This book is an introduction to the parsing scheme of the Treebank Semantics Parsed Corpus (TSPC). The TSPC is a corpus of English for general use with hand worked syntactic parse analysis for approaching half-a-million words. All annotations are discussed and illustrated with examples.

Table of contents

Chapter 1

Introduction

This book is an introduction to the parsing scheme of the Treebank Semantics Parsed Corpus (TSPC). The TSPC is a corpus of English for general use with hand worked syntactic parse analysis for approaching half-a-million words.

1.1    A hierarchy of layers

Parse analysis reveals smaller and smaller layers of structure. As a beginning, consider layers introduced by following possible paths through picture (1.1), from the highest sentence layer through to the lowest word layers, via clause and possibly phrase layers, with the arrow X ⟶ Y interpreted: ‘X immediately contains Y’. Dashed arrows indicate optionality.

(1.1)

    A sentence is at the top of the hierarchy, so it is the largest unit considered. As internal layers inside the hierarchy, a CLAUSE is reached from the outermost sentence layer, and subsequently a PHRASE can be reached from a clause layer. In addition, clauses can occur within clauses, phrases can occur within phrases and clauses can occur within phrases. A word is at the other end of the hierarchy. Chapter 2 details the range of possible words, but for considering the hierarchy, it is helpful to start from an outlook where words are of two kinds:

• +VERB words (and so immediately contained within a clause layer)
• -VERB words (non-verb words)

1.1.1    Identifying sentence and word layers

Sentence and word layers are fairly easily identified by conventions of the English writing system. Sentences are delimited by an initial capital letter and a final full-stop (or question-mark or exclamation-mark). Words are created with single character strings that are delimited by a space on each side (or punctuation mark other than a hyphen or apostrophe). But these conventions are not always followed. An important exception is when a sequence of words behaves syntactically in an idiomatic way (e.g. ‘as long as’, ‘with regard to’, ‘by the way’). Such word sequences will be analysed as if single words, with each component separated by underscore characters in a single character string, as in (1.2).

(1.2)

as_long_as
with_regard_to
by_the_way

    To allow for integration into the parse analysis, punctuation points, quotation marks, and brackets (‘.’ ‘?’ ‘!’ ‘:’ ‘;’ ‘,’ ‘-’ ‘(’ ‘)’ etc.) are treated as words; see section 2.11 for details.

1.1.2    A simple sentence

A simple sentence consists of one clause. A clause contains minimally a verb. Following picture (1.1), paths start: SENTENCE,CLAUSE. This initial clause layer is known as the matrix clause and is symbolised most generally with IP-MAT. IP is an abbreviation for Inflectional Phrase, which is another way to say ‘clause’. IP-MAT may be replaced by a more specific name. For example, IP-IMP is used instead of IP-MAT to identify a sentence as an imperative. Consider the imperative sentence (1.3), which is comprised of a single verb contained within a single clause.

(1.3)

Look!

This can be given the analysis of (1.4) from sentence layer to words, with the punctuation treated as a word.

(1.4)

    A way to write the parse information pictured in (1.4) from sentence layer to words is with (1.5).

(1.5)

IP-IMP,+VERB,Look
IP-IMP,-VERB,!
ID,101_a_saint_exupery_1943

With (1.5), each word of the parse comes at the end of its own line, with each line presenting path information from the root (IP-IMP) layer through to a word layer. The last line is the ID node for the parse.

1.1.3    A complex sentence

Consider (1.6) as an example of the interplay between internal clause and phrase layers.

(1.6)

He wanted to show what happened to anyone who tried to start trouble.

Analysis of (1.6) gives the tree structure of (1.7).

(1.7)

When viewed in terms of paths from sentence layer to words, the analysis of (1.7) becomes (1.8).

