The purpose of the correction annotation is to make the learner corpus searchable for different types of deviations from a standard language norm. The annotation of the learner texts according to the correction-taxonomy of the SweLL project is hence an important step in making the learner language assembled in the corpus analyzable for research and educational purposes.
This document has three interconnected purposes:
It documents choices made by the SweLL project group regarding the principles of the correction annotation.
It provides guidance for annotators on how to apply the SweLL annotation taxonomy in order to ensure the greatest possible inter-annotator agreement.
It provides the essential material for a thorough description of the correction annotation for future users of the Swell corpus. This projected description will accompany the published corpus.
The rest of this document is divided into two main sections: one section with general directions for the correction annotation, and one section on the SweLL correction annotation taxonomy.
By correction we mean a difference between the original learner text and the normalized version of the text. The correction annotation is thus a categorization of such differences.
This means that the correction annotation only indirectly indicates properties of the source text. What is directly indicated is the relationship between the original version and the normalized version of the text. The correction annotation is thus highly dependent on the preceding normalization.
While the purpose of the correction annotation is to make the texts searchable for deviations from a standard language norm, such deviations are only possible to categorize on the basis of assumptions of the learner’s intended content, along with judgements of the acceptability and suitability of standard language expressions communicating that content. This means that any annotation of “deviations”, “errors” etc. in learner texts is actually an annotation of a relationship between (a segment of) the analyzed text and an assumed standard language version of the text (segment). By the choice of the term correction annotation (rather than, for instance, error annotation) the SweLL project emphasizes these conditions of learner language analysis. The normalized text versions make explicit this necessary assumption about the standard version of the text to which the original text is related. (The principles underlying the normalization are described in the normalization guidelines.)
A consequence of the fundamental principle of annotating corrections, understood as differences between the original text and one specific interpretation of this text, rather than “errors in general”, is that certain clear deviations from the norms of written standard Swedish in the original texts are left without annotations – because they are not deviations in relation to the normalized text. This occurs for instance when a misspelled word in the original text has been corrected to another word altogether. Such a correction will be annotated as an instance of a faulty choice of word (L-W), and since the word in the original text cannot be analyzed as a misspelling of the word in the normalized text, the spelling error will be left without annotation.
The correction annotation is created by marking links between the original text and the normalized text with tags representing the categories in the SweLL correction annotation taxonomy, which is presented below. The tags are available in the Svala annotation tool in a list to the left on the screen.
A link may be tagged with one or several tags.
The links on which the tags are placed are visualized as lines in the Svala annotation tool. They run between elements of the original text and corresponding elements of the normalized text. The linked text elements typically consist of one token in each text version, but they may also consist of more than one token in one of the texts or in both texts. Moreover, links may run between a text element in one text version and “nothing”. This is normally the case when tokens have been added or removed in the normalized version of the text.
During the normalization process, preliminary links between the original and the normalized texts are created automatically. These links may be adjusted during the correction annotation process. The correct link must be in place before a tag is inserted. Links may be adjusted in the following way:
In order to break a link, place the marker on one of the linked text elements or on the link itself, and press the orphan button on the menu to the left.
In order to create a new link, mark all of the tokens which you wish to link (at least one token in the original text and at least one token in the normalized text) by clicking the tokens while holding ctrl (PC) or shift (Mac). Then press the group button on the menu to the left.
The Document comment field provides an opportunity to comment on deviations from the standard norm regarding text properties which cannot be adequately reflected by tags on individual links. This field may for instance be used when the verb tense choices in the text are inconsistent at a global text level, but when corrections of individual verb forms have generally not been made, since there is consistency more locally in the text.
The field may also be used for any kind of general comment on the text which the annotator regards as essential for the future corpus user.
In this section the Swell correction annotation taxonomy is presented, and directions on how to apply and interpret the tags associated with the correction categories are provided.
First, the general structure of the taxonomy is described. After that, the annotation categories are presented in the order in which they occur in the Svala annotation tool. After the presentation of all the correction categories follows a section on a few specific categorization issues, cutting through several correction categories. This includes the following sections:
Compounds vs multi-word expressions
Non-Swedish words and sequences
Verbal particles and reflexives
The SweLL correction annotation taxonomy contains five main categories of correction types:
Orthographic corrections (O), including “regular spelling corrections” (corrections of the string of letters), corrections between upper and lower case, and corrections of the use of spaces and hyphens between words.
Lexical corrections (L), including 1) corrections of the choice of word and 2) corrections of the internal morphological structure of word stems – i.e. corrections regarding the formation of words through compounding and derivation.
Morphological corrections (M), by which we specifically mean corrections concerning inflectional morphology. This category also covers some extra-morphological corrections which are closely related to inflectional forms and grammatical categories involving inflections.
Punctuation corrections (P), including corrections of the choice of punctuation marks as well as the adding or removal of punctuation marks, and also instances of merging or splitting sentences.
Syntactical corrections (S), i.e. corrections regarding the structure of multi-word phrases and clauses, including corrections of missing or redundant words, word order, the choice between a compound and a multi-word expression, and more complex syntactic corrections such as corrections between a finite clause structure and an infinitive phrase.
In addition to the correction tags included in these five main correction categories, the Svala tool provides six other tags, including tags for corrections made as a consequence of other corrections (C), corrections not covered by any of the defined correction categories (Unid), unintelligible strings (X), strings cited from a foreign language (Cit-FL), and finally two tags for notes and comments – OBS! for internal work notes, and Com! for comments intended for the corpus users.
In the following the annotation categories and their tags will be presented in the order in which they appear in the Svala annotation tool – i.e. the O tags first, followed by the four other main correction categories in alphabetic order (L, M, P, S), and, finally, the remaining six tags under the heading Other tags.
The O tags represent the orthographic correction category. It includes three sub-categories.
The plain O tag is used for regular spelling corrections, i.e. when the string of letters is different in the original text and the normalized text, due to a spelling mistake.
Examples
Det tåg 6 timmar från Teheran till Göteborg -> tog
Det finns monga affärer -> många
The O-Cap tag is used for corrections regarding the choice between upper and lower case letters.
Examples
Det fanns en affär och vi gick dit. vi köpte flera saker -> Det fanns en affär och vi gick dit. Vi köpte flera saker.
På Måndag är det den 3 Mars. -> På måndag är det den 3 mars.
The O-Comp tag is used for corrections which involve the removal of a space between two words which have been interpreted as making up a compound in the normalized text version, or, more rarely, the adding of a space between two words. It may also be used for corrections regarding the use of hyphens in compounds.
This tag should only be used for corrections concerning the mere orthographic rendering – with or without a space or a hyphen – of a compound or a multi-word expression, and not for corrections which are rather to be interpreted as involving an actual alternation between a compound and a multi-word expression. The latter case is covered by the S-Comp tag. (See the S-Comp section, and the section on compounds vs multi-word expressions.)
