International Journal of Lexicography, 2021, Vol. 34, No.

3, 382–399
doi: 10.1093/ijl/ecab009
Advance Access Publication Date: 20 June 2021
Article

Article

Downloaded from [Link] by OUP-USA Mirror user on 20 February 2025

NEW VERBS AND DICTIONARIES: A
METHOD FOR THE AUTOMATIC
DETECTION OF NEOLOGY IN SPANISH
VERBS
Ana Castro
Pontificia Universidad Católica de Valparaı́so, Chile ([Link].paez1@[Link])

Rogelio Nazar
Pontificia Universidad Católica de Valparaı́so, Chile ([Link]@[Link])

Irene Renau
Pontificia Universidad Católica de Valparaı́so, Chile ([Link]@[Link])

Abstract
The appearance of new verbs can be observed regularly, but verbs are not frequent-
ly investigated in neology, and they are difficult to detect automatically. In this study,
a corpus-based method is proposed to detect Spanish verbs with a series of algo-
rithms that analyse the morphology of regular verbs. The vocabulary was drawn
from a large corpus and contrasted with a major dictionary of Spanish. Then, a ser-
ies of filters were applied to distinguish between valid neologism candidates and
spelling mistakes. Around 88% of the neologisms proposed by the method were cor-
rect and we estimate that the system detected 76% of the neologisms present in the
corpus. This procedure can be included in the workflow of a lexicographic project as
a regular part of the task, as a systematic way of collecting new verbs from the data
and avoiding under-representation or bias.

Key words: computational lexicography, corpus-based lexicography, neology, Spanish, verbs

1. Introduction
In this paper, we present a methodological proposal for the detection of new Spanish verbs
in a general corpus, and for their incorporation in lexicographic tasks concerning the activ-
ity of updating dictionaries. Verbs are an open class of words and new units are regularly

C 2021 Oxford University Press. All rights reserved.

V
For permissions, please email: [Link]@[Link]
New Verbs and Dictionaries 383

added to the vocabulary, as we are observing now due to the coronavirus crisis: verbs such
as cuarentenar ‘to be in quarantine’ or desconfinar ‘to deconfine’ are new verbs created by
combining existing morphological structures in an innovative way. That is why Cook
(2010: 13) explains that ‘POS tagging of unknown words, including neologisms, can benefit
greatly from exploiting word structure’. However, morphological analysis is one of the
main sources of error in automatic neology detection, as many authors have reported.

Downloaded from [Link] by OUP-USA Mirror user on 20 February 2025

Thus, Renouf (2016) states that POS taggers struggle with words not found in a dictionary.
Janssen (2009) also explains that POS taggers can be less effective in neologisms and have a
much higher error rate. Amore et al. (2018) report problems regarding POS tagging of
Italian verbs, that, as the rest of Romance languages, have tens of possible inflected forms
for each unit. Langemets et al. (2020: 79) also conclude that, ‘in order to make neologism
discovery more effective we need more advanced tools for automatic language processing.
There were a large number of mistakes in tokenisation, lemmatisation, POS tagging, Name
Entity Recognition (NER) etc.’.
In Spanish, as in many other languages, verb morphology is composed of the stem, car-
rying the semantic information, and the inflection, carrying the grammatical information
(RAE and ASALE 2009). Thus, in new verbs, it is only the inflection that allows us to auto-
matically identify these units as verbs in the corpus. By comparing the list of the verbs
obtained this way with the headwords present in a dictionary, it is possible to detect those
that are not yet recorded and, as a consequence, might be candidates for neologism. This
allows us to obtain not only the neologisms themselves, but also corpus-based evidence of
their use, such as their frequency and dispersion, which is important for taking informed
decisions on whether or not to include them in the dictionary. The approach described, con-
sisting basically of the automatic extraction of neologisms from the corpus and their con-
trast against a dictionary, is well-established (see Section 2). We propose methods to reduce
the noise that such contrast usually produces. Moreover, our method for verb detection,
which is rule-based and simple both conceptually and technically, is not present in current
POS-taggers for Spanish, for which verbs are still a challenge, particularly those with
enclitics (Parra Escartı́n and Martı́nez Alonso 2015). For that reason, we consider this study
a contribution to the field of natural language processing, computational lexicography and
automatic detection of neologisms.
This paper is structured as follows: in the next section we offer a brief account of previ-
ous research on neology detection and classification. In Section 3 we present our method-
ology. In Section 4 we describe the results and, finally, in Section 5 we offer our
conclusions.

2. The siege of neology

2.1. Theoretical and methodological considerations about neology
A neologism can be broadly defined as a new formal or semantic linguistic unit (Rey 1976),
and it can manifest in various forms: as Klosa-Kückelhaus and Kernerman (2020: 2) state,
‘the general understanding of neologisms includes new words, new multiword units, new
elements of word formation, and new meanings of any of them’. For example, cases such as
the already mentioned verbs cuarentenar and desconfinar are classified as formal neolo-
gisms, as they are new lexical units that are added to the vocabulary. Formal neologisms
can originate using already available linguistic resources (internal neology), or can be
384 Ana Castro et al.

acquired from other languages as loans (external neology). Furthermore, a new meaning
can be added to an already existing word. An example of this, called semantic neology,
would be the case of the verb rastrear ‘to follow the trace, to investigate’, which is currently
being used with the meaning of ‘tracking the contacts of a person who has been infected
with covid-19 in order to put them in quarantine’.
Different researchers have dealt with the theoretical and methodological considerations

Downloaded from [Link] by OUP-USA Mirror user on 20 February 2025

on how to discriminate between a true neologism and hapax legomena, nonce word forma-
tions (occasionalisms), or simply spelling mistakes. A new word may appear a few times in
the corpus as a creative resource in a specific situation, but that is not enough to qualify it
as a neologism (Cabré 2015, Abel and Stemle 2018). Cabré (1999) established some non-
mutually exclusive criteria to determine whether a word is a neologism: a) if the unit has
arisen in recent years; b) if it is not listed in dictionaries; c) if it exhibits formal instability
and d) if it is perceived as new by native speakers. Later, however, Cabré and Nazar (2012)
acknowledged how problematic these criteria can be, since a word can fulfil some of them
and still not be considered a neologism. Firstly, a term may have a long history of use
among a reduced circle of experts and remain unknown to a wider audience. Secondly,
being in the dictionary is also not a reliable criterion, especially in the case of semantic neol-
ogy. And thirdly, someone may not be familiar with a particular term due to age, origin or
for cultural or sociological reasons.
One cannot find an entirely satisfactory answer to this problem, as the process of lexic-
alisation of a neologism is a continuum and there is no sharp distinction between a word
that is used for the first time and a neologism that becomes a part of the vocabulary of a
language (Kerremans and Prokic 2018). From the methodological point of view, this discus-
sion is often solved at the last step of the process by human analysis of neologism candi-
dates extracted from the corpus by some automatic system.
Regarding the problem of automatic neology detection, the comparison of corpus data
against a list of words obtained from dictionaries and other sources is a widely used meth-
odology for automatic detection of lexical neology (Janssen 2009, Abel and Stemle 2018).
Generally speaking, this approach consists of identifying lexical units that are not present
in a reference list. There are several systems for neology detection that use this approach:
Neoveille (Cartier 2016) for Chinese, French, Greek, Polish, Portuguese, Russian and
Czech; NeoTrack (Janssen 2005) for Portuguese; Sextan (Vivaldi 2000) and BuscaNeo
(Cabré and Estopà 2009) for Spanish and Catalan; AVIATOR (Renouf 1993) and
NeoCrawler (Kerremans and Prokic 2018) for English; the Korean Neologism Investigation
Project (Nam et al. 2020) for Korean; two online lexicographic resources for German
linked to a neologism detector: the Wortwarte (Lemnitzer 2010) and the
Neologismenwörterbuch (Klosa-Kückelhaus and Lüngen 2018), among others. Costin-
Gabriel and Rebedea (2014) use the Google Books N-Gram corpus for neologism detection
in English. Cartier (2016) and Kerremans and Prokic (2018) seem to work at the token
level, i.e., without tagging. However, in general in this approach one must deal with the
problem of POS-tagging when tokens have many different inflections, as lemmatisation is a
key consideration if one wants to compare two lists of words. For example, in the case of
Spanish verbs, it would not be accurate to compare only tokens, because some of the verb
inflections have the same form as an adjective or a noun (e.g., amo ‘I love’ has the same
form as the masculine noun amo ‘master’).
New Verbs and Dictionaries 385

