the adjective‑adverbs chaud 'hot' and froid 'cold' refer to a quality of the direct object (the soup) at the moment the action described by the verb manger 'to eat' takes place.

Occasionally, the adverb refers to an implicit direct object of verbs like Portuguese votar 'to vote', as in the sentence

(13) O povo brasileiro […] continua votando errado, sempre (Pt_APM_DeG, 1993, Rio de Janeiro, parte escrita)
'the Brazilian people always continue voting wrong(ly)'.

The adjective‑adverb errado 'wrong(ly)' does not describe solely the manner of voting (the action of voting itself is not invalid or done in an incorrect manner), but refers to the implicit direct object (e.g. they voted the wrong party), hence it is also annotated as attribution target: verb and subject.

Besides, adverbs also modify other parts of the sentence than the verb. The following attribution target‑categories do not consider the verb of the sentence, which is, consequently, not tagged:

• attribution target: adjective

Adverbs may modify adjectives, e.g. the Romanian adjective‑adverb foarte 'strong(ly)' modifies the adjective mare 'big' in this example:

(14) Ţara Ungurească mai denainte vreme era foarte mare (1725, Grigore Ureche: Letopiseţul Ţărîi Moldovei, Ro_ADP_aaif)
'The country of Hungary was really large before'.

In this example, foarte is annotated with the category attribution target: adjective; in turn, the adjective mare itself is not tagged (since adjectives with adjectival syntax are not included in the annotation model).

Adjective‑adverbs may also be inflected and show agreement with the inflectional categories of the adjective they modify (which itself concords with the subject): the inflected (fem. sg.) Neapolitan adjective‑adverb meza (standard Italian lemma: mezzo) modifies the adjective vregognosa (modern standard Italian: vergognosa) in the following sentence:

(15) steva meza vregognosa (It_AP_aaif, 1684, Napoli, Sarnelli, La Posilecheata)
'she was half shy/ashamed'.

As noted above, the adjective modified by the adverb (that is, the adjective as an attribution target) is not tagged itself. Nevertheless, deverbal adjectives (e.g. participles) are classified and tagged as verb forms (see section 3 for the tagging of the word‑class verb), since this allows for annotating and lemmatizing of the participle. For example, in the Spanish sentence

(16) como un cordero recién esquilado (2007, El corazón helado, Sp_AP_SH3)
'like a recently sheared sheep'

the adjective‑adverb recién (attribution target: adjective) modifies the deverbal adjective esquilado, which itself has been annotated with the tag for the word‑class verb and lemmatized as esquilar.

• attribution target: adverb

Adverbs may also modify other adverbs: the syntactic scope of the Spanish adjective‑adverb harto in interpretar harto duramente ('to interpret/perform extremely hard') is the mente‑adverb duramente, which, in turn, is a manner adverb that modifies the verb interpretar. Therefore, harto is tagged with the category attribution target: adverb. Depending on the types of adverbs tagged in each dataset (e.g. additionally to adjective‑adverbs also mente‑adverbs), the adverb modified by an adverb has also been tagged itself. In the following Portuguese example, the mente‑adverb praticamente modifies the adjective‑adverb sozinho:

(17) a pessoa trabalha praticamente sozinho ... na sua ... na sua criação (1993, Natal, Pt_APM_DeG, oral corpus)
'the person works practically/nearly alone on its creation'.

In this example, the adverb praticamente has been tagged with the category attribution target: adverb and the adjective‑adverb sozinho with the category attribution target: verb and subject.

In Romanian, an adjective‑adverb (e.g. foarte 'strong(ly)') can also modify another adjective‑adverb (e.g. greu 'hard(ly), difficult(ly)'):

(18) foarte greu au stătut strâmbătate (1774, Bertoldo, Ro_APD_aaif)
'with real difficulty they stood crooked'

In this sentence, both the adjective‑adverbs foarte (attribution target: adverb) and greu (attribution target: verb) have been tagged.

• attribution target: noun or syntagma without verb-reference

This category includes all types of nouns or noun phrases as well as prepositional phrases or any type of phrase which does not include the verb of the sentence (in the annotation model, we use the term syntagma for phrase). For example, the Spanish adjective‑adverb sola (inflected: fem. sg., lemma solo) modifies the syntagma una palabra (which is a noun phrase) in the following example:

(19) con sola una palabra (1499, La Celestina, Sp_AP_SH3)
'with just/only one word'.

Likewise, in the Portuguese example

(20) principalmente no verão (1993, Rio de Janeiro, parte oral, Pt_APD_DeG)
'particularly during the summer'

the adverb principalmente modifies the prepositional phrase no verão.

Finally, in the Neapolitan example

(21) simmo iusto quatto (1934, Napoli, Salvatore Di Giacomo: Tutto il teatro, It_A_aaif)
'we are exactly four'

the adjective‑adverb iusto (standard Italian lemma giusto) refers to the numeral. Hence, this heterogeneous category includes different types of phrases or words (nouns, numerals, etc. — all word‑classes except adjective and adverb for which there is a proper category, see above). The common criterion is that the modified phrase or word does not include the verb, hence, the adverb does not characterize the action described by the verb.

• attribution target: sentence

Adverbs of this category modify the sentence as a whole. For example, in the Neapolitan phrase

(22) sì lo conosco, lo conosco sicuro (1925, Eduardo Scarpetta, Tutto il teatro, It_A_aaif)
'yes, I know him, I know him surely/ for sure'

the adjective‑adverb sicuro does not refer as a manner adverb to the verb conoscere, but its scope is the whole sentence; sicuro is a sentence‑adverb that characterizes the whole proposition —the fact of knowing him— as 'sure'.

