Many thanks. I see that the book is very recent.

Would you please provide the full citation for that book. I may want to take a look.

The constituent cannot be a TP because it has to exclude the subject. The constituent must also exclude the too*/*тоже, which again indicates that it cannot be a phrasal XP constituent. I therefore assume that it must be a bar-level constituent, and of T because it includes the finite verb. However, I do not like X-bar syntax, nor do I for a second that the strict binarity of branching assumed in most modern X-bar theoretic accounts is at all plausible. That is just not a good way to model syntactic structures.

I am not sure what you are asking. The ellipsis mechanism at work in your example is similar to what occurs in English (my knowledge of Russian is very basic): John came home. -- Me <came home> too. The ellipsis mechanism in such cases is mysterious, since I am not aware that it has been identified in the literature; at least it does not have a name that I am aware of. In any case, one could perhaps argue that came home is a constituent, but to do so, it would have to be T' constituent due to the presence of too. I think much of phrase syntax would be loathe to do that.

Your Russian examples involve ellipsis. Many ellipsis mechanisms clearly do not elide constituents (e.g. gapping, sluicing, pseudogapping, answer ellipsis, etc.), a fact that I have argued extensively elsewhere in my writings. Thus, пришел домой would not be a constituent in the sentence Я тоже пришел домой. Interestingly, though, идти домой in Я тоже буду *идти домой is arguably a constituent; as the complement of буду, the non-finite VP идти домой can and should be viewed as a complete subtree, that is, as a constituent.

I think there is ample evidence that the complements of prepositions are constituents. Since some prepositions also double as subordinators (e.g. before, after, because of), it makes sense to view the complements of subordinators as constituents, too. Note that subordinator = complementizer -- the use of terminology varies depending on the grammarian and the theory of syntax at hand. Therefore I do not think it matters so much what you want to call the constituent (i.e. IP/TP/VP or simply clause), that unit is arguably a constituent.

@Keelan. I have added the disclaimer.

I have heard that the best dependency parsers have an accuracy rate that reaches 90%, or even a little more. Your parse here is a phrase structure parse, and hence the syntactic analysis is much more complicated than the dependency parse would be. More nodes in the syntactic analysis means more potential for incorrect parsing choices.

@Ellie Xia. Concerning Sportiche et al.'s account, I note on page 24 that they conveniently and opportunistically omit de- from their structural analysis of denationalization. Why? My guess is that they did not know how to put it in the tree. If de- takes a complement as they suggest, then it should be head over nationalize. I think they were reluctant to commit to such an analysis. They perhaps sensed that de- should not be a head because it has no impact on category status; it should therefore only ever be a dependent.

@Ellie Xia. Another difficulty concerns the implication about re- in your comment. re- is a prefix that does not alter the category status of the lexical core of the word. Therefore, I do not think it can be viewed as taking a complement, but rather it simply attaches to the lexical core in the same manner as, for instance, an adverb attaching to a verb -- the verb is head over the adverb. Thus, use is head over re- in reuse.

@Ellie Xia. The point becomes clear when one considers other examples, for instance denationalization, the example that Sportiche et al. start with. The analysis should show that denational is not a word, which it can and indeed does do. But the implication, then, is that the analysis one chooses should always show which words can be acknowledged inside the greater word. Therein lies the problem.

@Ellie Xia. Right. But there is still a problem. Analysis (b) suggests that useable is not a word, which is clearly wrong, because it certainly is a word. t

Before the advent of X-bar Theory, most phrase structure syntax treated at least some (what would later become known as) X' units as phrases. This is particularly true of the finite VP string, e.g. visited Mary in Max visited Mary. That string was designated simply a VP.

Right. The issue is dense and difficult., but interesting and very much worth scrutiny Good luck.

Concerning Mel'cuk's semantic dependencies cited in the addendum to the question, that account does not consider the manner in which meaning is mapped to units of syntax. Mel'cuk's semantic level of representation is abstract. His account does not acknowledge predicates and their arguments as concrete units of actual sentence structures. I have in fact asked Mel'cuk about this directly. He stated unequivocally that his predicates exist at a semantic level of representation only.

The assumption that units of meaning can and must be mapped to constituents only is part of what has motivated the strictly binary branching phrase structures that are so prevalent among formal semanticists. Dependency grammar cannot pursue an analogous account of the semantics-syntax mapping because its syntactic structures are so minimal, acknowledging many fewer constituents. I think this is in fact an area where someone could make their mark. If a plausible DG theory of the semantics-syntax interface were developed, that would push DG ahead.