Examples
Jag kände mig jätte konstig -> jättekonstig
Det ligger ett kultur hus nära min bostad -> kulturhus
En före nänmt problem nynorska i norge har tas upp i artiklen “ Obligatoriska språkprov sliter på norsk tvåspråkighet “ av Bengt Östling ( 11.11.2013 ) där han har jämförts ” tvångsvenska “ debatten i Finland med nynorska debatten i Norge -> Ett tidigare nämnt problem som nynorskan i Norge har tas upp i artikeln “ Obligatoriska språkprov sliter på norsk tvåspråkighet “ av Bengt Östling ( 11.11.2013 ) , där han har jämfört “tvångssvenska”-debatten i Finland med nynorska-debatten i Norge
kultur-hus -> kulturhus
New York bor -> New York-bor
New Yorkbor -> New York-bor
tv program -> tv-program
engång -> en gång
för två årsen -> för två år sen
In the following example, the change from a comma to a long dash between Finland and Sverige is tagged with P-W, and the change from a space to a hyphen between Sverige and historien is tagged with O-Comp:
Corrections which involve both a removal of a space and a change of the form of the first part of the compound are tagged both with O-Comp and L-Der:
The L tags represent the lexical correction category. It includes four sub-categories.
The L-Der tag represents the correction category deviant word formation. It is used for corrections of the internal morphological structure of word stems, both with regard to compounding and to derivation.
The L-Der tag is exclusively used for links between one-word units, where the normalized word has kept at least one root morpheme from the original word, but where another morpheme has been removed, added, exchanged or had its form altered.
Examples
A derivational affix has been corrected, e.g.: De är stressiga på grund av studier -> De är stressade på grund av studier
A derivational affix has been added, e.g. ändring -> förändring
A derivational affix has been removed, e.g. förstöra -> störa
The form of the first part of a compound has been corrected (for instance by adding or removing a “foge-s”), e.g. tvångsvenska -> tvångssvenska; sagabok -> sagobok
A verbal particle of a compound form of a particle verb has been added, removed or corrected, e.g.:
Internet uppmanar vår förståelse -> Internet utmanar vår förståelse
För att sammanfatta och svara på frågan om likheter och skillnader finns mellan nynorskans ställning i Norge och svenkans ställning i Finland man måste förstå att debatten i båda länder är jätte lik varandra , bara hantering avskiljer -> För att sammanfatta och svara på frågan om det finns likheter och skillnader mellan nynorskans ställning i Norge och svenkans ställning i Finland måste man förstå att debatterna i de båda länderna är jättelika varandra , bara hanteringen skiljer
I annan text “ Tre röster om svenskans ställning i Finland “ Parnass 2013:3 som är skriven av författaren Catharina Söderbergh beskriver dem tre olika ställningar som finns om svenska inom Finland -> I en annan text , “ Tre röster om svenskans ställning i Finland “ ( Parnass 2013:3 ) , som är skriven av författaren Catharina Söderbergh , beskriver författaren de tre olika inställningar som finns till svenska i Finland
(See the special section on verbal particles and reflexives.)
A lexical root morpheme has been exchanged, e.g.: dagsskolan -> förskolan
A verbal particle has been exchanged for a derivational affix, e.g.: efterlyser -> belyser
A root morpheme and a derivational morpheme has been added, e.g.: maktfull –> maktfullkomlig
Note (1): Corrections of verbal particles of phrasal forms of particle verbs are not tagged with the L-Der tag, but with the L-W tag. (See the special section on verbal particles and reflexives.)
Note (2): The L-Der tag is not used for corrections which involve the addition of a -t suffix to an adjective which is used as an adverb, since this correction category is treated as a morphological correction with its own tag, M-Adj/adv.
Note (3): When the correction of a word tagged with L-Der involves a change of phrase type or part of speech, the correction is tagged with S-Type, in addition to L-Der.
Examples
norska bokmål -> norskt bokmål
en nybyggnad lägenhet -> en nybyggd lägenhet
Jag ska solen och bada -> Jag ska sola och bada
Jag behöver pengar för liv och betalning av min hus -> Jag behöver pengar för att leva och betala mitt hus
Changes from båda to både, or the other way around, are tagged as L-Der (and S-Type) rather than as L-W:
The L-FL tag is used for words from a foreign (non-Swedish) language which have been corrected to a Swedish word. It may also be applied to words which have certain non-Swedish traits due to influence from a foreign language.
Examples
Det fanns flera rum, två kök, balkony och trädgård -> balkong
Jag och min family -> familj
varför man ens lär sig svenska idag i Finland som beror på historiska event -> varför man ens lär sig svenska idag i Finland , vilket beror på historiska händelser
Bostad i D-hemland är litet het topik för att diskussera -> Bostäder i D-hemland är ett lite hett ämne att diskutera
The L-FL tag is used for corrections with the following characteristics:
The word in the original text is recognizable as a word from a foreign language to the annotator (who is obviously not equally proficient in all languages). Alternatively, the annotator recognizes certain non-Swedish traits in the word which are due to influence from a foreign language.
The word in the original text may or may not be correctly formed and used according to the norms of the assumed influencing language.
The word in the original text may have both Swedish and foreign traits, but if it is identifiable as an existing well-established Swedish word which has been replaced by an altogether different word, it should not be marked as L-FL, but as L-W. This holds even if the deviant choice of words is clearly due to lexical influence from the foreign language. For instance, the following example should be marked as L-W, although it is likely that the faulty use of busiga is influenced by the English word busy. (See the section on non-Swedish words and sequences below.)
It is not reasonable to interpret the correction as a correction of a spelling mistake or a word formation mistake. In that case the correction should be tagged with the O tag or the L-Der tag respectively. In other words: The O and L-Der tags have higher priority than the L-FL tag. (See the section on non-Swedish words and sequences below.)
The L-Ref tag is used for anaphoric expressions (particularly pronouns and pronominal adverbs) which have been corrected in order to have the grammatical form (gender, number, reflexive/non-reflexive), semantic content (masculine/feminine, directional/locational etc.), and specificity which suits its correlate and its textual position.
The L-Ref tag has higher priority than the M-Num, M-Gend and L-W tags.
The L-Ref tag has lower priority than the M-Def tag, and should only be applied in cases when the M-Def tag cannot be applied.
Examples
Jag har en bror, hon heter xx -> han heter xx
Stämmer det att våra sociala media skapar individualism och ensamhet? Eller skapar det kontakter och trivs? -> Eller skapar de kontakter … (In cases like this, the L-Ref tag has higher priority than the M-Num tag.)
Innan trädet blev borta dem fick fars äpplen och sin fru lagade saft och mos av äpplena -> Innan trädet var borta fick de fars äpplen och hans fru lagade saft och mos av äpplena
Östling skriver om politiker som tycker att obligatorisk språkundervisningen och prov på nynorska krävs för att försäkra ett minoritets språk som har ett gamal kulturarv knutet till den . -> Östling skriver om politiker som tycker att obligatorisk språkundervisning och prov på nynorska krävs för att skydda ett minoritetsspråk som har ett gammalt kulturarv knutet till sig .
