Comments from reviewers for GL 2009

Review 1

The paper studies a class of nouns that the authors label as “unsaturated”: nouns that need semantic arguments in order to be interpreted. The topic is intensively studied in current linguistic research, in studies interested in the predicative and aspectual properties of nominals and nominalizations, the interplay of semantic and distributional properties in the definition of major lexical classes (verbs vs. nouns), the syntactic and semantic description of different elements within the predicate and bigger linguistic units (frames, scenes), etc. Also, the issue of argument-taking nouns fits very well into the range of topics proposed for the conference, among other things, because it proves that the qualia structure is an important source of proper arguments and co-arguments of nouns. Several important issues related to semantically unsaturated nouns are dealt with, such as their definition and identification, degrees of (un)saturatedness, verbal and nominal properties of this class of nouns, the behavior of unsaturated nouns in hypernym hierarchies, etc. The general impression is that this is a work in progress, since most of the mentioned issues need further research. On the one hand, this is understandable at least in some cases, given the current state of the art on the topic. On the other hand, a minimal explanation of the most relevant points would be appreciated. For instance, a “working hypothesis” is put forward in section 3.3.3, according to which “the clearer the qualia structure of the noun has, the stronger its co-arguments are”, but no motivation is given for this statement.

Comments on concrete points:
  1. Definition of ‘semantically unsaturated noun’ in section 2.3: “class of nouns that require co-arguments”. It is further proposed to apply this term to the following classes of nouns: relational nouns, nouns with strong co-arguments, deverbal nouns and nouns derived from adjectives and adjectivals. It is not clear why deverbal nouns are included here in view of the definition of ‘co-argument’ given in section 2.2: “Noun A has co-arguments {B1, …, Bn} relative to predicate P iff B and B1, …, Bn are arguments of P and P is a LU for frame F”. This definition does not work with many deverbal nouns, which inherit the argument structure of the verb and have PROPER arguments therefore. The difference between both kinds of arguments is captured by Gentner and Kurtz (2005) in terms of ‘role categories’ and ‘schema categories’.
  2. Quantitative definition of unsaturatedness in section 2.3.3. It is difficult to assess the validity of the heuristics proposed to measure the unsaturatedness because it has not been tested.
  3. A more coherent organization of the material would be appreciated. For instance, it is not clear why section 3.2 (Modifiers as sources of unsaturatedness), which deals with general properties of unsaturated nouns, is included in section 3 (“Classification of Japanese nouns and verbs”).

References: Gentner, D., & Kurtz, K. (2005). “Learning and using relational categories”, in W. K. Ahn, R. L. Goldstone, B. C. Love, A. B. Markman & P. W. Wolff (Eds.), Categorization inside and outside the laboratory. Washington, DC: APA.

Review 2

The paper is not well organised and is not so easy to follow the reasoning in it. There are also statements that are rather obvious. And some parts are very confused.

The topic is interesting and there may be some good ideas, but they are exposed in such a confused way that it is impossible to follow them.

It should be checked for the language.

Review 3

This seems to be an interesting paper with something to say about so-called "unsaturated" nouns. However, I found the presentation to be very difficult to follow, even circular at times (e.g., specification of the need for co-arguments seems to be used to define "unsaturated" nouns, and vice versa). In general, the organization is awkward and topics seem to be introduced and left somewhat randomly. One example is the long explanation of the hypernym chains and cleaning of the automatically acquired data at the outset, which is itself never used in the paper (it obviously motivates the work, but the long description is unnecessary and misleading).

That said, the paper does have some interesting points that are uncovered after re-reading and trying to fit the pieces together. If the paper were to be re-written for clarity and organization, it may well be very good indeed.