Similarly, the Spanish adjective‑adverb cierto 'certainly' modifies the whole sentence (note that the adverb is also separated by commas, hence there is a prosodic pause):

(23) Verdad sea que, cierto, duele más la pérdida (1499, La Celestina, Sp_AP_SH3)
'Certainly, it is true that the loss hurts more'.

Likewise, the Portuguese adverb infelizmente 'unfortunately, unhappily' refers to the whole sentence (expressing the speaker’s point of view) in the following example:

(24) Meu grupo infelizmente não foi o vencedor (1993, Natal, parte escrita, Pt_APM_DeG)
'Unfortunately, my group was not the winner'.

The tag attribution target: sentence also applies for discourse markers, for example, affirmation markers or phatic markers (Spanish bueno, claro, Portuguese bom, exatamente, Neapolitan certo).

2.2.2. Modification of the adverb

The category modified: true applies if the adverbial itself is modified (mostly intensified) by another adverbial (hence, the default category for all non‑modified adverbs is modified: false). For example, the Romanian adjective‑adverb scurt 'short' is modified by the adverb mai 'more' in the following example:

(25) alţii au scris mai scurt (1620, Cronica universală, Ro_APD_aaif)
'others wrote more briefly'

Similarly, in the French example

(26) car ilz parloient bien bas (1489, Philippe de Commynes, Fr_A_DHAA)
'because they spoke very quietly'

the adjective‑adverb bas is modified by the adverb bien 'very'. The feature modified: true is annotated for the adjective‑adverbs scurt and bas, whereas the modifiers themselves, mai 'more' and bien 'very', are not tagged. As explained above, the aaif‑database only annotates and lemmatizes adverbials which are based on an adjectival root. Therefore, modifiers such as Spanish muy, tan, bien, French très, trop, bien, Portuguese tão, bem, demais, Romanian mai, aşa or Neapolitan tanto, cchiú (standard Italian più), così are not tagged and annotated themselves.

In the case that the modifier itself is actually based on an adjectival root, both adverbials are tagged and annotated (see section 2.2.1 attribution target: adverb), as in the case of the following Portuguese example:

(27) a pessoa trabalha praticamente sozinho ... na sua ... na sua criação (1993, Natal, Pt_APM_DeG, oral corpus)
'the person works practically/nearly alone on its creation'

Both adverbs are annotated: praticamente (attribution target: adverb; modified: false) and sozinho (attribution target: verb and subject; modified: true).

Likewise, both adjective‑adverbs, foarte and greu, in the Romanian example foarte greu au stătut strâmbătate 'with real difficulty they stood crooked' (see example 18 above) were tagged, and greu was annotated with the category modified: true.

2.2.3. Coordination of adverbs

If two or more different adverbs are coordinated, they share the same syntactic function and syntactic scope. For example, two manner adverbs can be coordinated, as in the following French example:

(28) des accents qui sonnent beaux et étranges (2016, Fr_A_Web)
'accents which sound beautiful and strange'

In this example, the two inflected adjective‑adverbs beau 'beautiful(ly)' and étrange 'strange(ly)' are both manner adverbs which both modify the verb form sonnent 'to sound'. The adverbs beau 'beautiful(ly)' and étrange 'strange(ly)' are annotated separately and both appear separately in the table of results in the search interface. Additionally, they are tagged with the category coordinated: true (hence, the default tag for all non‑coordinated examples is coordinated: false).

In the following Romanian example, two prepositional phrases are coordinated:

(29) mai pe largu şi de agiunsu scriu (1725, Letopiseţul Ţărîi Moldovei, Ro_ADP_aaif)
'I write longer/more in detail and sufficient'.

Both adverbials, pe largu 'long' (which is, by the way, modified by mai) and de agiunsu 'detailed' (modern standard Romanian lemma: de‑ajuns), are annotated and both are tagged as coordinated: true.

In line with this, more than two adverbs can be coordinated, as in this Neapolitan example:

(30) abbiannose chiano chiano adaso adaso e palillo palillo (1634‑1636, Lo cunto de li cunti, Napoli, It_A_aaif)
'they set off [standard Italian verb avviarsi] slowly, quietly and gently'

The three coordinated adverbs are each reduplicated (see section 2.1.4). Hence, one of the instances of the adjective‑adverb chiano [standard Italian lemma piano] is lemmatized and tagged as reduplicated: true and coordinated: true, and the same holds for the reduplicated and coordinated adjective‑adverb adaso [standard Italian adjective‑adverb lemma adagio] and for the reduplicated and coordinated noun palillo.

Coordination is also possible between different adverb types, i.e. with different morphosyntactic structures. Depending on which types of adverbs are tagged in a given subcorpus (adjective‑adverbs, derived adverbs, prepositional phrases, etc.), the coordinated adverb may be tagged or not. For example, in the following Portuguese sentence, a mente‑adverb is coordinated with a prepositional phrase (preposition + noun: com clareza) and both are annotated because this subcorpus includes both types of adverbials:

(31) onde mostra objetivamente e com clareza como pratica‑se o esporte (1994, Relato de interação, Pt_APM_DeG)
'where he/she shows objectively and clearly how this sport is practised'

On the contrary, adverbs like Spanish bien 'well' are not tagged, as in the following example:

(32) Francisco manejaba rápido y bien (1996, Los últimos días de "La Prensa", 1996, Sp_A_CDH)
'Francisco drove quickly and well'

Therefore, only rápido is tagged and categorized as coordinated: true.