“ Finns det något som är mer värt än pengar ? ” Jag vet faktiskt inte , den är en svår fråga . -> “ Finns det något som är mer värt än pengar ? ” Jag vet faktiskt inte , det är en svår fråga . (In cases like this, and the following examples, the L-Ref tag has higher priority than the M-Gend tag.)
Om man jobbar åtta timmar varje vardag för att få pengar , sedan är det lite hycklande att säga något som , ” Ja , det finns jo viktigare saker i livet ! ” Den är en tredjedel av din dag ! -> Om man jobbar åtta timmar varje vardag för att få pengar , är det sedan lite hycklande att säga något som : ” Ja , det finns ju viktigare saker i livet ! ” Det är en tredjedel av din dag !
Jag tycker om Gotland och den ligger i Sverige . -> Jag tycker om Gotland , och det ligger i Sverige .
I artikeln “ Svenskan tynar i Finland “ ( Forsknings & Framsteg 2009:2 ) Cecilia Christner Raid om varför man ens lär sig svenska idag i Finland som beror på historiska event . -> I artikeln “ Svenskan tynar i Finland “ ( Forskning & Framsteg 2009:2 ) skriver Cecilia Christner Riad om varför man ens lär sig svenska idag i Finland , vilket beror på historiska händelser . (In cases like this, and the following examples, the L-Ref tag has higher priority than the L-W tag.)
Man kan flyga till Gotland eller åka till Nynäshamn och åka båt till där -> Man kan flyga till Gotland eller åka till Nynäshamn och åka båt dit .
The L-Ref tag may also be used when a noun which is used anaphorically has been exchanged for a pronoun, or the other way around, in order for the specificity of the anaphoric expression to suit its textual position:
The L-W tag represents the correction category wrong word or phrase. It is used when a word or phrase in the original text has been replaced by another word or phrase in the normalized version. The L-W tag is thus placed on strings which are exchanged rather than corrected (see note below for further explanation).
The L-W tag is only applied when at least one of the strings (the original string and the normalized string) is (an attempt at) a word or a fixed phrase.
The L-W tag has lower priority than the L-Ref tag.
One word replaced by one word:
Uppfinningen som transformerade hela kommunikationsområdet -> Uppfinningen som förändrade hela kommunikationsområdet
Jag bor i D-område på en lägenhet -> Jag bor i D-område i en lägenhet
På det sättet kan kommunen motionera alla medborgare att träna → På det sättet kan kommunen motivera alla medborgare att träna.
vi solade och badade mycket och träffäde några personer från flera länder -> Vi solade och badade mycket och träffade människor från flera länder .
Alla blir busiga med sina sociala medier. -> Alla blir upptagna med sina sociala medier (See the section on non-Swedish words and sequences for further comments on this example.)
One word replaced by a multi-word expression:
Multi-word expression replaced by one word:
Multi-word expression replaced by another multi-word expression:
Phrasal verb replaced by another phrasal verb, both verb and particle exchanged:
Compound particle verb replaced by a phrasal verb, both verb and particle replaced:
Verbal particle replaced by another verbal particle, but verb kept; the verbal particle rather than the whole phrasal verb is tagged with L-W:
Fixed expression consisting of a lexical word (e.g. a noun) and a grammatical word (e.g. a preposition) replaced by another fixed expression consisting of the same lexical word but another grammatical word; the grammatical word rather than the full fixed expression is tagged with L-W:
Note (1): Corrections consisting in the mere removal or addition of the reflexive sig or a verbal particle are not tagged with L-W, but with S-R or S-M respectively – even if both the bare verb and the phrasal verb may be characterized as lexical units.
Note (2): When a correction tagged with L-W involves a change of phrase type/part of speech, the correction is also marked with the additional tag S-Type.
Examples
plötsligt bröt muren sinsemellan -> plötsligt rasade muren mellan dem
jag behover pengar f$r liv och betalning av min hus -> jag behöver pengar för att leva och betala för mitt hus
Tycker om min plats har en köpa mat ett , litet centrum en förskolan en vårdcentralen -> Jag tycker om platsen där jag bor , den har en mataffär , ett litet centrum , en förskola , en vårdcentral
Man kan promonera lång tid finns det blåser -> Man kan promenera länge när det blåser
Några tider vi kan titta och lyssna på hur funkar det -> Ibland kan vi titta och lyssna på hur det funkar .
Finns många nya lagenheterna i dyrare delar i huvudstaden , men detta är lång distans från räker nummer . -> Det finns många nya lägenheter i dyrare delar i huvudstaden , men detta är långt ifrån tillräckligt .
jag bor in lägenhet plan ett -> Jag bor i en lägenhet på plan ett
den obligatoriska svenskundervisningen i dem finska skolorna -> den obligatoriska svenskundervisningen i de finska skolorna (The pronoun form dem is changed to the form de used as a definite article. Cf. M-Case.)
Note (3): An expression tagged with L-W should be replaced rather than corrected. This means:
If both the original string and the normalized string are multi-word expressions containing one and the same lexical word, as well as one or more grammatical words which are changed, the L-W tag should be placed on the grammatical word(s) and not on the lexical word, as in the example above where under tiden is exchanged for med tiden.
If both the original string and the normalized string contain the same lexical morphemes but with another internal structure, the S-Comp tag should be applied rather than the L-W tag. Example: avsnittet av texten -> textavsnittet
If both the original string and the normalized string are multi-word expressions with the same main word (for instance the central noun of a noun phrase), one or several S-tags should be used rather than the L-W tag. The following correction is thus not marked as L-W but as S-Clause: min plats -> platsen där jag bor
If both the original and the normalized string are single words, all root morphemes of the word should be changed. Otherwise the L-Der tag should be applied. Example: dagsskola -> förskola
The M tags represent the category morphological corrections. It covers corrections related to inflections. This includes primarily corrections of individual inflectional forms, but in some cases also corrections of more complex grammatical constructions closely related to inflectional forms. The latter concerns basic definiteness constructions (see M-Def), the periphrastic comparative and superlative adjective constructions (see M-F), and tense-related verbal constructions involving auxiliaries (see M-Verb).
The category of morphological corrections includes eight sub-categories.
The M-Adj/adv tag is used for corrections involving the change of an adjective to its t-form, when the t-form is called for due to the adjective being used as an adverb.