(1.8)

IP-MAT,PHRASE,-VERB,He
IP-MAT,+VERB,wanted
IP-MAT,CLAUSE,-VERB,to
IP-MAT,CLAUSE,+VERB,show
IP-MAT,CLAUSE,CLAUSE,PHRASE,-VERB,what
IP-MAT,CLAUSE,CLAUSE,+VERB,happened
IP-MAT,CLAUSE,CLAUSE,PHRASE,-VERB,to
IP-MAT,CLAUSE,CLAUSE,PHRASE,PHRASE,-VERB,anyone
IP-MAT,CLAUSE,CLAUSE,PHRASE,PHRASE,CLAUSE,PHRASE,-VERB,who
IP-MAT,CLAUSE,CLAUSE,PHRASE,PHRASE,CLAUSE,+VERB,tried
IP-MAT,CLAUSE,CLAUSE,PHRASE,PHRASE,CLAUSE,CLAUSE,-VERB,to
IP-MAT,CLAUSE,CLAUSE,PHRASE,PHRASE,CLAUSE,CLAUSE,+VERB,start
IP-MAT,CLAUSE,CLAUSE,PHRASE,PHRASE,CLAUSE,CLAUSE,PHRASE,-VERB,trouble
IP-MAT,-VERB,.
ID,154_susanne_n12

Notably, the paths of (1.8) conform to picture (1.1) above. They start with a sentence and clause layer (IP-MAT), and then can have clauses within clauses, phrases within clauses, phrases within phrases, clauses within phrases, verbs with occasional non-verb words within clauses, and non-verb words within phrases.

1.2    Outlook

Each of the units of structure in pictures (1.4) and (1.7) above, and paths of (1.5) and (1.8) above, is either a clause or a phrase. It is typical to refer to such units as constituents.

    We have so far categorised words into being either:

• +VERB words (e.g. look, wanted, start)
• -VERB words (e.g. he, to, town)

Because of this categorisation, we have already been able to distinguish clause constituents from phrase constituents:

• an immediately contained +VERB word gives a clause layer
• without an immediately contained +VERB word, a non-word unit is a phrase

    We will want to further distinguish phrases into being of various kinds based on the words they contain. To this end, we can divide -VERB words into classes such as:

• noun (N)
• adjective (ADJ)
• adverb (ADV)

With such divisions, phrases can be sub-divided into:

• noun phrases (NP) that minimally contain a noun
• adjective phrases (ADJP) that minimally contain an adjective
• adverb phrases (ADVP) that minimally contain an adverb

    Establishing the range of word classes is the aim of chapter 2. Looking at phrase classes in detail is pursued in chapter 3. A big part of chapters 2 and 3 will be to raise and answer questions of form, that is, how words and constituents look — their shape or appearance, and what their internal structure happens to be. But form is not the full story required for parsing: In order to gain motivation for having structure, we will often need to be equally concerned to explore the roles that words and constituents perform in the larger structures which contain them.

    The concern for form and roles in chapters 2 and 3 will be carried over to an exploration of clause constituents in chapter 4. Chapter 5 details options available for realising complex sentences with clause-within-clause subordination.

    The parsing practise developed over chapters 3, 4 and 5 will involve specifying form and roles for the constituents named PHRASE and CLAUSE above.

    Chapters 6, 7, and 8 integrate verb codes that signal the clause requirements of verbs.

    Chapter 9 discusses exceptional scope, control and anaphora. Chapter 10 extends coverage to constructions with displacements where full analysis can require indexing. Chapter 11 mentions loose ends.

1.2.1    Notational conventions

Table 1.1 details useful notational conventions used in this book.

