OLDES

This corpus is developed under the project ONTOSEM (FFI2010-15006).

Corpus description

This diachronic corpus of Spanish, developed by Cristina Sánchez-Marco, includes 674 texts that cover from the 12th century to the 20th century containing a total of 44,470,288 words. The texts are sourced from open repositories and collections available in various univeristy libraries. Efforts have been made to ensure that the texts are representative and balanced in terms of genres and the presence of material from different centuries that are covered. The genres represented include: poetry, history, laws, didactics, prose, religion, medicine, letters, theater, and press. The objective of the project is to develop a tool that helps understand linguistic change in general and the evolution of Peninsular Spanish in particular; therefore, all texts belong to that variety.

Tagging of the corpus

The corpus has been tagged with an existing linguistic analyzer, Freeling. This tool is designed for contemporary Spanish, and it was necessary to expand it to handle the lexical and orthographic variability of a diachronic corpus. Here is how this process was carried out.

Freeling uses the Parole tagset, developed by the EAGLES group, as a common tool for the computerized processing of various European languages. However, in some aspects, the tags used in this corpus differ from the original Parole ones. Some changes have been introduced for two main reasons: on one hand, to adapt the tagset to the different stages of Spanish; and, on the other hand, new tags have been implemented so that the analysis result can function as suitable input for a syntactic analyzer, which is the next step of this project. On this page you can consult the list of tags used in this corpus. If you are already familiar with the Parole system, please refer to the list to see the changes that have been introduced.

Consulting the corpus

The corpus can be accessed through the Corpus Query Processor (CQP), both in its command-line version and in its graphical interface CQPweb. In any case, CQP allows conducting searches by complex patterns using positional attributes related to a single item (word, lemma, tag), or structural attributes related to phrases (length, position in the text...) or texts (date, title, author, genre...). The available attributes can be consulted here. These attributes can be used to refine searches through the 'Restricted Query' option, instead of 'Standard Query', in the CQPweb interface. In the current state of development, the CQPweb interface allows automatically restricting searches by century, genre, and author.

CQPweb allows using the CQP syntax, as you would in a terminal, or in a simplified language called 'Simple Query'. The manual for CQP, with the description of its syntax, can be found here, and a brief guide for conducting searches using 'Simple Query' is available here. This corpus can be accessed using the "guest" account on CQPweb.

Tagset

1. Adjectives

Position	Attribute	Value	Code
1	Category	Adjective	A
2	Type	Qualificative	Q
		Ordinal	O
		Possessive	X
3	Degree	-	0
3	Degree	Appreciative	A
4	Gender	Masculine	M
		Feminine	F
		Common	C
5	Number	Singular	S
		Plural	P
		Invariable	N
6	Person	-	0
		First	1
		Second	2
		Third	3
7	Possessor	-	0
		Singular	3
		Plural	P

Qualificative and ordinal adjectives are tagged following the Parole system. However, there is a new category: possessive adjectives (AX). This includes words labeled as Possessive Pronouns in the original Parole. The reasons for this are primarily distributional: Freeling tags very frequent sequences like la mía with a null nominal category, as determiner + pronoun; however, these combinations should in principle be avoided in the language (*el las, *el tú), and a syntactic analyzer should be able to recognize the ungrammaticality of such a sequence. Since distributionally these elements (mío, tuya, nuestra...) behave like adjectives (and it could even be argued that semantically they are), we have made the decision to tag them as possessive adjectives.

Another change that must be taken into account regarding elements with participial morphology (-ado/ada/ados/adas) is that they all now form a separate new category (T; see below).

2. Adverbs

Position	Attributes	Value	Code
1	Category	Adverb	R
2	Type	General	G
2	Type	Negative	N

There are no changes to the original Parole tagging system.

