ESD Test Suite Examples

Abrams arrived Tuesday morning.
Abrams arrived June 3rd.

Linguistic Characterization

Time expressions in English exhibit a wide range of idiosyncracies, both lexical and syntactic, but also conform to several sub-regularities which the grammar captures and normalizes to a high degree. Time expressions in the ERG encompass, among others, day and month names (Tuesday, June), days of the month (the third), season and holiday names (Spring, Easter), times of day (morning), and numbered hours and minutes (ten past four, four fifty). From these building blocks, complex time expressions can be derived in a number of ways, where the grammar aims to assign parallel semantics to different syntactic constructions where they are close parapraphses

Two of these lexical rules have semantic effects, illustrated in the examples above. Both rules relate a lexical entry for a name of a month or a day-of-week to a derived entry which supplies two quantifiers, to bind the instance variables of the name’s predicate and the neighboring word’s predicate, and also a two-place of_p relation linking the inherent arguments of these two predicates.

One lexical rule, used for e.g. Tuesday in Tuesday morning, relates a lexical entry for the name of a day of the week to the derived entry supplying quantifiers and the of_p relation, so that when it combines with the word for a part of a day, such as morning or night, a well-formed semantics emerges which is very similar to the paraphrase in on the morning of Tuesday, the second.

Similarly, a second lexical rule, for e.g. June in June third, relates a lexical entry for the name of a month of the year to the derived entry supplying the quantifiers and the of_p relation, ready to combine with the semantics of the word for a day of the month, to produce a semantics which corresponds closely to the alternative phrasing the third of June.

Motivating Examples

The following are examples involving this phenomenon:

  • The morning of Tuesday, the third, was cloudy.
  • The third of June was cloudy.

ERS Fingerprints

Analyses of these two types of time expressions are characterized by (1) the EP of_p, which is a relation between two individuals, one for a named day or month, and the other for a day of the month or part of a day; and (2) two quantifiers, a def_implicit_q for the named entity, and a def_explicit_q for the day of month or part of a day.

  of_p[ARG1 x1, ARG2 x2]
  dofm_or_mofy[ARG0 x2]
  def_explicit_q[ARG0 x1]
  def_implicit_q[ARG0 x2]

Note that the of_p also identifies its label with that of the relation whose ARG0 is its ARG1, as with other intersective modifiers.

Interactions

The primary motivation for this lexical rule analysis is to minimize the differences between the semantics of these expressions and their variants with an overt of-PP, as illustrated above.

Reflections

It might be worthwhile to further minimize the differences between the pairs of paraphrases, by replacing the def_implicit_q predicate with proper_q.

I believe we should greatly extend our test suite for this phenomenon; time expressions are well worked out in the ERG, and there is a non-trivial price we pay for (almost) always aiming to disambiguate different usages of, for example, numbers. At the same time, normalizing time expressions is important to many applications.

Open Questions

Expert External Commentary

Grammar Version

  • 1214

References

Flickinger, Dan (1996). “English Time Expressions in an HPSG Grammar,” in T. Gunji, ed., Studies on the Universality of Constraint-based Phrase Structure Grammars, pp. 1–8. Ministry of Education, Science, and Culture, Japan.

More Information

Last update: 2015-09-22 by StephanOepen [edit]