As we have observed, dictionaries are both receivers of neologisms and material for
neology detection, working as stop lists. Concerning lexicographic projects, the fact that a
word is missing from the lemma list does not necessarily mean that it is a neologism, as the
criteria for the creation of the macrostructure of the dictionary involve many complex fac-
tors, including the type of user and their needs, the size and type of dictionary, the prescrip-
tive approach of the dictionary, the low frequency and dispersion of the word, etc. (Alvar

Downloaded from [Link] by OUP-USA Mirror user on 20 February 2025

Ezquerra 2007, and Atkins and Rundell 2008, for dictionary projects in general). Indeed,
Baayen and Renouf (1996: 69) state that dictionaries ‘are not a reliable source for studying
morphological productivity’, because lexical productivity is a wider phenomenon. Cartier
and Sablayrolles (2008) also argue that there are technical limitations related to the list of
headwords, which are not displayed in all their forms in the dictionary. However, as
Ca~nete et al. (2019) state, the inclusion of a new word in a dictionary can be considered as
the last step of the lexicalisation of a neologism and its inclusion in the general vocabulary
of a language, as it is sufficient evidence of the institutionalisation of the word. As Freixa
and Torner (2020) point out, this is not equivalent to saying that all institutionalised neolo-
gisms belong to the lemma list of a dictionary. The argument works the other way around:
when conducting an accurate lexicographic procedure, there should not be unstable, very
infrequent lexical units in a dictionary. In this study, our purpose is not to study language
innovation but rather to contribute to the methods for updating dictionaries. For this rea-
son, we consider that the approach of comparing a corpus against a lemma list is in this
case appropriate.

2.2. New verbs in Spanish

Concerning, specifically, new verbs in Spanish, the most common mechanism is converting
already existing nouns or verbs (either from Spanish or from other languages) into new
verbs by affixation (Bohrn 2010, Fuentes et al. 2009). Regarding suffixation, very product-
ive suffixes take part in the process, especially -ificar, -ear and -izar (Adelstein and Kuguel
2008, Bohrn 2010, RAE and ASALE 2009), e.g., pesificar ‘to turn into pesos’ (from peso
and -ificar), googlear ‘to google’ (from Google and -ear, probably inspired by the English
verb to google) and oscarizar ‘to give an Oscar’ (from Oscar and -izar). They are also often
created by the addition of very productive prefixes to already existing verbs, such as re-
(revictimizar ‘re-victimize’, reconfinar ‘re-confine’), des- (desconfinar ‘de-confine’, desesca-
lar ‘de-escalate’), etc. Lavale-Ortiz (2016) also reports a substantial number of cases of
parasynthetic new verbs in her corpus, such as enturbantar ‘to put a turban’ (made with en-
and -ar added to the root) or aproblemar ‘to cause problems’ (made with a- and -ar added
to the root).
Compared to nouns and adjectives, the creation of new verbs in Spanish is a less com-
mon phenomenon (Fuentes et al. 2009, Ortega Martı́n 2001, Sanmartı́n 2010). And yet,
they contribute to a significant part of the processes of neology. Regarding affixation spe-
cifically, this mechanism has proven to be a current and constant way of creating new verbs
(Lavale-Ortiz 2016). We described some studies on the topic, but the fact is that there is still
plenty of room for research devoted specifically to Spanish new verbs. The situation does
not seem to be very different in studies in other languages. Nevertheless, we would like to
mention the work by Amore et al. (2018), who apply a semantic distributional model to
conduct a corpus analysis of Italian new verbs. Also, Ginebra and Rull (2010) provide a
quantitative analysis of new verbs in Catalan, and conclude that there is a tendency for
386 Ana Castro et al.

these units to be transitive in a higher percentage of those verbs included in the dictionary -
a tendency that, according to Bohrn (2010), can also be observed in Spanish verbs. Oliveira
(2020) deals with the process of creation of new verbs in Portuguese, which take place
mainly via suffixation, as it does in Spanish (and with equivalent suffixes: -ificar, -ear, -izar,
etc.).
L’Homme (2015) has argued that, in the case of terminology, verbs have been often

Downloaded from [Link] by OUP-USA Mirror user on 20 February 2025

neglected, because studies are more focused on knowledge representation, in which nouns
seem to be very well established. This argument may be applied to neology as well, as a
way to explain the relative lack of interest in verbs. However, verbs are a key part of speech
in lexicography, requiring complex corpus analyses and specific considerations on their lex-
icographic treatment that take into account their argument structure, polysemy, colloca-
tions, pedagogical aspects, etc. (Atkins et al. 1988, Battaner and Torner 2008, Boas 2001,
El Maarouf et al. 2014, Hanks 2013, L’Homme 2003, Marello 2010, among many others).
This complexity requires that studies in neology focus more on verbs and propose specific
methods for detection and representation of new verbs in dictionaries. For example, Atkins
et al. (1988), Hanks (2013) and Marello (2010) deal with the differences in the syntactic
structure of verbs (e.g. the causative/inchoative alternation), which has to be taken into ac-
count when conducting automatic neology detection of verbs (see section 3.2.1 for how we
dealt with the problem of pronouns used for creating inchoative structures in Spanish).
Thus, a method for neology detection should not only detect the new verbs, but also the
most frequent structures and other features in which the verbs are involved –e.g. alterna-
tions, prevalence of participle over other inflections, and collocations, among others. All
these aspects could be of great help for lexicographers in their quest to create new verb
entries.