1.3    About the annotation approach

This book's annotation approach builds on alternative annotation schemes for English. A central goal has been to consolidate strengths of existing methods in a single annotation scheme with a high degree of normalised structure. Ideas were taken from the following annotation schemes:

• The Penn Historical Parsed Corpora Scheme (Santorini 2016)
• The Penn Treebank Scheme (Marcus, Santorini and Marcinkiewicz 1993)
• The SUSANNE Corpus and Analytic Scheme (Sampson 1995)
• The ICE Parsing Scheme (Nelson, Wallis and Aarts 2002)

    The Penn Historical scheme, which is a substantial refactoring of the Penn Treebank scheme, has informed the ‘look’ of the annotation. This includes:

• labelling: CP and IP for clause layers; ADJP, ADVP, NP and PP for phrase layers
• label extensions after CP, IP, ADJP, ADVP, NP and PP
• the presentation of conjunction structure with CONJP layers

Label names aside, the word class tag set of the TSPC scheme is most compatible with the SUSANNE scheme. This includes tags that closely overlap with the Lancaster word class tagging systems, especially the UCREL CLAWS5 and CLAWS7 tag sets used for the British National Corpus (BNC Consortium 2005).

    A large percentage (nearly three quarters) of the Treebank Semantics Parsed Corpus (TSPC) exists as data that was converted from annotation that had been following the SUSANNE scheme.

    The SUSANNE scheme is closely related to the English grammars of Quirk et al. (1972, 1985). The ICE Parsing Scheme similarly follows the Quirk et al. grammars. In addition, ICE is notable for its rich range of features. The TSPC annotation supports the ability for many of these features to be automatically derived.

    The annotation also contains plenty that is innovative. Notably, normalised structure is achieved with internal layers at:

• clause levels (ILYR)
• noun phrase levels (NLYR)
• adjective phrase levels (AJLYR)
• adverb phrase levels (AVLYR)

These internal layers are required when their corresponding levels of structure include conjunction (CONJP) layers.

    Another area of innovation is the verb code integration of chapters 6, 7 and 8. The codes of chapter 6 classify catenative verbs; cf. Huddleston and Pullum (2002, p.1220). These are verbs of a verb sequence that are prior to the main verb of a clause and intermediate layers of structure labelled with a -CAT extension are included with the clause annotation to support this verb type. Additional codes of chapter 6 with very particular distributions are included from Hornby (1975). The verb codes for main verbs in chapter 7 are from the mnemonic system of the fourth edition of the Oxford Advanced Learner's Dictionary (OALD4; Cowie 1989).

    The OALD4 system is the apex of a long history of verb code systems integrated into English as a Foreign Language (EFL) dictionaries. Such systems were first proposed in the 1930s in pioneering work by H.E. Palmer (Palmer 1938) and A.S. Hornby (Hornby 1954); see Cowie (1999) for an historical overview and comparison of systems.

    The most innovative aspect of the annotation gives the TSPC its name: It can be fed to the Treebank Semantics evaluation system (Butler 2021). Treebank Semantics processes constituency tree annotations and returns logic-based meaning representations. While the annotation seldom includes indexing, results calculated with Treebank Semantics resolve both inter and intra clause dependencies, including cross sentential anaphoric dependencies. It is sometimes especially helpful to see meaning representations from Treebank Semantics as dependency graphs (e.g., from the on-line corpus interface) to make visually apparent connections that the design of the annotation captures.

    As an example of sending parse structures from this book to the calculation routine of Treebank Semantics, consider (1.9) as input and its result shown as a dependency graph in (1.10).

(1.9)

(1.10)

In (1.9), John is surface subject of the main clause, but, in (1.10), John is logical object ([arg1]) for deceive. Results like (1.10) unlock a fuller story of the information content the annotation captures.

    The tree of (1.11) is (1.9) with the dependencies of (1.10) included as overlaid edges.

(1.11)

Chapter 2

Words

2.1    Introduction

This chapter details the range of possible words as the bottom layer of sentence structure, while chapter 3 considers phrase layers, and chapter 4 ascends to clause layers. This unravelling of grammatical structure will show all elements as inter-related components: An understanding of one layer requires integrating knowledge from all the other layers. But discovery has to start somewhere, hence the focus here on words as foundational components.

    With this chapter, words are categorised into classes informed by how words contribute to phrases and clauses. Word analysis involves saying with a tag the word class to which a given word in context belongs. This task is known as tagging.