Examples
Användaren mår dålig -> dåligt
Hur påverkar detta våra måenden långsiktig? -> långsiktigt
Jag tror att man måste hitta balansen mellan rikdom och andra saker som är riktiga viktiga -> Jag tror att man måste hitta balansen mellan rikedom och andra saker som är riktigt viktiga
The M-Adj/adv is also used for similar changes, when an adjective or adjective-like word is changed to a morphologically closely related but distinct adverb form:
Moreover, the M-Adj/adv tag is used when an adjectival form of the word liten is changed to the adverb form of the same word, i.e. lite. This holds also when the form litet is changed to the form lite. Although the form litet is occasionally used as an adverb in standard Swedish, it is too archaic to be used in most of the Swell text genres, and adverbial uses of the form litet are thus normally corrected to the form lite during normalization. Such changes are also tagged with M-Adj/adv:
The M-Case tag is used for corrections regarding the choice of case form for nouns (nominative vs genitive) and pronouns (nominative vs accusative).
Examples
50 kilometer avstånd -> kilometers
Som kan bidra till samhället utveckling -> samhällets
Det ger man en positiv energi -> Det ger en en positiv energi
Folk hinner inte prata med de -> Folk hinner inte prata med dem
Ingen förstår hon -> Ingen förstår henne
Note: When the form dem is changed to the form de used as a definite article, the correction is tagged as L-W, not as M-Case:
The M-Def tag is used for corrections regarding definiteness constructions. The kinds of corrections which are involved in this correction category are:
Change between indefinite and definite forms of nouns (bok vs boken; rum vs rummet) In these cases the M-Def tag has higher priority than the L-Ref tag.
Change between, removal of and addition of indefinite and definite articles (en, ett; den, det, de). In these cases the M-Def tag has higher priority than the S-M and S-R tags.
Change between indefinite and definite forms of adjectives and adjective-like pronouns (stor/stort vs stora; egen/eget vs egna)
Examples
Vi gick till McDonalds och åt maten -> mat
Jag trivs bättre på jobb här -> jobbet
Jag har läst ditt mejlet -> ditt mejl
min stor dag -> min stora dag
De gränserna som fanns mellan kvinnor och män har nästan försvunnit -> De gränser som fanns
När kommer den buss -> När kommer bussen
Då kommer svenska språket försvinna ur Finska samhället -> Då kommer svenska språket försvinna från det finska samhället
Han har egen rummet och jag egen -> Han har ett eget rum och jag ett eget
The M-F tag is used when a declension/conjugation form (typically a suffix) which is used to express a specific grammatical category (e.g. number or plural) has been corrected to a form belonging to another declension/conjugation within the same grammatical category. The tag is used for the following correction types:
Nouns:
Note: Unsuffixed noun forms will be interpreted as singular when corrected to a suffixed plural form, although unsuffixed plurals exist. Corrections like båt -> båtar will thus be tagged with the M-Num tag and not with the M-F tag.
Verbs:
Correction of a present tense suffix: jag lager mat -> jag lagar mat
Correction of a past tense suffix: böjade -> böjde
Correction of a supine suffix: kunnit -> kunnat
Correction from a verb form with a basic stem form and a suffix (past tense or supine) to a stem-changed verb form (with or without suffix): sjungde -> sjöng; drickit -> druckit
Adjectives:
Correction of a comparative suffix: högare -> högre
Correction of a superlative suffix: högast -> högst
Correction from an adjective form (comparative or superlative) with a basic stem form and a suffix to a stem-changed form: långare -> längre; ungast -> yngst
Correction of the choice between a periphrastic and an inflectional comparative or superlative construction: I Sverige är mer kallt än i xx -> I Sverige är det kallare än i xx
The tag may also be used for some analogous changes of pronouns:
The M-Gend tag is used to mark corrections of gender forms (neuter vs non-neuter) of nouns, articles, adjectives, and pronouns with adjective-like functions.
Examples
Det är en mycket trevlig område -> Det är ett mycket trevligt område
Den viktigaste är att … -> Det viktigaste är att …
Den första telefonnät -> Det första telefonnätet
Kommunikationen är möjligt -> Kommunikationen är möjlig
ifall undervisning i svenska blev frivillig -> frivilligt
Han har egen rummet och jag egen -> Han har ett eget rum och jag ett eget (cf. the same example under M-Def)
många i båda länder tycker att minoritet språken borde förbjudas eller att minoritetspåken har för mycket makt inom landen -> många i båda länderna tycker att minoritetsspråken borde förbjudas eller att minoritetsspråken har för mycket makt i landet
The M-Gend tag is also used for corrections of the overuse of the distinctly masculine form of adjectives:
(Since the masculine form is never obligatory, corrections from the feminine/common form to the masculine form are not made during the normalization, and thus do never occur in the correction annotation process.)
Note: Gender corrections of pronouns which are due to their anaphoric reference will be covered by the L-Ref tag.
The M-Num tag is used to mark number corrections of nouns, articles, adjectives, participles, and pronouns with adjective-like functions.
Examples
Stress kan komma i många form -> Stress kan komma i många former
Så fort bilder är tagen -> Så fort bilderna är tagna
Alla sitter bakom sina skärmens sken -> Alla sitter bakom sina skärmars sken
Note: Number corrections of pronouns which are due to their anaphoric reference will be covered by the L-Ref tag.
The M-Other tag is used for corrections involving inflectional morphology for which none of the other M tags are suited, or for ambiguous cases when different sound interpretations of the correction lead to different M tags.
The usage of the M-Other tag covers corrections between the comparational forms of adjectives, including corrections between non-morphologically related words functioning as different comparational forms of the same adjective (e.g. dålig and sämre or många and fler):
I slutligen kan jag säga utifrån texterna jag läste att det finns många nackdelar än fördelar med att avskaffa den obligatorisk svenskan i Finland -> Slutligen kan jag säga , utifrån texterna jag har läst , att det finns fler nackdelar än fördelar med att avskaffa den obligatoriska svenskan i Finland
Morfar ville visa de att det finns nånting som är viktigaste , deras förändringen t.ex. -> Morfar ville visa dem att det finns nånting som är viktigare , deras förändring t.ex.
Note: Take care not to overuse the M-Other tag. The other M tags should be carefully considered before choosing this one.
The M-Verb tag covers corrections regarding inflectional verb forms and basic tense constructions involving auxiliaries.
The verb-related grammatical categories involved are primarily tense, mode and voice.
The verb forms involved are the non-finite verb forms (infinitive and supine), the basic tense forms (present and past), the imperative, and the s-forms (both when used in passive constructions and in other uses).
The extra-inflectional constructions involved are the tense-related constructions including the auxiliary verbs ha, skola and komma and the non-finite verb forms.
Examples
Jag trivs där för att jag hade fler vänner -> Jag trivdes där för att jag hade fler vänner
Slutligen tror jag att sociala medier blev en essentiell del -> Slutligen tror jag att sociala medier har blivit en essentiell del
Det betyder inte att vi glomma bort vår tradition Det betyder inte att vi ska glömma bort vår tradition
Världen utvecklar i teknologin i varje minut -> Världen utvecklas i teknologin i varje minut
The P tags represent the category of punctuation corrections, including instances of of merging or splitting sentences. It has four sub-categories.