3. Methodology
The general approach of our method was to extract verbs from corpora and compare them
with the lemma list of a Spanish dictionary. The algorithm for the automatic extraction of
Spanish verbs from corpora is based on a set of rules to detect verbal inflection. The basic
idea was to hand-code a list of verb endings and to observe if the same root appeared in the
corpus in combination with multiple endings. In the case that it did, we reconstructed an in-
finitive form and contrasted it against a reference dictionary. If no match was found, we
submitted the candidate to a battery of tests to distinguish between true neologisms and
spelling mistakes.

3.1. Materials
Our materials comprised a corpus, for which we used the EsTenTen (Kilgarriff and Renau
2013) and a dictionary, used to contrast the verbs extracted from the corpus. As for the dic-
tionary, we used the Diccionario de la lengua espa~ nola, DLE (RAE 2014), mainly for tech-
nical reasons. The lemma list of this dictionary can be downloaded via Enclave RAE
([Link] a website that offers an advanced interface for the DLE. The list of
verbs from the dictionary comprised 11,893 verb lemmas.
The EsTenTen corpus consists of approximately 1010 running words from randomly
downloaded web pages of Spanish-speaking countries. This corpus is already tokenised, but
we ignored the tagging and used only the word forms. We only kept in the list those form
New Verbs and Dictionaries 387

types occurring at least 5 times in the corpus, a threshold we considered convenient to min-
imise accidental occurrences. We used 25% of this corpus to develop and test our method,
and the rest to evaluate its performance. From the first part we obtained 1,054,411 differ-
ent word forms, while from the rest of the corpus we obtained 2,954,973 different word
forms. It should be clear that what we mean by this is form types and not form tokens, and
also that form types are not to be confused with lemma types.

Downloaded from [Link] by OUP-USA Mirror user on 20 February 2025

3.2. Methods
3.2.1. Hand-coding a list of Spanish verb endings
We obtained a list of verb endings from RAE and ASALE (2009), retaining only those with
a minimum length of three letters. We also included a list of prefixes and enclitics, i.e., pro-
nouns like me, te, se, lo, los, la, las, le, les or os, that can be attached to verbs, as in cómpra-
selo ‘buy him [something]’ / ‘buy it from him’ or peinarse ‘to comb oneself’, etc.
According to RAE and ASALE (2009), verb endings are organised in paradigms or verb
tenses and moods. Tenses are divided into simple and composites according to their lexical
structure, with one or two elements respectively. We only used the simple tenses. Verbs are
also classified in regular and irregular. Three regular conjugations exist, with infinitive
forms ending in -ar as in amar ‘to love’, -er, like in temer ‘to be afraid of’ and -ir, like in par-
tir ‘to depart’. In the three cases, the root of the verb is kept intact in all inflected forms,
with some exceptions due to accentuation. Irregular conjugations, in turn, correspond to all
other verbs that do not conform to any of these three models, and their root is altered in
their inflection. According to RAE and ASALE (2009), 90% of Spanish verbs correspond to
those ending in -ar, and practically all cases of word formation with suffixes are also of this
type (cases such as -ear, -izar and -ificar, as discussed in section 2).
We discarded word endings with fewer than three letters because they are error prone.
For instance, a word ending like -o as in yo amo ‘I love’, corresponding to the first person
singular of present tense indicative form of the verb amar ‘to love’, would coalesce with
other word forms also ending in -o but pertaining to other grammatical categories, like
adjectives (e.g. atrevido ‘cheeky’) or nouns (e.g. sentimiento ‘feeling’). Table 1 shows the
full list of 106 elements that were used.

3.2.2. Data processing

As already mentioned, we used 25% of the corpus to develop and test our methodology,
and left the rest of the corpus to evaluate the method’s performance. This is important be-
cause one cannot evaluate a method with the same corpus that was used in its development.
With the list of word forms extracted from the corpus and the hand-coded list of verb
endings, we implemented a script to extract all word forms matching a verb ending. Every
time a match was found, the script separated the root from verb ending and stored them in
a two-dimensional hash table. For illustration, consider Table 2, which shows some combi-
nations of verb roots and endings.
Here we depicted a root form like rape- as well as some typical verb endings, such as -
aba. Together, they form the verb form rapeaba ‘he/she was rapping’, which was found in
the corpus. If at least four different verb endings were observed attached to the same root,
as in rapeaba, rapeaban, rapeada or rapeadas, then a new verb entry was created. The num-
ber of four combinations is certainly arbitrary, but is based on empirical testing.
388 Ana Castro et al.

Table 1. List of regular verb endings and enclitic pronouns.

Regular verb endings

-aba -áramos -arı́amos -eré -iendo ieses

-abais -aran -arı́an -eréis -iera imos
-ábamos -arán -arı́as -eremos -ierais irá

Downloaded from [Link] by OUP-USA Mirror user on 20 February 2025

-aban -aras -aron -erı́a -iéramos -irán
-abas -arás -ase -erı́ais -ieran -irás
-ada -are -aseis -erı́amos -ieras -iré
-adas -aré -ásemos -erı́an -iere -iréis
-ado -areis -asen -erı́as -iereis -iremos
-ados -aréis -ases -ı́ais -iéremos -irı́a
-áis -aremos -aste -ı́amos -ieren -irı́ais
-amos -áremos -asteis -ı́an -ieres -irı́amos
-ara -aren -éis -ı́as -ieron -irı́an
-ará -ares -emos -ida -iese -irı́as
-arais -arı́a -erá -idas -ieseis -iste
-arı́ais -erán -ido -iésemos -isteis
-erás -idos -iesen -ndo

Enclitics

me te se nos le
les la lo las los

Table 2. Examples of verb root and endings from verbs rapear ‘to sing a rap’, retener ‘to retain’
and suprimir ‘to delete’.

Root Endings

rape- -aba, -aban, -ada, -adas, -ado, -ados

reten- -emos, -ida, -ido, -idos, -iendo, -éis, -ı́an
suprim- -amos, -ida, -idas, -ido, -idos, -iendo, -iera, -ieran, -ieron, -iese, -iesen, -irán, -iré, -irı́a, -ı́an

Of course, what we needed to obtain was a list of verb lemmas and not a list of inflected
forms. Therefore, once this part of the procedure was finished, the next step was to find the
infinitive form of the verb, which in Spanish can only end in -ar, -er or -ir, as explained ear-
lier. The way to find the infinitive form was to attach the newly found root to each of these
endings and select the most frequent root-ending combination. For instance, in the case of a
verb root like reten-, shown in Table 2, the algorithm would test lemma candidates with
-ar: *retenar; with -er: retener; and with -ir: *retenir, but retener, the correct lemma, is the
one that appears most frequently in the corpus. Table 3 shows an example of the result of
this process, with the lemma, its frequency, root and endings. The DLE column states
whether the lemma is found in the dictionary.
If the newly found verb lemma matched the list obtained from the reference dictionary
(the DLE), then there was no doubt we were looking at a legitimate Spanish verb.
New Verbs and Dictionaries 389

Table 3. Examples of the reconstruction of the lemma forms of the verbs and their comparison
against the dictionary lemma list. Only rapear ‘to sing a rap’ was not present in the dictionary.