    Section 2.2 warns how the same character string might need different tags. One criteria for choosing a tag is to identify words that can be minimal components for phrase layers: words for noun phrases are discussed in section 2.3; words for adjective phrases, in section 2.4; and words for adverb phrases, in section 2.5. Section 2.6 covers other word types that appear under a noun phrase as modifiers.

    Section 2.7 presents tags for verbs which are minimal components for clause layers. Section 2.9 adds other word types that occur under a clause layer. Section 2.10 considers words that serve as connectives to combine layers. Section 2.11 looks at the treatment of punctuation. Section 2.12 deals with interjection, reaction signals, and formulaic expressions. Section 2.13 notes other possible elements of a parse.

2.2    Beware: Same character string, different tag

Section 1.1.1 identifies words through the conventions of the English writing system as character strings delimited by a space on each side, while noting exceptions. Word identification is not a complete word analysis: The same character string might be used as a word in different ways in different contexts. Tagging completes the word analysis. For example, down occurs as a noun (N) in (2.1), an adjective (ADJ) in (2.2), a verb (VB) in (2.3), an adverbial particle (RP) in (2.4), and a preposition (P-ROLE) in (2.5).

(2.1)

(2.2)

(2.3)

(2.4)

(2.5)

    The examples of (2.1)–(2.5) show how there might be different tags for what are different instances in different contexts of the same character string. Tag differences due to context can happen even within the same sentence. For example, that in (2.6) occurs as a subordinating conjunction (P-CONN), determiner (D), noun (N) complementizer (C), and demonstrative pronoun (D;_nphd_).

(2.6)

2.3    Words that can head a noun phrase

Certain words form the minimal component required for the phrase layers that immediately contain them. Such a minimal component is called the head of the phrase. The word class tags of Table 2.1 distinguish words that can head a noun phrase (NP; see section 3.2).

The rest of this section describes morphosyntactic properties for identifying the words of Table 2.1.

2.3.1    Suffixes that derive nouns

There are many suffixes that derive nouns from other nouns or from words of other classes. Some examples include: {age} (package, usage), {er} (officer, teacher), {ness} (illness, awareness), {ship} (championship, relationship), {tion} (action, organisation), {ty} (ability, responsibility).

2.3.2    Count nouns and plural marking

Many nouns take the suffix {(e)s} when they refer to plural items:

(2.7)

(N hedgehog)      (NS hedgehogs)
(N revelation)    (NS revelations)
(N time)          (NS times)
(N wish)          (NS wishes)
etc.

Such nouns are instances of count nouns, so called, because they can combine with numerals (e.g., three wishes). A few count nouns take the irregular suffix {(r)en} in the plural:

(2.8)

(N child)         (NS children)
(N ox)            (NS oxen)

There are also count nouns that use the same form in the singular and plural. For example, nouns for the animal species in (2.9) look like singular nouns, and so are tagged N, but can also be used as plurals (e.g., three deer); while (2.10) is tagged NS for looking like a plural, but might also refer to a single pair.

(2.9)

(N deer)
(N fish)

(2.10)

(NS scissors)

2.3.3    Mass nouns

In contrast to count nouns, there are mass (or noncount) nouns that can neither combine with numerals nor inflect for number. For example:

(2.11)

(N gold)
(N information)

It is possible for the same noun to belong to both count and mass categories: in Her hair is brown, hair is a mass noun, but in I found a hair in my soup, it is a count noun.

2.3.4    Proper nouns

Proper nouns are written with an initial capital letter. Inflection for number is rare, but is possible in talk about more than one entity with the same name, e.g., There are two Heathers in my class.

(2.12)

(NPR Heather)
(NPRS Heathers)

2.3.5    Genitive case

Both proper nouns and common nouns can be case marked — but only for genitive case — by the genitive suffix {'s} (or {'} if the noun already has the plural suffix {s}). For parsing, the genitive marking is made to form a distinct word that is tagged GENM:

This marker is placed leftmost under an NP-GEN projection that comprises the genitive content. For example:

(2.13)

(2.14)

(2.15)

Double marking of a genitive is possible when the content of the genitive has both a following genitive marker and a preceding of preposition, as in (2.16).