3. Determiners

Position	Attribute	Value	Code
1	Category	Determiner	D
2	Type	Demonstrative	D
		Possessive	P
		Interrogative	T
		Exclamative	E
		Indefinite	I
		Article	A
		Relative	R
		Numeral	N
3	Person	First	1
		Second	2
		Third	3
4	Gender	Masculine	M
		Feminine	F
		Common	C
		Neuter	N
5	Number	Singular	S
		Plural	P
		Invariable	N
6	Possessor	Singular	S
6	Possessor	Plural	P

There are no changes to the original Parole tagging system.

4. Quantifiers

Position	Attribute	Value	Code
1	Category	Quantifier	Q
2	Gender	Masculine	M
		Feminine	F
		Common	C
3	Number	Singular	S
		Plural	P
		Invariable	N

This category did not exist in the original Parole tagging system. It includes, as lemmas, mucho, todo, and cada. Despite their differences, these three elements share common features:

They can license a nominal phrase: cada niño, todo niño, muchos niños.
Some can combine with a determiner: todos los niños, los muchos niños.
Some can behave as predicative adjectives or adverbs: son muchos, me gusta mucho.
They are very frequent.
Both traditional grammar and the Parole tagging system have difficulties treating them, often resorting to assigning them multiple grammatical categories (D, P, A, Adv...).

The Q tag is intended to provide a unified treatment for them and to generate suitable input for the syntactic analyzer. It's important to note that all occurrences of todo, cada, and mucho are analyzed as quantifiers. However, the exact definition of this category and the extension of the Q tag to other similar elements will be defined in the next phase of the project.

5. Nouns

Position	Attribute	Value	Code
1	Category	Noun	N
2	Type	Common	C
2	Type	Proper	P
3	Gender	Masculine	M
		Feminine	F
		Common	C
4	Number	Singular	S
		Plural	P
		Invariable	N
5-6	Semantic classification	Being-Person	SP
		Organization	O0
		Place	G0
7	Degree	Appreciative	A

There are no changes to the original Parole tagging system.

6. Verbs

Position	Attribute	Value	Code
1	Category	Verb	V
2	Type	Main	M
		Auxiliary	A
		Semiauxiliary	S
3	Mode	Indicative	I
		Subjunctive	S
		Imperative	M
		Infinitive	N
		Gerund	G
4	Tense	Present	P
		Imperfect	I
		Future	F
		Past	S
		Conditional	C
		-	0
5	Person	First	1
		Second	2
		Third	3
6	Number	Singular	S
6	Number	Plural	P
7	Gender	Masculine	M
7	Gender	Feminine	F

Examples of complete verbal paradigms:

Tense	Main verbs
Tense	Form	Lemma	Tag
Present Indicative	canto	cantar	VMIP1S0
	cantas	cantar	VMIP2S0
	canta	cantar	VMIP3S0
	cantamos	cantar	VMIP1P0
	cantáis	cantar	VMIP2P0
	cantan	cantar	VMIP3P0
Past Imperfect	cantaba	cantar	VMII1S0
	cantabas	cantar	VMII2S0
	cantaba	cantar	VMII3S0
	cantábamos	cantar	VMII1P0
	cantábais	cantar	VMII2P0
	cantaban	cantar	VMII3P0
Simple Past	canté	cantar	VMIS1S0
	cantaste	cantar	VMIS2S0
	cantó	cantar	VMIS3S0
	cantamos	cantar	VMIS1P0
	cantasteis	cantar	VMIS2P0
	cantaron	cantar	VMIS3P0
Future Indicative	cantaré	cantar	VMIF1S0
	cantarás	cantar	VMIF2S0
	cantará	cantar	VMIF3S0
	cantaremos	cantar	VMIF1P0
	cantaréis	cantar	VMIF2P0
	cantarán	cantar	VMIF3P0
Conditional	cantaría	cantar	VMCP1S0
	cantarías	cantar	VMCP2S0
	cantaría	cantar	VMCP3S0
	cantaríamos	cantar	VMCP1P0
	cantaríais	cantar	VMCP2P0
	cantarían	cantar	VMCP3P0
Present Subjunctive	cante	cantar	VMSP1S0
	cantes	cantar	VMSP2S0
	cante	cantar	VMSP3S0
	cantemos	cantar	VMSP1P0
	cantéis	cantar	VMSP2P0
	canten	cantar	VMSP3P0
Imperfect Subjunctive	cantara	cantar	VMSI1S0
	cantaras	cantar	VMSI2S0
	cantara	cantar	VMSI3S0
	cantáramos	cantar	VMSI1P0
	cantarais	cantar	VMSI2P0
	cantaran	cantar	VMSI3P0
	cantase	cantar	VMSI1S0
	cantases	cantar	VMSI2S0
	cantase	cantar	VMSI3S0
	cantásemos	cantar	VMSI1P0
	cantaseis	cantar	VMSI2P0
	cantasen	cantar	VMSI3P0
Future Subjunctive	cantare	cantar	VMSF1S0
	cantares	cantar	VMSF2S0
	cantare	cantar	VMSF3S0
	cantáremos	cantar	VMSF1P0
	cantareis	cantar	VMSF2P0
	cantaren	cantar	VMSF3P0
Gerund	cantando	cantar	VMG0000
Imperative	canta	cantar	VMMP2S0
	cante	cantar	VMMP3S0
	cantemos	cantar	VMMP1P0
	cantad	cantar	VMMP2P0
	canten	cantar	VMMP3P0
Infinitive	cantar	cantar	VMN0000