Lemma Frequency Root DLE Endings

rapear 206 rape- 0 -aba, -aban, -ada, -adas, -ado, -ados

retener 26390 reten- 1 -emos, -ida, -ido, -idos, -iendo, -éis, -ı́an

Downloaded from [Link] by OUP-USA Mirror user on 20 February 2025

suprimir 32436 suprim- 1 -amos, -ida, -idas, -ido, -idos, -iendo, -iera,
-ieran, -ieron, -iese, -iesen, -irán, -iré, -irı́a, -ı́an

Otherwise, it could be a neologism candidate, an orthographic error or a word form per-

taining to a different grammatical category. The next step of the procedure was then to ana-
lyse those candidates with value 0 in the DLE column in order to separate true neologisms
from erroneous forms.

3.2.3. Error reducing strategies

In this part of the procedure we submitted all the verb candidates that did not match the
reference dictionary to a series of tests to further separate true neologism candidates from
misspellings and errors of other nature. Only if a candidate survived the full battery of tests
was it considered a neologism candidate.

Prefixes. We used a list of prefixes in this process. If a prefix was detected in an infinitive,
and the verb was listed in the dictionary without the prefix, then the verb was considered a
neologism candidate, and then no other test was necessary. We adopted this strategy be-
cause one way for creating derivative neologisms is by adding prefixes to an existing word,
such as in recalendarizar ‘to re-calendarise’, from the existing verb calendarizar ‘calenda-
rise’ and the prefix re-. We extracted the list of prefixes from DLE, also by downloading it
from Enclave RAE. The full list is shown in Table 4.
Table 5, in turn, shows some examples of verbs not listed in the dictionary but bearing
one of the prefixes. As can be seen in the Frequency columns, the lemma without the prefix,
which does appear in the dictionary, is always far more frequent than the form with the
prefix.

Minimum length. Those candidates that did not match any prefix were subjected to the rest
of the battery tests. If a given candidate did not pass all these tests, it was discarded. The
first of such eliminating tests was to measure the length of the lemma in characters. As we
considered word endings of at least three letters, we discarded in this step all lemmas with
less than four characters. As explained by RAE and ASALE (2009), Spanish verbs tend not
to be too short because they carry at least four morphological segments: the root, the the-
matic vowel, the tense/mood and the person/number. Moreover, verbs with affixation and
enclitic pronouns are not infrequent, thus it is difficult to find verbs that can compress all
this morphological information in less than four letters. Examples of verb lemmas that
were eliminated in this step are *bar, *kir and *oir. None of these appear as verb lemmas
in the dictionary. The first form coincides with the noun bar ‘bar’, the second is not a word
and the last one resembles the Spanish verb oı́r ‘to hear’ but lacking the diacritical mark
due to misspelling.
390 Ana Castro et al.

Table 4. List of prefixes used in the prefix detection process.

ana-, anti-, auto-, cata-, cis-, co-, con-, contra-, cuasi-, de-, des-, di-, dia-, dis-, em-, en-, entre-, es-, ex-,
extra-, geo-, hiper-, hipo-, in-, infra-, inter-, intra-, para-, per-, peri-, pluri-, pos-, post-, pre-, pro-,
re-, requete-, res-, rete-, semi-, sin-, so-, sobre-, son-, sub-, super-, tele-, tera-, trans-, tras- ,ultra-

Downloaded from [Link] by OUP-USA Mirror user on 20 February 2025

Table 5. Examples of verbs with their prefixes.

Candidate Frequency Prefix Base verb Frequency

autoevaluar 476 auto- evaluar 350,844

desjerarquizar 66 des- jerarquizar 9,668
recalendarizar 188 re- calendarizar 655

Reinsertion of first letter. Our second attempt to separate neologism candidates from spell-
ing mistakes was to reinsert a first letter, as we observed this is a very common occurrence.
Table 6 shows a few examples of false verb candidates that were obtained from the corpus,
the forms *aber, *econocer and *guantar. They certainly meet all the morphological criteria
described so far (e.g. they present a root which combines with different word endings) and
yet, they are not neologisms but spelling errors.
The idea was thus to try to reinsert an initial letter (such as a, e, i, o, u, h, b, c, d, f, g, j,
k, l, m, n, p, q, qu, r, s, t, v) and then check again against the reference dictionary. If a
match was found this time, then the candidate was discarded. Otherwise, it was submitted
to the following tests. For example, in the cases of Table 6, by adding an initial letter, the
system could match *aber, a mistake, with the real verb haber ‘to have’, which is not a
neologism candidate and was thus discarded.

Letter substitution. The remaining candidates were then screened for the detection of other
forms of typographic errors. A typical case of misspelling is the substitution of one letter
for another. For instance, a very common orthographic mistake in Spanish is the confusion
between b and v (e.g. it was common to find the verb avanzar ‘to move forward’ misspelled
as *abanzar). In order to detect and discard this type of error, we applied a special case of a
spell-checking algorithm consisting of a combination of letter substitution rules based on
Spanish orthography (RAE and ASALE 2010). Table 7 shows the entire catalogue of rules.
It should be read from left to right: for example, the letter or sequence of letters on the left
is replaced by the letter or sequence on the right.
If a letter substitution could be made in some verb candidate, the new form was then
compared with those in the dictionary. If a match was found, then the candidate was dis-
carded. Table 8 shows some examples of cases in which this happens.
Notice a very frequent case which is the incorrect addition of the letter h, a very com-
mon spelling mistake in Spanish because h is silent. The system eliminated this letter, result-
ing in the verb abrir ‘to open’, which is not a neologism. The table also shows the frequency
of the correctly spelled form in the corpus. If no match was found with these substitution
rules, the candidate was then submitted to the rest of the tests.
New Verbs and Dictionaries 391

Table 6. Examples of false candidates that were eliminated by inserting a first letter.

Candidate Reinserted letter Correct lemma

*aber h- haber
*econocer r- reconocer
*guantar a- aguantar

Downloaded from [Link] by OUP-USA Mirror user on 20 February 2025

Table 7. List of pairs of characters used by the substitution rules.

s!c n!m nb ! mb ps ! s n ! pn ó ! o
c!s s!z mb ! nb s ! ps pt ! t o ! ó
b!v z!s mb ! nv ii ! i t ! pt ú ! u
v!b z!c nv ! mb i ! ii bs ! s u ! ú
j!g y ! ll ep ! pe de ! des s ! bs u ! ü
g!j ll ! l pe ! ep di ! dis ns ! s ü ! u
c!k l ! ll p ! pe il ! ll s ! ns i!e
k!c ll ! y pe ! p li ! ll st ! s e!i
k ! qu r ! rr nf ! mf in ! ins s ! st a!e
qu ! k rr ! r x!j in ! inc á ! a e!a
n~ ! n h!- j!x oo ! o a ! á
n ! n~ y!i ee ! e o ! oo é ! e
N~ ! n~ i!y e ! ee gn ! n e ! é
ny ! n~ mp ! np x!s n ! gn ı́ ! i
m!n np ! mp s!x pn ! n i ! ı́

Table 8. Examples of misspelled verbs that were detected using substitution rules.