The P-M tag is used for corrections involving the addition of a punctuation mark.
Examples
Jag har fyra barn två pojkar och två flickor -> Jag har fyra barn, två pojkar och två flickor
År 1972 togs den sista manuella växeln ur bruk anger Björkkvist -> År 1972 togs den sista manuella växeln ur bruk, anger Björkkvist
Det ger en positiv energi därmed kan man bli av med stress -> Det ger en positiv energi. Därmed kan man bli av med stress
Min mamma lagar så god mat jag trivs med hennes mat -> Min mamma lagar så god mat. Jag trivs med hennes mat
Note: Additions of hyphens between words are not included in this category, but in the O-Comp category.
The P-R tag is used for corrections involving the removal of a punctuation mark.
Examples
Note: Removals of hyphens between the constituent words in a compound are not included in this category, but in the O-Comp category.
The P-Sent tag is used for corrections involving splitting a sentence or merging two sentences into one, when this correction involves more than the pure insertion or removal of a punctuation mark – in the typical case the adding or removal of a conjunction.
Examples
In this example, the P-Sent tag is placed on a link between och in the original text and the period in the normalized text. The link between vi and Vi is tagged with C, as a consistency correction necessitated by another correction.
In this example, the P-Sent tag is placed on a link between som in the original text and the period in the normalized text. The link between hon and Hon is tagged with C, in the same way as in the example above.
The P-W tag is used when a punctuation mark in the original text has been replaced with another punctuation mark in the normalized text.
Examples
Note (1): Instances where a space has been corrected to a hyphen between the constituent words in a compound are not marked with this tag, but with O-Comp or S-Comp. The same holds for instances where a hyphen between words has been corrected to a space.
Note (2): Instances where a hyphen has been used in the original text where a dash would be more appropriate are left uncorrected and should thus not appear as corrections to be annotated.
Note (3): Possible errors involving the incorrect placement of a space before a punctuation mark will not be corrected in the normalization process, since spaces are always inserted before punctuation marks for the sake of tokenization. Consequently, such errors will not be tagged.
Note (4): Possible errors involving the lack of a space between a punctuation mark and the following word are corrected in the normalization process (a space is inserted), but are nevertheless left untagged.
Example:
In this example, a space is inserted between the period and tack, and the t in tack is changed from lower to upper case. Tack will be tagged with O-Cap, but the insertion of the space will not be tagged.
The S tags represent the syntactical correction category. It contains eleven sub-categories.
The S-adv tag is used for corrections involving the placement of an adverbial.
This word order tag has the highest priority of the three word order tags (S-Adv, S-FinV and S-WO), and should be applied whenever a word order correction may be interpreted as concerning the placement of an adverbial. Particularly, word order corrections regarding the relative ordering between an adverbial and a finite verb, should be marked as S-Adv rather than as S-FinV.
Examples
Jag ofta är vaken länge på kvällarna. → Jag är ofta vaken länge på kvällarna.
Å ena sidan idag har… → Å ena sidan har idag…
Jag är jättetrött eftersom jag sover inte på nätterna. → Jag är jättetrött eftersom jag inte sover på nätterna.
som i sin tur har tydligt påverkat … → som i sin tur tydligt har påverkat …
och med hela världen dela sina idéer och tankar. → och dela sina idéer och tankar med hela världen.
om hur under 600 år Finland var en del av Sverige → om hur Finland var en del av Sverige under 600 år
The S-Clause tag is used for corrections involving changes of the most basic clause structure. The corrections in this category may be divided into the two following main types:
Subject changed to a prepositional complement in an adverbial:
Changes between a passive construction and an active construction:
Subject changed to egentligt subjekt (’object-positioned subject’):
på andra sidan finns människor som har så mycket pengar att de kan köpa halva världen -> Å andra sidan finns det människor som har så mycket pengar att de kan köpa halva världen (cf. S-Msubj and note below)
Fler faktorer för detta finns , exempelvis mellan 1950 och 1980 skedde en inflyttning då många svenskar från Finland flyttade tillbaka till Sverige -> Fler faktorer bakom detta finns , exempelvis att det mellan 1950 och 1980 skedde en inflyttning då många svenskar från Finland flyttade tillbaka till Sverige (Att is marked with S-type, det is marked with S-Msubj, and en inflyttning _is marked with S-Clause. Cf. S-Msubj and S-Type.)
The structure of a noun phrase is changed by changing a framförställt attribut (‘preceding attribute’) to an efterställt attribut (‘succeeding attribute’) in the form of a relative clause
The structure of a clause is changed in a way which involves adding an extra level of subordination. In the following example an ambiguous clause structure is normalized in a way which involves adding a finite verb (är), and thus an extra clause, to the structure. The following clause is changed from a main clause to a subordinate clause (by changing så to att då). Är is tagged with S-Clause, and a link between så in the original text and att då in the normalized text is tagged with S-Type:
Note: When a clause is changed by 1) adding an expletive det as a subject and 2) changing a subject to an egentligt subjekt (‘object-positioned subject’), two tags are needed to mark the corrections:
The S-Msubj tag is placed on the added det.
The S-Clause tag is placed on the element which has been changed from subject to egentligt subjekt (‘object-positioned subject’).
The S-Comp tag is used for:
corrections regarding the choice between a compound and a multi-word expression. This includes the choice between a compound and a phrasal structure for the combination of a verb and a verbal particle. (See the special section on verbal particles and reflexives.)
det vardagliga livet -> vardagslivet
livsskillnaden -> skillnaden i liv
svenska undervisningar -> svenskundervisning
avsnittet av texten -> textavsnittet
Enligt Hyltenstam så kan minoritetsspråk räddas om man inblandar dem äldre som kan språket -> Enligt Hyltenstam så kan minoritetsspråk räddas om man blandar in de äldre som kan språket
Om det händer att det finns planhalvor så kan det bero på blyghet, osäkerhet, rädsla eller socialfobi -> social fobi (See the section on compounds vs multi-word phrases for further discussion of this example.)
Hejdlös sociala medier användning orsakar ensamhet -> Hejdlös användning av sociala medier orsakar ensamhet (See the section on compounds vs multi-word phrases for further discussion of this example.)
other corrections involving the restructuring of the same lexical morphemes within a phrase
Skillnader är stor av Sveriges bostad -> Skillnaderna är stora mot bostäder i Sverige
Cecila Christner skriver om hur det svenska språket i skolor blir kallad tvångsvenska i samma tid svenska har blivit icke populärt i Finland -> Cecilia Christner skriver om hur det svenska språket i skolorna blir kallat tvångssvenska samtidigt som svenska har blivit impopulärt i Finland
corrections involving the change between an expression with a lexical negation and an expression with a derivational negation:
Note: Corrections regarding the mere orthographic rendering of a string with or without a space should not be marked with the S-Comp tag but with the O-Comp tag. (See O-Comp and the section on compounds vs multi-word expressions below.)