(2.16)

2.3.6    Reciprocal, reflexive, and personal pronouns

This section presents an overview of reciprocal and reflexive pronouns (PNX), and personal pronouns (PRO). Such words are understood in the context of their occurrence, often by taking already mentioned noun phrases as antecedents. They can appear in full noun phrase positions, with the range of grammatical markings detailed in Table 2.3.

    The reciprocal pronouns each_other and one_another are used to indicate a relationship between conjoined nouns, for example, the love relationship in (2.17).

(2.17)

Reflexive personal pronouns can indicate that the subject and object are the same entity, as in (2.18).

(2.18)

Reflexive forms are also used for emphasis, as seen with (2.19), where myself is annotated as an adverbial noun phrase with reflexive role (NP-RFL).

(2.19)

    A look across the columns of Table 2.3 shows that the ‘person’ category is useful when differentiating the various reflexive and personal pronouns. The first person refers to the speaker(s), the second person refers to the hearer(s), and the third person refers to other entities. Person is further distinguished into singular and plural reference. Further distinction is possible with third person singular pronouns (reflexive and personal), since they inflect for gender: This gives: masculine he/him(self)/his, feminine she/her(self)/hers, and neuter it(self)/its.

    A look down the rows of Table 2.3 shows that pronouns are marked for more grammatical cases than common nouns or proper nouns. In addition to genitive case, pronouns are marked for nominative case (marking the subject of the clause: I shrugged), and accusative case (marking the object acted on by the verb: James tickled me).

(2.20)

(2.21)

    Genitive pronouns can be either:

• dependent (Genitive 1 of Table 2.3), i.e., standing for the possessor only, and appearing with a following noun (for example, our friends)

(2.22)

• independent (Genitive 2 of Table 2.3), i.e., standing for the whole NP, so occurring without a following noun (for example, it is theirs)

(2.23)

As shown by examples (2.22) and (2.23), with parsing, a genitive pronoun (dependent (PRO;_genm_) or independent (PRO;_ppge_)) is the only element of an NP-GEN layer. Furthermore, for PRO;_ppge_, the NP-GEN layer is the only element of the containing noun phrase.

2.3.7    Other types of pronouns

Other types of pronouns found in English include:

• Interrogative pronouns (tagged WPRO) are used in questions, and stand for the entity under question, for example, Who came?, What did it mean?

(2.24)

(2.25)

• Relative pronouns (tagged RPRO) introduce a relative clause, for example, an iron hook which he unfastened.

(2.26)

• Indefinite pronouns with quantification (tagged Q;_nphd_, while quantifiers are tagged Q; see section 2.6) come in two broad types: compounds, for example, as everyone knows, nobody knows, and words which often occur with the preposition of, such as, all of them, none of them.

(2.27)

(2.28)

(2.29)

(2.30)

• Indefinite pronouns not subclassified as Q;_nphd_ (tagged D;_nphd_, while indefinite determiners are tagged D; see section 2.6), for example, in case somebody comes, but also some of them. The indefinite pronoun one, for example, as one would expect, typically occurs in formal contexts. The pronoun you (always tagged PRO) can also be used indefinitely, for example, as you can see.

(2.31)

(2.32)

(2.33)

(2.34)

• Demonstrative pronouns (tagged D;_nphd_; see Table 2.4 for the full range) indicate a specific referent in terms of the distance (real or metaphorical) from the speaker, for example, I don't want that. I'll have these, Demonstratives also behave as definite determiners (tagged D) which modify nouns within a noun phrase, for example, this point (see section 2.6).

(2.35)

(2.36)

(2.37)

2.4    Words that can head an adjective phrase

For a word to be an adjective, it must be able to contribute the head of an adjective phrase (ADJP; see section 3.3), often as the only element of the phrase. Adjectives are tagged as in Table 2.5 to distinguish comparative and superlative forms from general forms.