Candidate Correct lemma Frequency

*abanzar avanzar 319,185

*blokear bloquear 42,481
*habrir abrir 375,674

Orthographic similarity coefficient. We placed the orthographic similarity measures after

the previously described rules because they are more computationally expensive. We com-
bined two different similarity coefficients in order to find the orthographic similarity be-
tween a verb candidate and each verb in the dictionary. The first one is the Jaccard index,
commonly used in natural language processing (Equation 1). For this we used two-letter
sequences as features.

jA \ Bj
JðA; BÞ ¼ (1)
jA [ Bj

Again because of the computational cost that this comparison entails, we limited the ap-
plication of such coefficient only to those verbs in the dictionary that began with the same
letter as the candidate (except in the case of h, given that this is, as already explained, a
392 Ana Castro et al.

Table 9. Examples of verb candidates that matched and entry in the dictionary using an ortho-
graphic similarity coefficient.

Candidate Similar lemma(s) and their frequency in the corpus

*abadonar abandonar (150,342)

*contatar contar (845,098); contratar (141,316); contentar (2,396);

Downloaded from [Link] by OUP-USA Mirror user on 20 February 2025

contactar (46,139); constatar (75,305)

very common occurrence), that were similar in length and which had a minimum frequency
of 15 occurrences in the corpus. Table 9 shows some examples of verb candidates that were
matched with verbs in the dictionary using the Jaccard similarity coefficient.
There we can see cases like *abadonar and *contatar, each of which found a match
with at least one entry in the dictionary using this coefficient: *abadonar matched abando-
nar ‘to abandon’, and contatar matched five possible verbs in the DLE. As in previous cases,
if at least one match with the dictionary was found, and such match was above the fre-
quency threshold, then the candidate was discarded. Otherwise, it continued to be sub-
jected to further tests.
To complement the previous similarity index, a new orthographic similarity coefficient
was implemented. As in the previous case, this new index takes two words as arguments
(the verb candidate and each of the entries in the dictionary) and proceeds to compare them
letter by letter. In a first loop, it compares the first letter of one with the first letter of the
other. If the letter was the same in both cases, a variable we call match was increased by
one, and then the algorithm moved to the next letter in both words. If, however, the letter
was not the same, then the first loop stopped and a second began, in a very similar fashion,
only that now instead of the first letter, it compared the last letter in both words. Again, if
the letter was the same, then the match variable was increased by one. Then, the pointer is
moved to the previous letter and the comparison is made again. The process finished when
another mismatch was found, now from the opposite direction. The result of the compari-
son was thus the quotient between the number of matches and the character length of the
larger word (2). As with the previous case, comparisons resulting in a value over an empir-
ically defined threshold were considered positive matches.

matchðA; BÞ
ortSimðA; BÞ ¼   (2)
max lengthðAÞ; lengthðBÞ

Table 10 shows two examples of verb candidates, *abodar and *colacar, that found a
match to some entries in the dictionary. The same criterion of frequency in the corpus was
used to avoid comparisons with very low frequency verbs of the dictionary.

Distributional profiling. The last filter we applied consisted of a distributional

similarity index, based on the intuition that a true verb will show a characteristic distribu-
tional pattern. This was by far the most computationally expensive of all, as it involved the
extraction from the corpus of contexts of occurrence of the verbs. For this reason, it was
left for the end, because only a reduced number of verb candidates passed the previous
tests.
New Verbs and Dictionaries 393

Table 10. Examples of matches using the second similarity coefficient.

Candidate Matches with dictionary

*abodar abocar (1,289); abogar (4,875); acodar (113); abordar (165,173)

*colacar colar (7,177); colocar (234,393)

Downloaded from [Link] by OUP-USA Mirror user on 20 February 2025

We defined what would be the typical profile of co-occurrence of a Spanish verb in its
infinitive form by examining contexts of occurrence in the corpus of genuine verbs. We
mainly resorted to verbal periphrases consisting of an auxiliary verb plus an infinitive form
(Gómez Torrego 1999). We manually compiled a list of 50 mainly modal periphrases, such
as tengo que ‘I have to’ or hay que ‘it is necessary to’; aspectual periphrases like comenzó a/
comienza a or empiece a/empieza a ‘he/she start/started to’, as well as other types, such as
se atrevió a ‘he/she dared to’; sirve para ‘it is used for’. We also included other expressions
that frequently precede infinitive forms, such as el arte de ‘the art of’; ganas de ‘lust for’;
máquinas de ‘machines for’; un aparato para ‘a device for’; etc. Using this list, the last test-
ing algorithm extracted contexts of occurrence of the candidate from the corpus and if it
found that there were no matches of any of these expressions, then it eliminated the candi-
date. With this method, the algorithm was able to eliminate false candidates such as *acuer,
*brujer or *shakespear, that is, forms that do appear in the corpus with some frequency,
but not with the predictors we defined for the category of verbs.

4. Results and evaluation

From the total of 2,954,973 form types with frequency > 5 in the corpus (i.e. the part that
we used for evaluation), we obtained a list of 10,034 verb candidates. From these, 5,685
(56,65%) were already listed in the DLE and 4,349 (43,34%) were submitted to the battery
of tests to discard orthographic errors. From this list, only 774 forms (17,79%) survived all
the filters and were therefore considered neologism candidates.
Considering the list of verbs that did not match the dictionary, the vast majority was dis-
carded by some filter, as shown in Table 11. Only the candidates with the categories of
‘prefixes’ and ‘survivors’ were considered true neologisms by the algorithm. The first cat-
egory comprises those with a prefix and other verbs that passed all filters. The table also
presents the results of manual evaluation of the performance of each filter or category.
The best performing filter was the ‘Minimum length’. This is also the less computation-
ally expensive one, as it only counts the number of letters of the candidates. The
‘Reinsertion of first letter’ and ‘Letter substitution’ filters also had good performance, with
98% and 99% precision respectively. The ‘Orthographic similarity 1’ filter was less reli-
able, with 89% precision. It resulted in errors such as discarding genuine neologisms like
bloguear ‘to blog’ because of its similarity with the verb bloquear ‘to block’. The
‘Distributional profiling’ filter was also less effective, with 90% precision. It incorrectly
eliminated valid neologisms such as flashear ‘to be shocked’ because they do not show the
typical distribution of normal verbs, or perhaps because we collected fewer periphrases
than were needed in the case of these less frequent verbs. One way to mitigate this problem
would be to focus on the evolution of the frequency curve in the timeline, as done by Nazar
394 Ana Castro et al.

Table 11. Results of the classification and application of filters.

