Semantic Web Machine Reading with FRED

The authors did in my opinion a very good job at improving the paper; the first few sections have been restructured and it is now much easier to grasp Fred's capabilities as well as its overall architecture. I also welcome the additional experimental results, which give a much clearer picture of the performance of the tool. Finally, the revised version contains a much more compelling conclusion, as it now includes a discussion on open issues and on lessons learnt. Overall, I suggest to accept the manuscript at this stage.

Short comments on the review dimensions defined for 'Tools and Systems Reports' (1) Quality, importance, and impact of the described tool or system (convincing evidence must be provided): this is in my opinion an important and high-quality tool. The pieces of evidence provided in terms of impact are largely informal; however, impact is always very difficult to measure for such academic tools and I feel like Fred already had much more impact than comparable pieces of software. (2) Clarity, illustration, and readability of the describing paper: the authors improved on those points; The paper is my opinion much clearer and more readable in this revised version.

The authors addressed the issues mentioned in my first review but the paper has been so significantly augmented and modified that it requires a new review. Overall, the new version is more convincing and better explained.

I have new remarks that should be easily addressed, so I request a minor revision before publication:

The first example given is not very demonstrative because it is ambiguous. Had it been an example of the ambiguities of human language, it would be fine. But is seems that its goal is simply to provide a simple starting exmaple to show what FRED is doing. The ambiguity is that "Valentina loves movies with Brad Pitt" can be understood as "Valentina loves the movies in which Brad Pitt is playing a role" or "Valentina loves movies when she is watching them with Brad Pitt". It is quite natural to interpret "loves" as a binary relation between the lover and the loved, e.g., " ". The first interpretation simply puts a constraint on what can be the object of such relation and can be written in OWL:

[
a owl:Restriction;
owl:onProperty ex:stars;
owl:hasValue ex:BradPitt
]
rdfs:subClassOf
[
a owl:Restriction;
owl:onProperty [owl:inverseOf ex:loves];
owl:hasValue ex:Valentina
] .

The love relation could also be interpreted as an instance of a Love class, but in any case, the fact the movie is starring Brad Pitt is not a characterisation of the relation, it characterises the film:

[] a ex:Love;
ex:lover ex:Valentina;
ex:loved ex:FightClub .
ex:FightClub ex:stars ex:BradPitt .

and the sentence can be translated into an OWL axiom:

[
a owl:Restriction;
owl:onProperty ex:stars;tripl
owl:hasValue ex:BradPitt
]
rdfs:subClassOf
[
a owl:Restriction;
owl:onProperty [owl:inverseOf ex:loved];
owl:someValuesFrom [
a owl:Restriction;
owl:onProperty ex:lover;
owl:hasValue ex:Valentina
] .

However, if the sentence is interpreted in the second way (Valentina loves movies when she is with Brad Pitt), the Brad Pitt characterise this particular love relation, not the movie, in which case it makes sense to model it like in Fig.1. In absence of context information, it is not possible to determine automatically whether the first or the second interpretation is correct, so FRED may very well produce the latter. Unfortunately, for a typical reader of our time, it is unlikely that the sentence be interpreted as if Valentina was hanging out with Brad. It results in something unnecessarily surprising, and the peculiarity of the model produced by FRED is not discussed in the paper at this point.

"Some examples are ... etc." -> no "etc."

In VerbNet, Love_31020100 is a relationship between an Experiencer and a Stimulus, not a Theme. Here, it seems that the movie is the stimulus.

Sec.2.1:
"The skolemization of the vent occurrence" -> of the event?

In Fig.3 (b) and (c), the class fred:People appears. Considering the convention used in writing class names as noons in singular form, this should be interpreted as the class of peoples. Those who experience movies are individual people (i.e., persons), not peoples (like tribes, etc).

Sec.2.2,
"""
"John did not go to school by car" can be interpreted as either as "There is no event in which John went to school by car" or "There is an event in which John went to school, but not by car"
"""
This is not an "either/or" dichotomy. The second interpretation entails the first one. In any case, if John did not go to school by car, there was no event in which he went to school by car. The second interpretation is simply adding context, or assumption to the meaning of the sentence. One could also interpret the sentence as "John usualy goes to school by car, but not this time", or differently. This misses the point, in my opinion.

Why is "should" a form of necessity?

Sec.2.3, "In fact, the adjectives in the two example sentences have the same syntax" -> no, in "Roberto Bolle is Italian", the adjective "Italian" is a predicative adjective, separated from the noun phrase it characterises by a verb, while in "Valentina is Giovanni's fifth daughter", the adjective "fifth" is an attributive adjective immediately attached to the noun daughter.

Sec.4.3, "performance of each such an application" -> of each such application

Sec.5, "expressed in triplet-based" -> triple-based

Sec.6, "This references demonstrate" -> These references

Ref.:
[22] "nlp" -> capital missing

Authors have significantly improved the paper, and addressed all my comments very well. The paper now much more clearly describes FRED, its purpose and functionality, and its work pipeline. The diagrams have been greatly improved as well, and therefore enhanced the readability of the paper as a whole. This is a valuable tool for the community.