The S-Ext tag is used for extensive, complex corrections. The syntactic structure of the normalized text segment may rather be described as created than as corrected, and the correction often also involves the addition of lexical words. The original text gives a fair indication of the intended meaning (otherwise the correction would be X-marked), but it gives a very poor basis for assuming a specific syntactic goal structure.
Examples
Det därför tycker jag om det är simma lungt och blåser -> Jag tycker om det för att jag tycker om att simma lugnt och när det blåser
Jag hoppas att du intressant för din ny livs -> Jag hoppas att du tycker att ditt nya liv är intressant
Så hur mycket pojkarna bettalad värt inte deras äppelträd eftersom trädet var viktig för dem -> Så hur mycket pojkarna betalade var mindre än vad deras äppelträd var värt eftersom trädet var viktigt för dem .
Det är lite bättre i huvudstad , när många manniskor bo tilsammans eftersom de kan e betala för -> Det är lite bättre i huvudstaden , när många människor bor tillsammans eftersom det gör att de kan betala (det gör att in the normalization is tagged with S-Ext; e in the original text is tagged with X and S-R; för is tagged with S-R.)
Note: The S-Ext tag should only be applied in cases when a correction has actually been made, and when the original text gives fairly sound support for the interpretation presented in the normalized version. Text segments which are so difficult to interpret that they are either left unchanged or normalized on the basis of guesses rather than interpretations should be tagged with X.
The S-FinV tag is used for corrections concerning the placement of a finite verb, unless the correction regards the ordering between the finite verb and an adverb, in which case the correction is tagged with the S-Adv tag. The S-FinV tag thus has lower priority than the S-Adv tag, but higher priority than the S-WO tag.
Examples
Vor du bor? -> Var bor du?
I morgon jag åker -> I morgon åker jag
Efteråt man surfar på internet -> Efteråt surfar man på internet
När jag ätiti, jag sover -> När jag ätit, sover jag
Eller det skapar kontakter och trivs? -> Eller skapar de kontakter och trivsel?
The S-M tag is used when a word is missing in the original text and has been added in the normalized version. This includes the addition of reflexives and verbal particles. (See the special section on verbal paricles and reflexive.)
The S-M tag has lower priority than the S-Msubj tag and the M-Def tag, and should only be applied in cases when neither of these other tags may be applied.
Examples
Hon fördelen med att behålla den obligatoriska svenskan -> Hon beskriver fördelen med att behålla den obligatoriska svenskan.
Sådana känslor gör användaren mår dåligt -> Sådana känslor gör att användaren mår dåligt
Jag gillar inte tapeterna i min kusins lägenhet men han inte byta dem eftersom det är hyresrätt → … han kan inte byta dem …
Jag trivs mycket bo med dem -> jag trivs mycket med att bo med dem
Man måste akta att man inte ramlar -> man måste akta sig så att man inte ramlar
men känner bara fysiskt närvarande -> men känner sig bara fysiskt närvarande
Jag slår ord i ordboken när jag inte vet -> Jag slår upp ord i ordboken när jag inte vet…
The S-Msubj tag is used to mark corrections involving the addition of a subject which is missing in a clause in the original text. This includes cases when the pronoun/subordinating conjunction som has been inserted as a subject.
The S-Msubj tag has higher priority than the S-M tag.
Examples
Regnar ute -> Det regnar ute
Det är inte bara arbetet och arbetslivet kan ge stress -> Det är inte bara arbetet och arbetslivet som kan ge stress
Det är viktigt att veta vad händer -> Det är viktigt att veta vad som händer
I annan text “ Tre röster om svenskans ställning i Finland “ Parnass 2013:3 som är skriven av författaren Catharina Söderbergh beskriver dem tre olika ställningar som finns om svenska inom Finland -> I en annan text , “ Tre röster om svenskans ställning i Finland “ ( Parnass 2013:3 ) , som är skriven av författaren Catharina Söderbergh , beskriver författaren de tre olika inställningar som finns till svenska i Finland
Finns många nya lagenheterna i dyrare delar i huvudstaden -> Det finns många nya lägenheter i dyrare delar i huvudstaden
The S-Msubj tag should be placed on a det which has been inserted as a subject, also in the following cases:
Det replaces a subject which has been changed to an egentligt subjekt (‘object-positioned subject’) (cf. S-Clause)
A heavy subject is moved to a position further back in the clause, and det is inserted in the subject position:
Det (and possibly att) is inserted as a subject before an infinitive phrase already functioning as a subject (and the infinitive phrase becomes an attribute to det):
Note: The S-Msubj tag should only be applied in those cases when the clause is already present in the original text. Thus, in the following example, the generic pronoun man is added as a subject as a part of a correction involving changing an infinitive phrase to a finite clause, and the correction, including the addition of man, is marked as S-type. The S-Msubj tag should not be applied.
The S-Other tag is used for syntactic corrections not covered by any of the other S-tags.
The S-R tag is used when a word is redundant in the original text and has been removed in the normalized version. This includes the removal of reflexives and verbal particles. (See the section on verbal particles and reflexives.)
The S-R tag has lower priority than the M-Def tag, and should only be applied in cases when the M-Def tag is not applicable.
Examples
På 2000- talet kom många kommunikationsappar , bland annat var Facebook , Twitter , Instegram och Snapchat , påpekar Rose ( 2012 ) -> På 2000- talet kom många kommunikationsappar , bland annat Facebook , Twitter , Instagram och Snapchat , påpekar Rose ( 2012 )
Man behöver inte att klä på sig -> Man behöver inte klä på sig
Det är ett personligt ansvar för att välja -> Det är ett personligt ansvar att välja …
X-stad är närmare till X-land än X-stad. -> X-stad är närmare X-land än X-stad.
De första mobiltelefonerna kom i 1957. -> De första mobiltelefonerna kom 1957.
De nya kommunikationssätten har medfört med stora möjligheter -> De nya kommunikationssätten har medfört stora möjligheter
sociala medier blev en essentiell del av våra liv som vi inte kan slänga det i den moderna världen -> sociala medier blev en essentiell del av våra liv som vi inte kan slänga i den moderna världen
De promenerade sig i parken -> De promenerade i parken
Hon gav bort honom en blomma -> Hon gav honom en blomma
The S-Type tag is used when a phrase or clause has in its entirety been changed to another phrase- or clause-type, such as:
changes between a noun phrase and an infinitive phrase:
changes between a finite clause and a noun phrase:
changes between an infinitive phrase and finite clause:
changes between a subordinate clause and a main clause:
Fler faktorer för detta finns , exempelvis mellan 1950 och 1980 skedde en inflyttning då många svenskar från Finland flyttade tillbaka till Sverige -> Fler faktorer bakom detta finns , exempelvis att det mellan 1950 och 1980 skedde en inflyttning då många svenskar från Finland flyttade tillbaka till Sverige (Att is marked with S-type, det is marked with S-Msubj, and en inflyttning is marked with S-Clause. Cf. S-Msubj and S-Clause.)