Filter/Selection Tag Total % correct % incorrect

Minimum length false 21 100% 0%

Reinsertion of first letter false 231 98,2% 1,7%
Letter substitution false 873 99,5% 0,4%

Downloaded from [Link] by OUP-USA Mirror user on 20 February 2025

Orthographic similarity 1 false 950 89,5% 10,4%
Orthographic similarity 2 false 568 97,3% 2,6%
Distributional profiling false 932 90,4% 9,5%
Prefixes true 398 96,7% 3,2%
Survivors true 376 79,5% 20,47%

and Vidal (2010). That is, if a word like flashear or bloguear shows a sharp rise in its fre-
quency of use, then this can be taken as a strong indication of neology. We leave this possi-
bility, however, for future work.
Table 12 presents the results of the overall evaluation, which informs how probable it is
that the correct true/false tag will be assigned to a given candidate.
We manually evaluated the total 4,349 verbs submitted to the battery of tests. We define
precision and recall in terms of true positives (tp), the case of a true neologism that is correctly
detected by the algorithm; false positives (fp), the case of a candidate that is not a neologism
but is anyway promoted by the algorithm; and finally false negatives (fn), the case of a neolo-
gism that was not detected by the algorithm (i.e., a true neologism that was discarded by
some filter). Using these three variables we calculated precision (3), which measures how like-
ly it is that a promoted candidate will be a true neologism, and recall (4), which measures the
exhaustivity of the method, i.e., how likely it is that a true neologism will be detected.
tp
precision ¼ (3)
ðtp þ fpÞ

tp
recall ¼ (4)
ðtp þ fnÞ

The overall precision of the detection of new verbs was 0.88 and recall was 0.76. Thus,
close to 88% of the final list of selected verbs are valid neologism candidates, resulting
from different word formation mechanisms.
With respect to the errors made by the algorithm, we found that they can be classified in
three groups: a) errors due to failure to detect missing accents (e.g. *sonreir was labelled a
neologism whereas it is a misspelling of sonreı́r ‘to smile’) and other typos (e.g. cosntruir
for construir ‘to build’); b) errors due to failure to detect verbs from other languages (e.g.
aproveitar, a Galician verb, was considered a valid Spanish verb), and c) errors due to in-
correct interpretation of a form as a verb, with the subsequent creation of a nonexistent
verb (e.g. *universar, *umbilicar).
Table 13 shows some examples of the list of neologism candidates, and the type of neol-
ogy mechanism according to Cabré (2015).
Some of these verbs are frequently used in everyday language, and the fact that they
have not been incorporated into the dictionary might be due to different factors, as
New Verbs and Dictionaries 395

Table 12. Proportion of correct / incorrect results in the detection of neologisms

Tag Correct Incorrect Total

<TRUE> 684 88,37% 90 11,62% 774

<FALSE> 3,364 94,09% 211 5,90% 3,575

Downloaded from [Link] by OUP-USA Mirror user on 20 February 2025

Table 13. Examples of new verbs detected with the proposed method.

Type of neology mechanism New verbs English translation

(Cabré 2015)

Prefixation (adding a prefix to coorganizar, desactualizar, ‘to co-organise’, ‘to make some-
an existing verb) intercomunicar, precalcular, thing out-of-date’, ‘to inter-
reasignar, etc. communicate’, ‘to pre-calcu-
late’, ‘to reassign’, etc.
Suffixation (adding a verb suffix aperturar, chocolatear, coopera- ‘to open’, ‘to add chocolate’, ‘to
to an existing noun) tivizar, ficcionar, gelificar, etc. create an association’, ‘to
turn into fiction’, ‘to turn
into gel’, etc.
Parasynthesis (adding a prefix adinerar, desvirtualizar, enrutar, ‘to enrich’, ‘to devirtualise’, ‘to
and suffix to an existing etc. put in a route’, etc.
noun)
Loans (adding a Spanish verb customizar, draftear, loguear, ‘to customise’, ‘to make a
suffix to a loan) resetear, spoilear, etc. draft’, ‘to log’, ‘to reset’, ‘to
spoil’, etc.

discussed in section 2. Freixa and Torner (2020), among others, point out that there can be
a mismatch between the real institutionalisation of a word and the fact that the word is pre-
sent in the dictionary. If we observe the 100 most frequent candidates in our results, most
of them could be included in the dictionary, as they are frequent enough (they appear from
169 to 40,409 times in the corpus), they are morphologically correct and they are wide-
spread across Spanish-speaking countries. This is the case of vivenciar ‘to experiment’,
direccionar ‘to address’, googlear ‘to google’, suplementar ‘to supplement’, referenciar ‘to
refer’, resetear ‘to reset’, customizar ‘to customise’, etc.
It is not the goal of this study to try to find out why these verbs were not included in the
dictionary used to test the method. However, we can say that the resulting list of candi-
dates, with very few errors, would allow a lexicographic team to work with empirically-
based data and obtain valid information for potential new verb entries for dictionary
updating. Once the team had this information, it could make informed decisions about
which of these lemmas should be included and why, instead of using non-systematic ways
of including new words in dictionaries (e.g. following social trends or through personal
findings in newspapers). As already observed, there are many reasons why a new word
should or should not be included in a dictionary. Some of them may be based on the charac-
teristics of the dictionary project, combined with the characteristics of the potential new
lemma. For example, in the examples in Table 13, most of the verbs could be included in a
396 Ana Castro et al.

general descriptive dictionary, while aperturar, enrutar, customizar, draftear, loguear, rese-
tear or spoilear, being unnecessary loans, would not be adequate for a prescriptive
dictionary.
Finally, an interesting and unexpected result was the great number of verb entries from
the DLE that were not present in the large corpus we used in our experiments. From the
total of 11,893 verb lemmas of the dictionary, we found 6,208 that did not appear with the

Downloaded from [Link] by OUP-USA Mirror user on 20 February 2025

minimum frequency. This represents a very large proportion of verbs listed in the dictionary
that are very infrequent or non-existent, and our intuition is that this is because they have
fallen into desuetude (Algeo 1993). For that reason, we believe that the study of desuetude
is complementary to the study of neology for updating dictionary lemma lists, but we will
leave the exploration of this topic for future work.

5. Conclusions
In this paper, we proposed a method for the detection of new verbs, taking into account the
difficulty of part of speech tagging and the less-than-perfect performance of current taggers.
The proposed method, which is relatively simple, performs well in the detection of verbs in
Spanish, and differentiates verbs that are already present in a dictionary such as DLE from
the ones that the dictionary did not register. We also highlight the effectiveness of the
method at reducing the noise of spelling mistakes in the corpus. The typification of the
most frequent errors offers much help in reducing the burden of the otherwise manual revi-
sion of the data. Another characteristic of the proposed method is that it is based mainly on
morphological features to reduce the error rate, avoiding the use of semantic features,
which are harder to systematise.
The proposed algorithm can be easily incorporated into a workflow for creating or
updating a lexicographic database, so it can be used regularly to detect new verbs and con-
tribute to a more dynamic and updated dictionary. The vision of the electronic dictionary
as ‘up-to-date’ and a ‘dynamic repository of knowledge’ was already presented by De
Schryver (2003: 159), but almost 20 years later there is still plenty of work to do in this re-
spect. The automatic interrogation of a corpus to find verbs not present in a list of head-
words is a reliable method for tracing new verbs and adding them to a list of possible future
verb entries to the dictionary, aiding the lexicographer in her/his decision-making process.
Information such as the first appearance in the corpus, as well as the frequency and disper-
sion of the word, can also be included in the entry, in order to orient the user to the neologi-
cal nature of a specific unit. Depending on the type of dictionary and the conditions of the
lexicographic projects (such as technical, methodological, economical, etc.), a dictionary
for the Internet can be updated regularly with information about the frequency of the
word, taking data from corpora of different periods of times, comparing them and showing
the results in each entry. The general approach is not new (Renouf 1993), but it has not yet
been fully incorporated into the lexicographic procedures of Spanish dictionaries, nor is it a
standard procedure in other lexicographic traditions. As future work, we would like to add
supplementary information about the new verbs to the method. For example, the method
should be able to recognise whether a verb is only used in its participle form, or if it is used
with a specific alternation, among other features that could help lexicographers to add ac-
curate information to the dictionary entry and having first clues about its macrostructure.
New Verbs and Dictionaries 397

