Review Reports - Dynamic Syntax in a Theory of Types with Records

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

See attached PDF.

Comments for author File: Comments.pdf

Author Response

Thank you for the detailed comments, which have improved the readability and quality of the paper.

The comments are listed below with our responses.

L.71: ‘section 4’ should be ‘section 3’?
- The paragraph has been rewritten
p.4 Eg (4): ‘Fo’ missing before \lambda x.arrive’(x).
- Added this.
a):p.6 eg (11): argument node has[fo=john’:e] while functor node has [fo=\lambda x.arrive’(x_:[Type,ty]] Make consistent.
- "e" has been changed to "ty"
p.6 fn. 5: The DS rule is deliberately unconstrained, leaving any constraints to be derived from the set of types involved in computational and lexical rules.
- The footnote has been removed
L. 239: page number missing from citation of Cann et al (2007) quote.
- page number has been added to the quote
P.12 eg (32): Should ANTICIPATION and ANTICIPATION appear in thE the actions string?
- The rule names have been added and the rules have been adjusted accordingly
L. 311: delete repeated string ‘the value … Cooper (2023)’.
- deleted
L. 337: ‘along’ should be ‘as long’.
- changed
L. 346: Insert ‘which’ between ‘rules’ and ‘will allow’.
- inserted 'which'
L. 403: Name of publication (and perhaps editors) missing after ‘In’.
- added name of publication
L. 404: delete ‘5 28’ from date of reference publication?
- deleted

Reviewer 2 Report

Comments and Suggestions for Authors

This paper addresses the formal finalisation of an encoding of Dynamic Syntax (DS) into Type Theory with Records (TTR).

The authors address the challenge of encoding all of DS within TTR, rather than enriching DS with TTR record types and functions, as has previously been more widely attempted, within DS-TTR (Purver et al 2010, inter alia). The authors set out the method for encoding DS grammar rules into TTR, both the computational/grammatical actions and lexical actions, and show how entire simple trees are encoded, though they do not show full derivations of an example parse.

The paper makes a valuable contribution as it commits to finally encoding how DS rules and representations can be encoded in TTR, including the computational actions on the parse state. The introduction is also compelling in terms of motivating incremental representations.

My main concerns before final publication are two-fold:

The lack of a full example of all the computational actions and lexical actions being applied appropriately with a relevant example (even to the initial example 'John arrives'). DS is primarily a process-oriented framework - showing how the parse unfolds should be made extremely clear in one set of transitions and representations, after the static presentation of the rules and trees as is already in the paper.
The somewhat terse comparison to the literature, particularly to the rather expansive work on DS-TTR since 2010 (the 'opposite' approach to the one the authors propose) which requires more than a brief comparison. I suggest at least the following work should be dealt with in the background and/or conclusion:
1. [The first time TTR was encoded in DS, for the purposes of adding speech act context] Purver, M., Gregoromichelaki, E., Meyer-Viol, W., & Cann, R. (2010, June). Splitting the ‘I’s and crossing the ‘You’s: Context, speech acts and grammar. In Proceedings of the Fourteenth Workshop on the Aspects of Semantics and Pragmatics of Dialogue (pp. 43-50).
2. [DS-TTR parsing fully incrementally modelled and explained] Purver, M., Eshghi, A., & Hough, J. (2011). Incremental semantic construction in a dialogue system. In Proceedings of the ninth international conference on computational semantics (IWCS 2011).
3. [Incorporating DS-TTR parsing within a wider TTR-encoded state] Hough, J., Jamone, L., Schlangen, D., Walck, G., & Haschke, R. (2020). A types-as-classifiers approach to human−robot interaction for continuous structured state classification. CLASP Papers in Computational Linguistics, 2, 28-40.
4. [A comprehensive overview of how DS-TTR can be applied to dialogue phenomena] Eshghi, A., Howes, C. & Gregoromichelaki, E (2022). Action coordination and learning in dialogue. In Bernardy, J-P, Blanck, R, Chatzikyriakidis, S et al (editors), Probabilistic Approaches to Linguistic Theory. CSLI Publications.

A final discussion finalising exactly how this move enriches and goes beyond some of the work above (and other DS-TTR work) would really strengthen the contribution. This is particularly relevant in that all that work above was explicitly motivated to incorporate TTR generated from DS parsing incrementally into a record-type state for dialogue and interaction. The paper, as it stands, is technically ambitious and well-explained, but needs more motivation for divergence from, and explicit comparison to, DS-TTR work.

Author Response

Thank you for the thoughtfull comments, which have helped improve the paper considerably. See below for comments are responses.

Comment 1: The lack of a full example of all the computational actions and lexical actions being applied appropriately with a relevant example (even to the initial example 'John arrives'). DS is primarily a process-oriented framework - showing how the parse unfolds should be made extremely clear in one set of transitions and representations, after the static presentation of the rules and trees as is already in the paper.

Response 1: A full example ("John arrives") has been added in Section 4.4

Comment 2: The somewhat terse comparison to the literature, particularly to the rather expansive work on DS-TTR since 2010 (the 'opposite' approach to the one the authors propose) which requires more than a brief comparison.

Response 2: A background section has been added (Section 2.2) addressing how the work presented here relates to the previous work mentioned.

Comment 3: A final discussion finalising exactly how this move enriches and goes beyond some of the work above (and other DS-TTR work) would really strengthen the contribution. This is particularly relevant in that all that work above was explicitly motivated to incorporate TTR generated from DS parsing incrementally into a record-type state for dialogue and interaction. The paper, as it stands, is technically ambitious and well-explained, but needs more motivation for divergence from, and explicit comparison to, DS-TTR work.

