MANUAL FOR MORPHOLOGICAL ANNOTATION

Lemma in PDT 1.0 has two parts. First part, the lemma proper, has to be a unique identifier of the lexical item. Usually it is the base form (e.g. infinitive for a verb) of the word, possibly followed by a number distinguishing different lemmas with the same base forms. Second part (optional) is not part of the identifier and contains additional information about the lemma, e.g. semantic or derivational information.

Note: There is a convention that if lemmas use numbers to distinguish lexical items with the same base form, they all have to use them- i.e. instead of sets of lemmas {X, X-1, X- 2} or {X, X-2, X-3}, there should be a set {X-1, X-2, X-3}

Note: The lemmas having different semantic suffixes should have different numbers. In this manual we behave as the annotator. We try to mark such improper numbers by roman font (other part of the lemma is in italics). For example stop in akce Stop million will be marked as stop-1_;m and not stop-1_;m).

Some names are sometimes declined, sometimes not (Bill – o Bill Clintonovi, o Billu Clintonovi, o Billovi). The tag for nondeclined form is NgXXA.

All lemmas of autosemantic words in compound names must have the category determined by the whole name (e.g. K, R). The lemmas of functional words contain default type category (m).

The problem is that a name of one type can occur as part of a name of a different type:

England is G noun in the first, K adjective in the second and R adj. in the third name.

If the lemma of the category you need does not exist and you have to insert a new one, do not care about numbering of lemmas, somebody else will do it (it would impossible to ensure that the numbers were unique across all annotators). That means, if there is other lemma having just different category (e.g. there is England_;G available, but you need England_;R), just change the category label.

The categorization is sometimes quite tricky – you do not know, whether to consider a phrase a name or a name plus normal word:

Using the separation of morphology and name categorization, this is quite easy:

Horses have all kind of names (e.g. Vinná réva, Deprivace, He Shall Reign, La Paloma Monitor, Frýdlant, Gold End, Lučina, Green Peace, Areál, First, Bounty), and quite often you do not know if it is female or male (sometimes even female like names belong to a male horse). One clue is, that in an Oak (a horse contest type), all horses are young mares – females.

In PDT 1.0 the names of horses where mostly not annotated correctly – simply any available name was selected (Otherwise, a new lemma with category Y would have to be inserted in each case: e.g. Deprivace would be Deprivace_;Y, annotated as deprivace, He Shall Reign annotated as normal English phrase: he_,t, shall_,t reign_,t).

Similar problem is with the names of musical groups and DJ's. For famous groups and DJ's enter separate lemmas, for others use normal available lemmas.

Name of the town in the club name: if only the town is noted, it is annotated as a geographic name (G), if the whole name of the club is noted, it is annotated as an institution (K). It is analogous to countries. (Česko vs. Německo are annotated as G)

Of course, it can be a problem to know it with foreign clubs. If you do not know, annotate it as an institution (K).

It is the adjective of Udine (town in NE Italy), the official name of the football club is Udinese Calcio (calcio = football). However in Czech, the name is perceived as a noun and as the name of that club, therefore it is probably better to use it in that way:

To determine, whether something is a name of a town or a club, you can try to find that name on a map (eg. http://www.expedia.com/pub/agent.dll?qscr=mmfn) and also find the club (e.g. http://www.soccerage.com).

The name of the sport club often contains some abbreviation. Some are common and present in the analyzer's lexicon (e.g. FC, AC) some are quite unusual (e.g. EV, ERC, EC, ERC, EG, VS, AS). If they are not present in the lexicon, entering them, suffixing the lemma by _:B_;K_;w and using NNNXX-----A---8 as tag,

Insisting on inclusion of name categories (K, R, etc.), implies explosion of number of lemmas. We follow each examples section by analogous examples using the above- proposed separation of morphology and name categorization (see Section 3.2).

Abbreviations can be used with different genders (e.g ODS – feminine (strana) or neuter). Any abbreviation can have neuter gender. If the gender cannot be disambiguated by the context, use the gender used elsewhere in article. If the author mixes genders or there are no disambiguating contexts, use the gender inherent gender of the abbreviation. In Czech, is usually easy to determine – it is the gender of the head of unabbreviated equivalent (e.g. ODS – strana → f). With foreign abbreviations it is much more problematic, different people use different genders (e.g. because of different translation). If you are not certain which of the gender is most widely used, use the default neutrum.

Normal abbreviations have sometimes as a lemma the abbreviation (and sometimes the original unabbreviated word. Usually the former method is used for abbreviation that are more common then the unabbreviated word (and for abbreviation of multi word expressions). But it is not always true.

For discussion about determining the gender of an abbreviation, see Section 4.1

Note: The following is still not official.

Isolated letters (e.g. A-konto) are handled as abbreviations. The only exception is if they are not in the name (zápas skupiny B). Many of the annotations suggested bellow are still not offered by the morphological analyzer. Moreover, sometimes the morphological analyzer is constrained to offer appropriate lemma and tag only if the letter is followed by a dot. Should be repaired.

You have to select (or insert) the lemma according to the semantic category:

Frequent abbreviations have their own lemmas, for example V – V-1`volt_:B or k: ABC k.s. – komanditní_:B_^(jen_komanditní_společnost).