    The comparative form of an adjective is typically indicated by the suffix {er}, whereas the superlative form is typically indicated by the suffix {est}; see Table 2.6.

2.5    Words that can head an adverb phrase

For a word to be an adverb, it must be able to contribute the head of an adverb phrase (ADVP; see section 3.4), often as the only element of the phrase. Adverbs are tagged as in Table 2.7 to distinguish comparative, superlative, and WH froms from general forms.

2.6    Other noun phrase level words

Section 2.3 above has already discussed words that can head a noun phrase. Table 2.8 gives tags for other words that can be immediate components of a noun phrase, but that can't be a noun phrase head. Rather, these words precede the head within noun phrases and modify the head in terms of definiteness, item under question or quantity.

    Prototypical nouns can take one of two articles:

• a (/an) or the in the singular
• some or the in the plural

Articles for the singular head noun wish and the plural head noun wishes are seen in (2.38).

(2.38)

(D a) (N wish)                   (D the) (N wish)
(D some) (NS wishes)             (D the) (NS wishes)

2.7    Verbs

Verbs occur at clause levels of structure in the annotation. There are tags to subclassify verbs in accordance to their form:

• lexical verbs (VBP, VBD, VB, VAG, VVN) (discussed in section 2.7.1)
• HAVE verbs (HVP, HVD, HV, HAG, HVN) (discussed in section 2.7.2)
• BE verbs (BEP, BED, BE, BAG, BEN) (discussed in section 2.7.3)
• DO verbs (DOP, DOD, DO, DAG, DON) (discussed in section 2.7.4)
• modal verbs (MD) (discussed in section 2.7.5)

2.7.1    Lexical verbs

Verbs can change in shape to show tense. For example, the verb SUPPORT in John supports Peter takes a third person present tense {s} inflection, while in John supported Peter it has a past tense {ed} inflection. A verb that has tense is called a finite verb.

    Tenseless forms of verbs are called nonfinite verbs, which are comprised of:

• infinitive forms
• present participle forms
• past participle forms

Note that participle forms are tenseless despite their full names! Infinitive forms occur in infinitive clauses, often preceded by the infinitive marker to (e.g., John happened to support Peter.). Present participles are used in the progressive construction (e.g., John is supporting Peter). Past participles are used in the perfect construction (e.g., John has supported Peter) and the passive construction (e.g., Peter is supported by John).

    The distinctions in verb forms just sketched are captured in the tags for lexical verbs of Table 2.9.

    Table 2.10 further illustrates with examples the distinctions between the different lexical verb forms. This includes examples of irregular verbs that do not have a regular past tense inflection.

2.7.2    HAVE

The forms of HAVE are tagged as in Table 2.11.

HAVE has the non-contracted inflections of Table 2.12.

2.7.3    BE

The forms of BE are tagged as in Table 2.13.

Table 2.14 presents an overview of the eight different non-contracted forms of BE. This is the widest range of distinct forms for the same verb lexme in English, with extra person-number contrasts in the past and present tenses.

2.7.4    DO

The forms of DO are tagged as in Table 2.15.

DO has the the non-contracted inflections of Table 2.16.

2.7.5    Modal verbs

Modal verbs express meanings such as certainty, ability, or obligation. The main modal verbs are WILL, WOULD, CAN, COULD, MAY, MIGHT, SHALL, SHOULD, MUST and OUGHT. A modal verb only has finite forms and has no suffixes (e.g., I sing — he sings, but I must — he must). Modal verbs are tagged as in Table 2.17.

2.8    Subject indicating words

2.9    Other clause level words

Besides verbs, other clause level components are words with the tags of Table 2.19.

    We can see some of these clause level components in the annotation of (2.39). This begins with there (EX) to create an existential construction, and includes the negative clitic particle n't (NEG;_clitic_). The IP-INF-CAT as selected complement of the existential verb (BED;_ex_cat_Vt_; see section 8.2.5) includes infinitive marker to (TO).