Konsekvenserna man skulle få ifall undervisning i svenska blev frivillig så skulle mer än hälften av finska befolkningen avskaffa svenskan som modersmålsundervisning och istället fokusera på finska då dem sällan använder svenskan -> Konsekvenserna man skulle få ifall undervisning i svenska blev frivilligt är att då skulle mer än hälften av den finska befolkningen välja bort svenskan som modersmålsundervisning och istället fokusera på finska då de sällan använder svenskan (Är is tagged with S-Clause; a link between så and att då is marked with S-Type.)
Note (1): The S-Type tag may be combined with the L-W tag and the L-Der tag. See these sections for examples.
Note (2): Word-order corrections covered by the word-order tags (S-Adv, S-FinV, S-WO) are not in themselves a basis for considering the correction a change of clause type (but rather as corrections to suit a clause type which has been indicated by other means). The S-Type tag should thus not be applied to a correction solely on the basis of an adverbial having been moved from a typical main clause position to a typical subordinate clause position (or vice versa). That correction is covered by the S-Adv tag. For an S-Type tag to be applied to a clause which has been changed from a subordinate clause to a main clause etc., the change of clause structure has to be indicated for instance by the addition of or removal of a subjunction.
The S-WO tag is used for word order corrections which are not covered by the S-Adv or the S-FinV categories.
In corrections regarding the relative placement of a phrasal head and a modifying element (for instance a noun and its attribute), the modifying element should be marked rather than the phrasal head (e.g. min rather than bostad in the example below):
In cases when the word order change may be interpreted as moving an element “out of” a normally fairly fixed structure, that element should be marked rather than an element included in the more fixed structure (dem rather than upp in the example below).
In other cases the placement of the tag may be chosen freely between the elements which have been moved relative to each other, but the readability of the resulting visualization should be taken into consideration.
The final group of tags, collected under the heading Other, contains six tags used for various purposes:
The C tag is used for corrections which are necessitated by other corrections, but which do not reflect mistakes in the original text.
The Com! and OBS! tags are used for notes and comments – the Com! tag for comments intended for the future corpus user, and the OBS! tag for work notes on pending analyses meant for internal project use.
The Cit-FL tag is used for segments of foreign language which have not been corrected during normalization.
The X tag is used for unintelligible strings.
The Unid tag is used for “unidentified” corrections, i.e. corrections which are not covered by any of the correction categories defined in the taxonomy.
Four of these tags, namely Cit-FL, Com!, OBS! and X are available already during the normalization process, and are normally inserted already by the normalizer.
The C tag represents consistency corrections, a category which covers necessary (follow-up) corrections in the text that come as a result of a previous correction, i.e. originally there was no mistake in the segment, but due to an introduced correction in the neighboring context, a correction is necessary in the segment. By using this tag we indicate that the error was not made originally by the learner.
In some instances it may not be self-evident which one of two related corrections that should be considered as necessitating the other, but by marking one of them with C we avoid marking a single mistake in the original text as two.
Examples
The shift of word order is marked as a word order correction (S-WO). The change from definite form (bostaden) to indefinite form (bostad) of the noun is made necessary because of the shift of word order, and is thus marked as a consistency correction (C).
The insertion of the full stop is tagged as a punctuation correction (P-M). The capitalization of the following D is made necessary because of the insertion of the full stop, and is thus marked as a consistency correction (C).
The change from bostanden to bostadsområde is marked as a lexical correction (L-W) and a morphological correction (M-DEF). The gender change of the pronoun, from min (non-neuter) to mitt (neuter) is made necessary because of the change of word (from the non-neuter bostad to the neuter bostadsområde), and is thus marked as a consistency correction (C).
The removal of är is marked as S-R, while the movement of går to the finite verb position (where it replaces the likewise finite verb är) is tagged with C.
The link between första and först is tagged with S-Adj/adv (see this section) and with C, which indicates that the word order change is a consequence of the change from adjective to adverb. (An alternative normalization would be to keep the adjective form första and add the noun gången.)
The addition of saker is tagged as S-M, and the correction of vad to vilka is tagged with C, since it is the congruence with saker which determines the choice of pronoun.
The Cit-FL tag is used for foreign (non-Swedish) words, phrases or text segments, which have been kept by the normalizer since their usage has been judged acceptable given the norms of the text type in question. This may be the case for instance for explicitly marked citations or intentional code switching appropriate for the genre. The Cit-FL tag is thus used to mark words and text segments which have not been corrected in the normalized version, but which nevertheless are not passable as standard Swedish.
Note that the only requirement for applying this code is that the word or text segment is recognizable as another language than Swedish, and that the choice to use this other language is judged appropriate for the genre and the text. No judgement or correction of the word or text segment is made relative to the norms of the foreign language in questions. For instance, spelling mistakes are left uncorrected.
The Cit-FL tag is usually added already during normalization.
Examples
Judged as appropriate code switching:
Clearly marked citation of norwegian passage:
The Com! tag is connected to an edge comment field, which is open for freely composed comments. It is available already during the normalization process.
The Com! tag is used for comments on specific tokens or text sequences which are relevant for future users of the corpus, and which are thus meant to be kept in the published corpus. (Comments regarding the text as a whole or recurring properties in the text may be added in the Document comment field.)
The Com! tag may for instance be used to mark text segments which are copied from the task formulation. If a significant portion of the text consists of copied text, this should preferably be indicated also in the Document comment field, in addition to the edge comment field connected to the Com! tag which is inserted on the specific text segment or the specific text segments.
The OBS! tag is, like the Com! tag, connected to an edge comment field, and is available already during the normalization process.
The OBS! tag is used to mark pending analyses to which the annotator wants to return, remarks which the normalizer wishes to pass on to the correction annotator, etc.
The X tag is used to mark unintelligible strings in the original text. The tag is available and usually added already during the normalization process. The marked original string may be left unchanged in the normalized version, or the normalizer may replace it with some more or less wild guess of the intended message.
The X tag may be used both in cases when there is no reasonable interpretation of the string, and when there are several somewhat reasonable interpretations, but none of these interpretations may be settled as better than the other.
Examples
En argumentera på är viktigt hur man bor är bor på en bra hem
och vi har 3 rum it huset dar rum är meka bra liv med maen familija dar huset familjbostder och jag vill sta och jog kan ante skriv meka ord -> och vi har 3 rum i huset . dar rum är mycket bra liv med min familj där i huset från Familjebostäder och jag vill stanna och jag kan inte skriva många ord
The text from which this example is collected has some german or possibly dutch traits, and a fairly resonable guess is that dar rum is meant to be darum. Another, less likely, possibility is that dar rum is intended to mean där rum, but that interpretation necessitates extensive changes in the rest of the text passage in order to create a syntactically functional string. By marking dar rum with X and keeping the rest of the sentence unchanged in the normalized text version, the normalization as a whole is better adjusted to the principle of minimal change.