Tag selection (or insertion):

An abbreviation preceding a noun is an adjective, an abbreviation following a noun is a noun. We would suggest to annotate them all as nouns (see Section 6.1.1). Does it mean that HIV in HIV virus and virus HIV have different POS.

Units called after some males person (V – volt, A – ampér, etc.), have inanimate gender. However, units using degrees (°C, °F) have masculine animate gender, because the word stupeň is always present (even if omitted in the written text). Absolute temperature uses as the unit called Kelvin (K) not degree of Kelvin. Therefore the unit has inanimate masculine gender. However, if the author uses it errorneously as degree, the tag as to be masculine animate.

If the C character is preceded by some character trying to look as the degree symbol ° (eg. -C, o C, O C), then you should mark it as an error – as lemma insert the degree^[12] symbol ° and as tag X@------------1. It should be converted into a punctuation mark.

The author's name abbreviations used in newspapers (e.g. Ber, mas, jst, ... ) have lemma as the form + -99_:B_;S and tag NNXXX-----A---8. There is X for gender because usually we do not know it. If the {M,F} gender is introduced, it should be used here. These abbreviations are not present in the lexicon, therefore you have to insert them.

Titles distinguish genders – there has to be one lemma for men, and one lemma for women (JUDr-1_:B_^(doktor_práv) vs. JUDr-2_:B_^(doktorka_práv)); to keep it consistent the masculine has number 1, the feminine has number 2. We think, the titles should have the same form for women and men. Just the tag should be different, with possibility to have X if the gender is not known (e.g. a letter subscribed as Dr. A. B.)

We tagged these words as if they were without -s and added -9 at the end.

Should not be treated as misspelling, but annotated as (colloquial) variant of official -á forms (variant '5').

Many foreign words used in Czech sentence can have different part of speech than in their original language. Usually the hint is how it behaves in different context, if it is declined as a noun, if agrees with its head, etc.

All foreign lemmas have _,t suffix.

It would be good to somehow distinguish foreign words in word or citation use.

English an should be a form of a.

Articles merged with a preposition (fra du, ita della, deu im, aufs, zur) are treated as prepositions (?Split into two words?)

Arabic short words (##)(?articles, ?prepositions) are treated as articles.

If a Slovak word has the same form as corresponding Czech one (e.g. prepositions), you should annotate it as if it were Czech. Otherwise it has to be annotated as any other foreign language.

Sometimes, foreign characters had been be screwed (e.g. Fran?oise), and therefore the morphological analyzer did not recognize the whole word. Mark it as a lemma error (do not edit the file), it has to be corrected and run thru the analyzer once more. There is a problem with letters not contained in Latin 2, they should be replaced by corresponding characters without diacritics. In the future, Unicode (2 bytes or UTF) should be considered.

Sometimes, the text contains o or I as bullets or separators. They should be marked for deletion (Press L (delete) in the lemma or tag list).

If omitted, the sentence stays grammatical. It is often possible replaceable by teprve.

Modifies functional word (should be probably TT)

It has to be modified (in the shallow syntax) by a noun in genitive. Has only two forms:

In general:

tak-2 + J^

Coordinative conjunction. If one of the clause is subordinative then tak has the meaning of an adverb: (Cf. Bál se, tak si pískal. – J^ vs. Kdyby se bál, tak si pískal – Db)

tak-3 + Db

v +

An adjective modifying a quantified expression agrees in case with the noun not the numeral.

1x. Lemma: as form. Tag Cv-------------

4x5. Should be split into three parts. E.g. 4x5 → 4, x, 5

If the hyphenated word ends with -o, and by a replacement of that -o by an adjective ending we obtain an adjective (normal or possesive), the lemma for the word is that adjective (e.g. česko-německý – česko → český, Karlo-Ferdinanova – Karlo → Karlův). Some word cannot be viewed as derived from adjectives, but rather from nouns (e.g. rap- jazzová – rap → rap vs. rapovo-jazzová – rapovo → rapový). However, the lemma of that noun cannot be used as a lemma for the hyphenated form, and a new lemma (having different number) has to be introduced.

Lemmas of possessive adjectives show how the get the noun they are derived from (see also Section 2.1.1). For example:

The base form should use -izmus ending, the form using -ismus is treated as variant '1'. Currently still some entries do not follow this convention.

Transcription of pronunciation. Lemma: as the form, tag: NNXXX

Isolated morphemes. Lemma: as the form, tag: NNXXX

Geometry. We can meet an article about a geometric theme sometimes. It means, that there occur a lot of triangles ABC, abscissas (lines) PQ, RS, AB and so on in that article. It is necessary to create a new lemma ending 98 for every mentioned figure.

Chess codes. Lemma: The code + -1_;w. Tag NNNXX-----A---8 (neuter because pole is neuter)

Crippled forms. 

This section contains especially examples of previously inserted lemmas/tag. Some of them are already in the dictionary, however they mainly serve as an inspiration, when inserting similar things.

ad hoc. ad-x_,t + RRX, hoc-x_,t + N

pele-mele . For example as a heading in a newspaper pele + TT, mele + TT

zprostředkovací vs. zprostředkovat. The morphology is rather shallow. It means, for example, that lemma for zprostředkovací is zprostředkovací (precisely zprostředkovací_ ^(*2t)), and zprostředkovat as it was in the past.