References

A. Dictionaries
Real Academia Espa~ nola. 2014. Diccionario de la lengua espa~ nola (Twenty-third edition.).
Madrid: Espasa.

Downloaded from [Link] by OUP-USA Mirror user on 20 February 2025

Real Academia Espa~
nola. Online. Enclave RAE. Accessed on 13 June 2020. [Link]

B. Other literature
Abel, A. and E. Stemle. 2018. ‘On the Detection of Neologism Candidates as Basis for Language
Observation and Lexicographic Endeavours: The STyrLogism Project’ In Cibej,  J., V. Gorjanc,
I. Kosem and S. Krek (eds), Proceedings of the XVIII EURALEX International Congress,
EURALEX 2018, Ljubljana, Slovenia, July 17-21, 2018. Ljubljana: University of Ljubljana,
535–544.
Adelstein, A. and I. Kuguel. 2008. De salariazo a corralito, de carapintada a blog. Nuevas palabras
en veinticinco a~nos de democracia. Los Polvorines: Universidad Nacional de General Sarmiento.
Algeo, J. 1993. ‘Desuetude among New English Words.’ International Journal of Lexicography
6.4: 281–293.
Alvar Ezquerra, M. 2007. ‘El neologismo espa~ nol actual’ In Luque Toro, L. (ed), Léxico espa~nol
actual. Actas del I Congreso Internacional de Léxico Espa~ nol Actual, Venice-Treviso, March
14-15, 2005. Venezia: Libreria Editrice Cafoscarina, 11–35.
Amore, M., S. McGregor and E. Jezek. 2018. ‘Distributional Analysis of Verbal Neologisms: Task
Definition and Dataset Construction’ In Cabrio, E., A. Mazzei and F. Tamburini (eds),
Proceedings of the Fifth Italian Conference on Computational Linguistics, CLiC-it 2018,
Torino, Italy. December 10-12, 2018. [Link], n. p.
Atkins, B. S., J. Kegl and B. Levin. 1988. ‘Anatomy of a Verb Entry: From Linguistic Theory to
Lexicographic Practice.’ International Journal of Lexicography 1.2: 84–126.
Atkins, B. S., and M. Rundell. 2008. The Oxford Guide to Practical Lexicography. Oxford:
Oxford University Press.
Baayen, R. H. and A. Renouf. 1996. ‘Chronicling the Times: Productive Lexical Innovations in an
English Newspaper.’ Language 72.1: 69–96.
Battaner, P. and S. Torner. 2008. ‘La polisemia verbal que muestra la lexicografı́a’ In Azorı́n D.,
B. Alvarado, J. Climent, M. I. Guardiola, R. Lavale Ortiz, C. Marimón Llorca, J. J. Joaquı́n
Martı́nez, X. A. Padilla, H. Provencio, I. Santamarı́a-Pérez, L. Timofeeva and E. Toro (eds).
Actas del II Congreso Internacional de Lexicografı́a Hispánica: el diccionario como puente
entre las lenguas y culturas del mundo. Alicante: Universidad de Alicante, 204–246.
Boas, F. 2001. ‘Frame Semantics as a Framework for Describing Polysemy and Syntactic
Structures of English and German Motion Verbs in Contrastive Computational Lexicography.’
In Rayson, P., A. Wilson, T. McEnery, A. Hardie and S. Khoja (eds), Proceedings of the Corpus
Linguistics 2001 Conference. Lancaster: Lancaster University, 64–73.
Bohrn, A. 2010. ‘La neologı́a verbal en el espa~ nol rioplatense.’ In Cabré, T., O. Domènech, R.
Estopà, J. Freixa and M. Lorente (eds), Actes del I Congrés Internacional de Neologia de les
Llengües Romàniques Barcelona: IULA, 917–927.
Cabré, M. T. 1999. Terminology. Theory, Methods and Applications. Amsterdam: John
Benjamins.
Cabré, M. T. and R. Estopà. 2009. ‘Trabajar en neologı́a con un entorno integrado en lı́nea: la
estación de trabajo OBNEO.’ Revista de Investigación Lingüı́stica 12: 17–38.
398 Ana Castro et al.

Cabré, M. T. and R. Nazar. 2012. ‘Towards a New Approach to the Study of Neology.’
Neologica 6: 63–80.
Cabré, M. T. 2015. ‘Bases para una teorı́a de los neologismos léxicos: primeras reflexiones’ In
Alves, I. and E. Sim~ oes (eds), Neologia das lı́nguas românicas. S~ ao Paulo: CAPES, Humanitas,
79–107.
Ca~nete, P., S. Fernández-Silva and B. Villena. 2019. ‘Estudio de los neologismos terminológicos
difundidos en el diario El Paı́s y su inclusión en el diccionario.’ Cı́rculo de Lingüı́stica Aplicada