Reponse 3: See the added Section 2.2, which we hope clarifies the relation to previous DS-TTR work. See also the new Sections 4.7, 4.8 and the paragraph in the Conclusions starting on line 498, which all discuss different aspects of how the work presented here relates to previous work on DS-TTR, including the possibility of separating out lexical entries, the possibility of using other semantic representations, and the idea of doing without the list of actions.

Reviewer 3 Report

Comments and Suggestions for Authors

Review of “Dynamic Syntax in Type Theory with Records” – Languages, MDPI

This is a well-written paper on the formal integration of Dynamic Syntax (DS) into Type Theory with Records, offering a unified type-theoretic foundation for incremental language processing. The author proposes that Dynamic Syntax (DS) be fully reformulated within Type Theory with Records (TTR).
The idea is a simple yet an ambitious one, as DS’s incremental syntax is assumed to be expressible through declarative type-theoretic inference. I must confess that it is not entirely clear too me how far this reinterpretation goes beyond a mere reformulation. In other words, does DS change dramatically in this new framework?
The paper is highly technical, and some passages are difficult for readers who are not specialists in DS or TTR. Although the author refers the reader to Cooper’s (2023) excellent book (an open access monograph that interested readers might wish to consult) and the paper by Howes & Gibson (2021) it is sometimes hard to follow the author’s arguments fully. Being a non-expert of either DS or TTR, I must confess I had difficulties grasping some parts of the paper and I think a little more than “working knowledge of both DS and TTR” (as the author states) is required. It is undeniable that the technical sections are well-structured and consistent but the paper lacks accessibility as several key formalisms are introduced with insufficient illustration. As far as I can tell, the framework is logically rigorous, yet it might lose some of the empirical immediacy of the original model grounded in psycholinguistic motivation.
That said, the paper is interesting and deserves attention (especially from scholars interested in formal semantics and dynamic theories of grammar), as it effectively interfaces DS with TTR and seeks to recast some of the key concepts of the former within the latter.
I recommend acceptance after minor revisions.

Specific passages – much of what I suggest is the intended to make it easier for the “average” reader to get familiar with the arguments presented.

p.2 l. 54: the sentence ‘John, I grew very fond of’ contains a fronted (topicalized) prepositional object: I suggest that the author should emphasise that John has been left-dislocated for some pragmatic reason.
p. 4 tree in (4): it might be useful to add that the more common <e,t> corresponds to ‘e -> t’
p. 4 l. 126: “now suppose…” check this sentence as its syntax is funny.
p. 8 ex (16): make fonts consistent with the rest of the paper (italics is in LaTeX Computer Modern instead of Palatino).
p. 10 ex (24): the formula is particularly difficult to break down. Some more explaining is in order here.
p. 11 l. 267: the symbol \ell is introduced without mentioning it stands for ‘label’.

Author Response

Thank you for the thoughful comments, which have led us to improve the paper considerably. Responses to comments are found below.

Comment 1: I must confess that it is not entirely clear too me how far this reinterpretation goes beyond a mere reformulation. In other words, does DS change dramatically in this new framework?

Response 1: We have tried to clarify the purpouse of recasting DS in TTR in the paragraph starting on line 74.

Comment 2: The paper is highly technical, and some passages are difficult for readers who are not specialists in DS or TTR. Although the author refers the reader to Cooper’s (2023) excellent book (an open access monograph that interested readers might wish to consult) and the paper by Howes & Gibson (2021) it is sometimes hard to follow the author’s arguments fully. Being a non-expert of either DS or TTR, I must confess I had difficulties grasping some parts of the paper and I think a little more than “working knowledge of both DS and TTR” (as the author states) is required. It is undeniable that the technical sections are well-structured and consistent but the paper lacks accessibility as several key formalisms are introduced with insufficient illustration.

Response 2: We have added a background section on TTR (Section 2.1) that introduces and explains the overall approach as well as the technical details relevant to the work presented here. Regarding DS, we understand from communication with the editors that we should not include an introduction to DS in this paper, as the volume as a whole will provide such an introduction.

Comment 3: As far as I can tell, the framework is logically rigorous, yet it might lose some of the empirical immediacy of the original model grounded in psycholinguistic motivation.

Response 3: We outline the general idea that TTR should function as a foundational theory of language and cognition in the paragraph starting on line 95. We have also addressed one specific aspect of the relevance to psycholinguistics of the work presented here briefly in the paragraph starting on line 510.

Comment 4: Specific passages.

Response 4: See responses to each individual bullet point.

p.2 l. 54: the sentence ‘John, I grew very fond of’ contains a fronted (topicalized) prepositional object: I suggest that the author should emphasise that John has been left-dislocated for some pragmatic reason.
- This has been added in the paragraph starting on line 56
p. 4 tree in (4): it might be useful to add that the more common <e,t> corresponds to ‘e -> t’
- A footnote has been added.
p. 4 l. 126: “now suppose…” check this sentence as its syntax is funny.
- We commented out the whole paragraph here as we in retrospect found the discussion confusing (as well as ungrammatical).
p. 8 ex (16): make fonts consistent with the rest of the paper (italics is in LaTeX Computer Modern instead of Palatino).
- This has been fixed.
p. 10 ex (24): the formula is particularly difficult to break down. Some more explaining is in order here.
- A brief explanation has been added
p. 11 l. 267: the symbol \ell is introduced without mentioning it stands for ‘label’.
- This is included in the new TTR introduction section (Section 2.1)