(2.39)

    It is also possible for a word tagged RP to occur as a clause level component. The RP tag is used to mark adverbial particles (e.g., up, off, out) and was seen in section 2.5 as the tag for a word that can head an adverb phrase. When an RP tagged word occurs as a clause level component it is part of a phrasal verb, as in (2.40).

(2.40)

2.10    Connective words

So far we have considered words that serve as components of either phrases or clauses. There is a further class of words with the tags of Table 2.20 that serve as the means to connect phrases and clauses.

When phrases and clauses are connected they are said to be complex.

2.11    Punctuation

Punctuation points, quotation marks, and brackets (‘.’ ‘?’ ‘!’ ‘:’ ‘;’ ‘,’ ‘-’ ‘(’ ‘)’ etc.) are treated as words for the purposes of word tagging with the tags of Table 2.21.

This makes punctuation part of a sentence in its own right. When creating constituent structure, punctuation is placed as high as possible. For example, a full stop that ends a sentence is treated as the last constituent of the highest clause layer (IP/CP/FRAG).

2.12    Interjection, reaction signals, and formulaic expressions

Interjections, reaction signals, and formulaic expressions are treated as single words with the tags of Table 2.22. They have a high placement in structure, typically occurring as elements of clause or fragment layers.

(2.41)

(2.42)

2.13    Other tags

Tags for other possible elements of a parse are given in Table 2.23.

Chapter 3

Phrases

3.1    Introduction

The phrase is an intermediate level of organisation between a word and the clause. Phrases can be characterised partly by their external role and partly by their internal form. Table 3.1 gives an overview of typical external roles for the phrases discussed in this chapter.

    By “form” is meant the way the structure of the phrase is made up of other, smaller elements. Each phrase is made up of a head word and any dependent(s). The head word may be the only element of the phrase. A phrase is named after the word class of the head word.

3.2    Noun phrase (NP)

This section considers the internal form and external roles of noun phrases.

3.2.1    NP heads and ellipsis

A constituent is structurally complete when not reduced by ellipsis. Ellipsis is the ommission of one or more elements that can be recovered or understood from the linguistic or situational context. All structurally complete noun phrases have a head word. Section 2.3 gives the list of tags for elements that can head noun phrases, including proper nouns and pronouns.

    As an example of noun phrase head ellipsis, consider (3.1), with its predicative noun phrase (NP-PRD) lacking a head noun. In a context discussing siblings, Alan is understood to be the youngest sibling.

(3.1)

While there is no additional marking to indicate the noun ellipsis in the annotation of (3.1), the overall structure gives this information, since there is a noun phrase projection that does not contain a head word.

3.2.2    NP dependents

A structurally complete noun phrase can consist of one head word (e.g. the pronoun we or the noun pizza), or it can consist of a head word with a number of dependents. Dependents modify the head word by adding information.

    Possible dependents include: determiners (e.g. the pizza), adjective phrases (e.g. healthy pizza), nominal modifiers (nouns which modify other nouns, e.g. vegetable pizza), and relative clauses (e.g. pizza that I made). It is possible to have multiple dependents. For example, (3.2) is a noun phrase containing determiner the, adjective phrase healthy, nominal modifier vegetable, head noun pizza, and relative clause that I made.

(3.2)

    With both nouns (e.g. vegetable) and adjective phrases (e.g. healthy) able to act as modifiers in a noun phrase, other criteria is needed to distinguish nouns that are nominal modifiers from single word adjective phrases. For example, adjectives do not take number marking, while vegetable can be made plural (vegetables).

    A dependent in the NP can also be categorised according to its position in the NP:

• Determiners (D), quantifiers (Q), genitive noun phrases (NP-GEN), numerals (NUM), adjective phrases (ADJP), nominal modifiers (N or NS), and some adverb phrases (ADVP), occur before the NP head (e.g., the young printer's apprentice, the computer skills course, every day, about the first week in June).

(3.3)

(3.4)

(3.5)

(3.6)