Downloaded from [Link] by OUP-USA Mirror user on 20 February 2025

a la Comunicación 80: 135–158.
Cartier, E. and J.-F. Sablayrolles. 2008. ‘Néologismes, dictionnaires et informatique.’ Cahiers de
Lexicologie 2008-2.93: 175–192.
Cartier, E. 2016. ‘Neoveille, système de repérage et de suivi des néologismes en sept langues.’
Neologica 10: 101–131.
Cook, C. 2010. Exploiting Linguistic Knowledge to Infer Properties of Neologisms. Ph.D. Thesis,
University of Toronto.
Costin-Gabriel, C. and T. E. Rebedea. 2014. ‘Archaisms and Neologisms Identification in Texts’
In 2014 RoEduNet Conference 13th Edition: Networking in Education and Research Joint
Event RENAM 8th Conference. IEEE, 1–6.
De Schryver, G.-M. 2003. ‘Lexicographers’ Dreams in the Electronic-Dictionary Age’.
International Journal of Lexicography 16.2: 143–199.
El Maarouf, I., V. Baisa, J. Bradbury and P. Hanks. 2014. ‘Disambiguating Verbs by Collocation:
Corpus Lexicography Meets Natural Language Processing.’ In Calzolari, N., K. Choukri, T.
Declerck, H. Loftsson, B. Maegaard, J. Mariani, A. Moreno, J. Odijk and S. Piperidis (eds),
Proceedings of the Ninth International Conference on Language Resources and Evaluation
(LREC’14). Reykjavik: European Language Resources Association, 1001–1006.
Freixa, J. and S. Torner. 2020. ‘Beyond Frequency: On the Dictionarization of New Words in
Spanish.’ Dictionaries: Journal of the Dictionary Society of North America 41.1: 131–153.
Fuentes, M., C. Gerding, A. CPecchi, G. A. Kotz and P. Ca~ nete. 2009. ‘Neologı́a léxica: Reflejo de
la vitalidad del espa~
nol de Chile.’ Revista de Lingüı́stica Teórica y Aplicada 47.1: 103–124.
Ginebra, J. and X. Rull. 2010. ‘Tendències en el règim sintàctic dels neologismes verbals: una
aproximació’ In Cabré, M. T., O. Domènech, R. Estopà, J. Freixa and M. Lorente (eds), Actes
del I Congrés Internacional de Neologia de les Llengües Romàniques. Barcelona: IULA,
377–390.
Gómez Torrego, L. 1999. ‘Los verbos auxiliares. Las perı́frasis verbales de infinitivo’ In Bosque, I.
and V. Demonte (eds), Gramática descriptiva de la lengua espa~ nola. Madrid: Espasa:
3323–3388.
Hanks, P. 2013. Lexical Analysis: Norms and Exploitations. Cambridge, MA: MIT Press.
Janssen, M. 2005. ‘Neo Track: Semiautomatic Neologism Detection.’ Communication presented
in XXI Encontro Nacional da Asociaç~ ao Portuguesa de Lingüı́stica. Lisbon.
Janssen, M. 2009. ‘Detección de neologismos: una perspectiva computacional.’ Debate
Terminológico 5: 68–75.
Kerremans, D. and J. Prokic. 2018. ‘Mining the Web for New Words: Semi-Automatic Neologism
Identification with the NeoCrawler.’ Anglia 136.2: 239–268.
Kilgarriff, A. and I. Renau. 2013. ‘esTenTen, a Vast Web Corpus of Peninsular and American
Spanish.’ Procedia-Social and Behavioral Sciences 95: 12–19.
Klosa-Kückelhaus, A. and H. Lüngen. 2018. ‘New German Words: Detection and Description’ In

Cibej, J., V. Gorjanc, I. Kosem and S. Krek (eds), Proceedings of the XVIII EURALEX
International Congress, EURALEX 2018, Ljubljana, Slovenia, July 17-21, 2018. Ljubljana:
University of Ljubljana, 559–569.
New Verbs and Dictionaries 399

Klosa-Kückelhaus, A. and I. Kernerman. 2020. ‘Global Viewpoints on Lexicography and

Neologisms: An Introduction.’ Dictionaries: Journal of the Dictionary Society of North
America 41.1: 1–9.
Langemets, M., J. Kallas, K. Norak and I. Hein. 2020. ‘New Estonian Words and Senses:
Detection and Description.’ Dictionaries: Journal of the Dictionary Society of North America
41.1: 69–82.
Lavale-Ortiz, R. 2016. ‘Hacia una revisión del concepto de neologismo aplicado a los verbos

Downloaded from [Link] by OUP-USA Mirror user on 20 February 2025

denominales aparecidos en la prensa espa~ nola.’ Revista Espa~ nola de Lingüı́stica Aplicada 29.1:
165–190.
Lemnitzer, L. 2010. Die Wortwarte. Wörter von heute und morgen. Eine Sammlung von
Neologismen. Accessed on 11 June 2020. [Link]
L’Homme, M. C. 2003. ‘Capturing the Lexical Structure in Special Subject Fields with Verbs and
Verbal Derivatives. A Model for Specialized Lexicography.’ International Journal of
Lexicography 16.4, 403–422.
L’Homme, M. C. 2015. ‘Predicative Lexical Units in Terminology.’ In Gala, N., R. Rapp and G.
Bel-Enguix (eds), Language Production, Cognition, and the Lexicon. Berlin: Springer, 75–93.
Marello, C. 2010. ‘Verbos con construcciones tanto transitivas como intransitivas y/o pronomi-
nales en los diccionarios monolingües y bilingües italianos y espa~
noles.’ In Castillo Carballo, M.
A. and J. M. Garcı́a Platero (coords), La lexicografı́a en su dimensión teórica. Málaga:
Universidad de Málaga, 411–433.
Nam, K., S. Lee and H. Jung. 2020. ‘The Korean Neologism Investigation Project: Current Status.’
Dictionaries: Journal of the Dictionary Society of North America 41.1: 105–129.
Nazar, R. and V. Vidal. 2010. Aproximación Cuantitativa a la Neologı́a. In Cabré, T., O.
Domènech, R. Estopà, J. Freixa and M. Lorente (eds), Actas del Congreso Internacional de
Neologı́a en las lenguas románicas (CINEO). Barcelona: IULA, 867–880.
Oliveira, B. 2010. ‘Verbos neologicos em un corpus jornalı́stico do português brasileiro.’ In
Cabré, T., O. Domènech, R. Estopà, J. Freixa and M. Lorente (eds), Actes del I Congrés
Internacional de Neologia de les Llengües Romàniques Barcelona: IULA, 635–646.
Ortega Martı́n, M. P. 2001. ‘Neologı́a y prensa: un binomio eficaz’. Espéculo. Revista de Estudios
Literarios 18.
Parra Escartı́n, C. and Alonso, H. Martı́nez 2015. Choosing a Spanish Part-of-Speech Tagger for a
Lexically Sensitive Task. Procesamiento del Lenguaje Natural. 54: 29–36.
Real Academia Espa~ nola (RAE) and Asociación de Academias de la Lengua Espa~ nola (ASALE).
2009. Nueva gramática de la lengua espa~ nola. Madrid: Espasa.
Real Academia Espa~ nola (RAE) and Asociación de Academias de la Lengua Espa~ nola (ASALE).
2010. Ortografı́a de la lengua espa~nola. Madrid: Espasa.
Renouf, A. 1993. ‘Sticking to the Text: A Corpus Linguist’s View of Language.’ ASLIB
Proceedings 45.5: 131–136.
Renouf, A. 2016. ‘Big Data and its Consequences for Neology.’ Neologica 10: 15–37.
Rey, A. 1976. ‘Néologisme, un pseudo-concept?’ Cahiers de Lexicologie 28.1: 3–17.
Sanmartı́n, J. 2010. ‘El neologismo castellano en un corpus de prensa editada en la comunidad
valenciana: >un hecho diferencial?’ In Cabré, M. T., O. Domènech, R. Estopà, J. Freixa and M.
Lorente (eds), Actes del I Congrés Internacional de Neologia de les Llengües Romàniques.
Barcelona. IULA, 693–709.
Vivaldi, J. 2000. ‘SEXTAN: prototip d’un sistema d’extracció de neologisms’ In Cabré, M. T., J.
Freixa and E. Solé (eds), La Neologia en el tombant de segle. Barcelona: IULA, 165–177.