Auxiliary verbs
Form	Lemma	Tag
soy	ser	VAIP1S0
eres	ser	VAIP2S0
es	ser	VAIP3S0
somos	ser	VAIP1P0
sois	ser	VAIP2P0
son	ser	VAIP3P0
era	ser	VAII1S0
eras	ser	VAII2S0
era	ser	VAII3S0
éramos	ser	VAII1P0
érais	ser	VAII2P0
eran	ser	VAII3P0
fui	ser	VAIS1S0
fuiste	ser	VAIS2S0
fue	ser	VAIS3S0
fuimos	ser	VAIS1P0
fuisteis	ser	VAIS2P0
fueron	ser	VAIS3P0
seré	ser	VAIF1S0
serás	ser	VAIF2S0
será	ser	VAIF3S0
seremos	ser	VAIF1P0
seréis	ser	VAIF2P0
serán	ser	VAIF3P0
sería	ser	VACP1S0
serías	ser	VACP2S0
sería	ser	VACP3S0
seríamos	ser	VACP1P0
seríais	ser	VACP2P0
serían	ser	VACP3P0
sea	ser	VASP1S0
seas	ser	VASP2S0
sea	ser	VASP3S0
seamos	ser	VASP1P0
seáis	ser	VASP2P0
sean	ser	VASP3P0
fuera	ser	VASI1S0
fueras	ser	VASI2S0
fuera	ser	VASI3S0
fuéramos	ser	VASI1P0
fuerais	ser	VASI2P0
fueran	ser	VASI3P0
fuese	ser	VASI1S0
fueses	ser	VASI2S0
fuese	ser	VASI3S0
fuésemos	ser	VASI1P0
fueseis	ser	VASI2P0
fuesen	ser	VASI3P0
fuere	ser	VASF1S0
fueres	ser	VASF2S0
fuere	ser	VASF3S0
fuéremos	ser	VASF1P0
fuereis	ser	VASF2P0
fueren	ser	VASF3P0
siendo	ser	VAG0000
sé	ser	VAMP2S0
sea	ser	VAMP3S0
seamos	ser	VAMP1P0
sed	ser	VAMP2P0
sean	ser	VAMP3P0
ser	ser	VAN0000

The tagging of verbs follows the Parole criteria, except for one important point: the treatment of past participles, which have been moved to a new category, T (see below).

