ErgSemantics_Inventory

Background

This page serves as the ‘table of contents’ (and to-do list), i.e. provides a collection of items that we believe should be described in documenting the semantic analyses of the ERG (see ErgSemantics). The initial list of candidate semantic phenomena originates from a two-step process. First, grammar entities (rules or lexical entries) were automatically inspected for indicators of ‘interesting’ semantic properties (via the discovery procedure described on the ErgSemantics/Discovery page) and grouped by their semantic synopses, i.e. the EP or set of EPs they contribute. Second, this list was paired with examples from the Redwoods Treebank instantiating the grammar entities in question.

Basic Components of Semantic Analyses

Some fundamental aspects of the MRS approach to meaning representation and its use in the ERG are independent of invididual semantic phenomena. The ErgSemantics/Basics page aims to define general terminology and introduce the relevant building blocks of ERG semantics, i.e. our version of things one might find in any logical representation language.

Semantic Phenomena Derived from Constructions

Reasoning over the grouping of grammar entities and corresponding linguistic examples, we distilled the following list of semantic phenomena, of which many correspond to one ‘cluster’ of grammar entities with identical (or very similar) synopses.

• Apposition: ‘appos’,

• Compounding: ‘compound’, ‘compound_name’, including titles and N-ed (J-N_CRD-T_C cuts across compounding and coordination)

• Conditionals: ‘if_x_then’

• Coordination: ‘conj’

• Foreign Expressions: ‘fw_seq’

• Fragments: ‘unknown’

• Imperatives: ‘pron’ (plus SFORCE)

• Implicit Locatives: (‘today’, ‘every time he arrives’): ‘loc_nonsp’

• Implicit Nominals: ‘generic_entity’

• Implicit Quantifiers: 'udef_q', ‘proper_q’, ‘pronoun_q', 'number_q'

• Instrumental Relatives: ‘with_p_rel’

• Measure Phrases: ‘measure_rel’

• Nominalization: ‘nominalization’, including verbal and nominal gerunds as well as some non-gerunds and WH nominalization (need to look at relation to free relatives)

• Non-Adverbial Clausal Modifiers: ‘subord’ (including depictives and absolutives)

• Number Sequences: ‘num_seq’ (for numeral juxtaposition)

• Parentheticals: ‘parenthetical’

• Partitives: ‘part_of’

• Quasi-Modal Infinitivals: ‘eventuality’

Semantic Phenomena Derived from Lexical Rules

Some of the phenomena listed under ‘Semantic Phenomena Derived from Constructions’ above are also instantiated, in some instances, by lexical rules. This section list those phenomena that are instantiated by lexical rules and not phrasal constructions (though possibly also by lexical types).

• Ellipsis: ‘ellipsis_ref’ (‘Can I?’), ‘ellipsis_expl’ (‘There's sun, isn't there?’)

• Passive: ‘parg_d’

• Intersective-Modifier Prefixes: ‘_co-_a_with’, ‘_counter-_a_anti’, ‘_mis-_a_error’, ‘_pre-_a_ante’, ‘_re-_a_again’, ‘_un-_a_rvrs’

• Scopal-Modifier Prefixes:

• Tag Questions: ‘id’, ‘ne_x’

• Time Expressions: ‘of_p’ ‘def_explicit_q’ ‘def_implicit_q’ (‘June third’, ‘Tuesday afternoon’)

Approximation of Phenomenon Inventory from Lexical Types

• Cardinal Adjectives and Number Names:

• Color Adjectives and Names:

• Comparatives

• Compositional Number Names:

• Control Relations

• Degree Specification (of Quantifiers, e.g. ‘nearly all’, ‘almost but not quite every’): [moved from high level concepts, because it seems more like a linguistic analysis]

• Existentials: ‘_be_v_there’

• Free Relatives: ‘free_relative_q’, ‘free_relative_ever_q’

• Generalized Quantifiers:

• Identity Copulae

• Modal Operators: ‘neg’, ‘_can_v_modal’, ‘_probable_a_1’, etc.

• MWE Quantifiers: any more et al

• Possessives:

• Pre-verbal Modifiers: [those that are syntactically analyzed as attaching to the subject; what happens to them semantically?]

• Propositional Arguments

• Pro-verbs (do so):

• Proper Names:

• Scopal Adverbs: certain uses of also, i.a.

• Verb particle constructions:

• WH Words:

Other Semantic Phenomena

Besides phenomena that can be identified by looking for ‘interesting’ constructions or lexical rules, the following is an emerging list of additional high-level semantic phenomena. [NB: Pages in this section are highly incomplete and may at this point contain only notes.]

• Resultatives (‘put the book in the box’, ‘make Abrams happy’, ‘have the report on my desk’)

• Semantically Vacuous Lexemes (i.e. In-Semantics)

• Relational nouns

• Idioms

Quasi-Semantic Phenomena

These are phenomena which are (currently) reflected in the MRS, but which either aren't directly about truth conditions or represent a way to create connected MRSs in the absence of syntactic analyses which would lead to the expected representations.

• Idiomatic Determinerless PPs: ‘idiom_q_i’

• It-Clefts: ‘_be_v_itcleft’

• Focus Movement: ‘focus_d’

• Passivization: ‘p_arg_d’

• Relative Clause Extraposition: ‘relative_mod’ (arguably should be fully parallel to non-extraposed relative clauses)

• Tag Questions: ‘ne_x’ ‘id’

• Vocatives: ‘addressee’

Semantically Vacuous Lexical Items

Another class of design decisions required in constructing a set of compositional semantic representations is the determination of which lexical items are to be treated as semantically vacuous (in the sense not contributing any elementary predications; they may yet contribute variable properties and/or serve to pass information between items in the process of composition). This section will document the semantically vacuous lexical items in the ERG. Examples include:

• Predicative copulae

• Relative pronouns

• Complementizers

• Dependent elements (both ... and, particles in verb-particle constructions)

• Fillers (um, uh, like)

Sources of inspiration

• ErgSemantics_Raw_Materials links to discussions and questions answered that could be used to further populate these pages.

More Information

• ErgSemantics Main Page

• How to Cite this Work