Stanna is a reasonable guess about the intention with sta, but not well founded enough for the correction to be marked as an orthographic correction. Many other interpretations are obviously also possible.
The X tag may be combined with the S-R tag when the unitelligible string is also redundant, and thus eliminated in the normalization:
The Unid tag is used for any type of correction which cannot be covered by any of the correction categories defined in the taxonomy.
Corrections concerning the forming of an expression as a compound or a multi-word expression are divided into two categories in the SweLL correction taxonomy: Corrections which are judged to concern the mere orthographic rendering with or without a space between two words are marked with the O-Comp tag, while corrections which are judged to concern the actual choice between a compound and a multi-word expression are marked with the S-Comp tag. Borderline or unclear cases between these two categories obviously exist, but for the most part one of the options is clearly the better one.
The O-Comp tag is primarily used for corrections of standard cases of särskrivning (the faulty writing of a compound with a space in between the two compounded words):
It may also be used for corrections which involve changing the first word of the compound into a specific compound form, in addition to the removal of the space between the two words. In such cases, the correction is marked with both the O-Comp tag and the L-Der tag:
More rarely, the O-Comp tag may be applied when a multi-word expression in the original text has been corrected through the adding of a space between two of the words:
The S-Comp tag is used whenever the correction made is more complex than the mere addition or removal of a space between two words (and possibly changing the form of the first word of a compound).
det vardagliga livet -> vardagslivet
livsskillnaden -> skillnaden i liv
svenska undervisningar -> svenskundervisning
Enligt Hyltenstam så kan minoritetsspråk räddas om man inblandar dem äldre som kan språket -> Enligt Hyltenstam så kan minoritetsspråk räddas om man blandar in de äldre som kan språket
However, in some cases the S-Comp tag is more suitable than the O-Comp tag, although the correction superficially merely involves the presence of a space. This is the case when two words may be correctly formed as a compound (without a space) and as a two-word expression (with a space), and the meaning of the compound version and the two-word version are either the same or else both plausible in the context. The correction made is thus not due to the chosen expression being unthinkable, but due to the other expressions happening to be the established lexical unit:
In this case, socialfobi is a perfectly well formed compound, but it is corrected to the two-word expression social fobi because this is the established way to express the intended meaning. The mistake made by the writer is therefore not judged to be a case of having missed a space in between two words in a two-word expression, but it is judged to be a case of the writer actually having chosen the compound expression instead of the established two-word expression. The correction is thus marked with the S-Comp tag.
Moreover, in some cases the S-Comp tag may be applied although neither the original nor the normalized string is a single orthographic word. This is the case when the string in the original text may be interpreted as an instance of a compound, although it includes spaces:
In this particular case, the string in the original text may be interpreted as a compound between a two-word phrase (sociala medier) and the word användning. According to the norms of written standard Swedish, such compounds should be written as sociala medier-användning etc. While such a minimal correction is not an unthinkable solution for the normalization, the normalizer has here judged a restructuring of the NP as a better solution, and the correction should be tagged S-Comp.
There are a number of ways to handle non-Swedish words during normalization and correction annotation. Many of the fundamental choices are made during the normalization process rather than during the correction annotation process.
The first judgement to be made when coming across a word stemming from another language in the material is naturally whether the word may be recognized as having been incorporated into written standard Swedish; in such cases the word is left uncorrected and untagged. This judgement is made during normalization.
When a word (or sequence) in an original text is recognized as belonging to a foreign language – or as having traits from a foreign language – and when this word may not be recognized as part of written standard Swedish, a number of options are at hand:
1) The word/sequence is judged as a genre appropriate usage of cited foreign language (explicitly signaled citations, code switching etc.). -> Not corrected, tagged Cit-FL during normalization.
2) The word is not judged as a genre appropriate usage of cited foreign language and is thus corrected to a Swedish word during normalization:
a) The form used may be interpreted as a misspelled Swedish word. -> Corrected during normalization, tagged O during correction annotation: **kaffee** -> **kaffe**; **can** -> **kan**
b) The form used may be interpreted as a Swedish word with an incorrect usage of derivational affixes etc. -> Corrected during normalization, tagged L-Der during correction annotation: **national helgdag** -> **nationell helgdag**
c) Neither a nor b applies. -> Corrected during normalization, tagged L-FL during correction annotation: **balkony** -> **balkong**; **family** -> **familj**; **gas bojler** -> **gaskokare**
Note: A word in the original text which is identifiable as a Swedish word, but which is used with another meaning in a way which is likely to be due to influence from a similar non-Swedish word, should be corrected and marked as L-W (not as L-FL):
In this example, it is likely that the incorrect usage of the correct Swedish word busiga is influenced by the word’s similarity to the English word busy – and it is partly based on this assumption that the writer’s intended meaning has been interpreted as ‘upptagna’. But since busiga is a correct Swedish word, with a distinctly Swedish morphological structure, the correction is tagged as L-W rather than as L-FL.
Several tags are used for corrections involving phrasal or compound verbs made up by a verb and verbal particle or a reflexive marker, primarily O-Comp, S-Comp, L-Der, L-W, S-M and S-R. This section provides an overview of the usage of these six tags for this category of corrections.
O-Comp: A space is removed between a verbal particle and a following verb, making up a compound verb:
S-Comp: A compound form of a particle verb is changed to a phrasal form, or the other way around:
Enligt Hyltenstam så kan minoritetsspråk räddas om man inblandar dem äldre som kan språket -> Enligt Hyltenstam så kan minoritetsspråk räddas om man blandar in de äldre som kan språket
Tåget går av från spår 3 -> Tåget avgår från spår 3
L-Der: A particle of a compound particle verb is changed, removed or added. Both the original and the normalized string consist of one single token. The original string may or may not be an existing Swedish word.
Internet uppmanar vår förståelse -> Internet utmanar vår förståelse
Han inhämtade väskorna från hotellrummet -> Han hämtade väskorna från hotellrummet
Hon minde honom om mötet på eftermiddagen -> Hon påminde honom om mötet på eftermiddagen
L-W, placed on the whole phrasal verb: Either the original string or the normalized string (or both) is a phrasal verb, and the verb itself is changed, not just the particle/reflexive marker:
L-W, placed on the particle: A verbal particle is replaced by another verbal particle, but the verb is kept:
S-M: An independent (non-compound) verbal particle or a reflexive marker is added:
men känner bara fysiskt närvarande -> men känner sig bara fysiskt närvarande
Jag slår ord i ordboken när jag inte vet -> Jag slår upp ord i ordboken när jag inte vet…
S-R: An independent (non-compound) verbal particle or a reflexive is removed.
De promenerade sig i parken -> De promenerade i parken
Hon gav bort honom en blomma -> Hon gav honom en blomma