7. Participles

Position	Attribute	Value	Code
1	Category	Participle	T
2	Type	Main	M
		Auxiliary	A
		Semiauxiliary	S
3	Gender	Masculine	M
		Feminine	F
		Common	C
4	Number	Singular	S
		Plural	P
		Invariable	N

Examples:

Form	Lemma	Form
cantado	cantar	TMMS
cantada	cantar	TMFS
cantados	cantar	TMMP
cantadas	cantar	TMFP
estado	estar	TAMS

As mentioned above, all elements with participial morphology, originally included in the categories of adjectives and verbs in Parole, have been brought together within this new category. The reason for this change is that our tagging system must be able to cover the early stages of a Romance language, Spanish, in which it is not always easy to determine whether an element with the morphology -ado/ada/ados/adas behaves like a verb or like an adjective. We have decided to unify their treatment, so it is important to note that all uses of these elements (even when their function is clearly verbal or adjectival according to traditional criteria) are labeled as T.

8. Pronouns

Position	Attribute	Value	Code
1	Category	Pronoun	P
2	Type	Person	P
		Demonstrative	D
		Possessive	X
		Indefinite	I
		Interrogative	T
		Relative	R
		Numeral	N
		Exclamative	E
3	Person	First	1
		Second	2
		Third	3
4	Gender	Masculine	M
		Feminine	F
		Common	C
		Neuter	N
5	Number	Singular	S
		Plural	P
		Invariable	N
6	Case	Nominative	N
		Accusative	A
		Dative	D
		Oblique	O
7	Possessor	Singular	S
7	Possessor	Plural	P
8	Politeness	Polite	P

In this category, a very significant change has been introduced compared to the original Parole tagging system: the category of personal pronouns has been divided into personal pronouns (commonly referred to as 'strong pronouns') and clitics (or 'weak pronouns'). The clitics now form a separate category (L). The following table shows the new distribution. Please note that not all elements that belong to the two categories are represented here.

Personal Pronouns (PP)	Clitics (L)
yo mi nosotros nosotras conmigo ti tú usted ustedes vos vosotras vosotros contigo él ella ellas ello ellos	me nos te os le las lo lo los les se y (archaic) en (archaic)

9. Clitics

Position	Attribute	Value	Code
1	Category	Clitic	L
2	Person	First	1
		Second	2
		Third	3
		Neuter	0
3	Gender	Masculine	M
		Feminine	F
		Common	C
		Neuter	N
4	Number	Singular	S
		Plural	P
		Neuter	N
5	Case	Accusative	A
		Dative	D
		Other (different forms of SE, etc.)	O

Examples:

Form	Lemma	Tag
me (me preocupo)	me	L1CSO
se (... se vino a la corte)	se	L3CNO
se (...conseio de se defender ...)	se	L0CNO
lo (lo vio)	lo	L3MSA
lo (lo siento mucho)	lo	L3CNA
os (os dio un caballo)	te	L2CPD
las (las vio)	lo	L3FPA

All clitics (weak pronouns) are now placed within this new category L, which did not exist in the original Parole system. Clitics can be an important object of study in diachronic research. In the Parole system, they are not clearly distinguished from the rest of the personal pronouns, and the length of the resulting labels sometimes makes working with them cumbersome. To simplify this, we have created this new category, even though it means that a few elements that are pronouns are now outside the 'pronouns' category. The lemmas are always the singular masculine version of each clitic.

10. Conjunctions

Position	Attribute	Value	Code
1	Category	Conjunction	C
2	Type	Coordinating	C
2	Type	Subordinating	S

There are no changes to the original Parole tagging system.

11. Interjections

Position	Attribute	Value	Code
1	Category	Interjection	I

There are no changes to the original Parole tagging system.

12. Prepositions

Position	Attribute	Value	Code
1	Category	Adposition	S
2	Type	Preposition	P
3	Form	Simple	S
3	Form	Complex	C
4	Gender	-	0
5	Number	-	0

There are no changes to the original Parole tagging system.

13. Punctuation marks

Position	Attribute	Value	Code
1	Category	Punctuation mark	F

There are no changes to the original Parole tagging system.

14. Numbers

Position	Attribute	Value	Code
1	Category	Number	Z
2	Type	"Millar"	d

There are no changes to the original Parole tagging system.