Joshua Johanson Lab 7 Other Since my grammar was not getting very much coverage, I decided to add two more features. I added a feature to allow pro-dropping by inheriting from the basic-head-opt-subj-phrase. I also added a rule that allowed the inversion of the verb and subject by creating a w-to-w rule that moved the subject to the complement. This doubled my coverage, from 12 parses to 24 parses. (No overgeneration.) Run 1 is my lab before any changes. Run 2 is my lab after the preceeding changes. Run 3 is my lab after the following changes. Interrogative and Declarative Clauses Italian does not make any syntactic distinction between matrix interrogative clauses and matrix declarative clauses. The only distinction is in the intonation. I have chosen to deal with this by giving the matrix clause the prop-or-ques_m_rel value for the message. That way, each parse could be interpreted as either a statement or a question. For example, the sentence 'Dorme' (sleeps) could be interpreted as a statement that you/he/she is sleeping or a question as to if you/he/she is sleeping. However, it does make a distinction in the embedded clauses. Italian marks embedded clauses with a complementizer that takes a sentence as its complement. Some complementizers, like 'che' can only take interrogative clauses, while other complementizers, like 'se' can only take declarative clauses. Because a matrix clause is underspecified as to whether it is declarative or interrogative, it can take on the type required by the complementizer. Certain types of verbs can only take certain types of embedded phrases. When the complementizer constrains the type of complement it can take, it also constrains what type of verb selects it for its complement. Building on the previous example, with both an interrogative and declarative phrase for "dorme", the grammar correctly assigns the following judgement. Domandiamo se dorme (We ask if sleeps) *Domandiamo che dorme (We ask that sleeps) Dico che dorme (I say that sleeps) *Dico se dorme (I say if sleeps) Domandiamo is a verb that only takes an interrogative phrase. Dico is a verb that only takes a declarative phrase. Dorme, which is underspecified, is constrained to be interrogative by 'se', and is constrained to be declarative by 'che'. It can therefore only be placed with Domandiamo if it is headed by 'se' and Dico if it is headed by 'che'. By the complementizer constraining the clause to become either interrogative or declarative, it is forced to take on the right semantics. My grammar did not increase its coverage with these new features because my test suite required several more features in order to parse. However, the glossed examples and others correctly parsed and got the right semantics. Imperatives Italian has 5 forms for imperatives. It has a different conjugation than a declarative sentence. The following are examples of the 5 types of imperatives and the accompanying conjugation. 2nd person singular - Canta (You sing) 3rd person (formal 2 person) singular - Canti (You sing) 1st person plural - Cantiamo (Let's sing) 2nd person plural - Cantate (Y'all sing) 3rd person (formal 2 person) plural - Cantino (All of you sing) The pronoun is typically dropped, but not required. However, with the 3rd person, it must be used as a plural you, not with another pronoun. Canti Lei - You sing (imp) *Canti lei - She sing (imp) *Canti lui - He sing (imp) Since we did not have to do 3rd person imperatives, I choose not to implement this feature of Italian. (Plus this grammar can't make distinctions between upper case and lower case) However, since 1st person plural did not have this problem, I choose to implement this for 1st person. Because of that, I did not require imperative clauses to be second person. Right now, it won't allow 3rd person imperatives because there is no lexical rule to change it, but that could be added in the future. Italian imperatives will not allow inflections in front of the verb. Instead, they are added to the end of the verb. Since I do not have argument optionality, I have have not implemented this feature. Italian also has a different structure for negative imperatives. Again, I did not implement this feature. I implemented this by creating a new feature for verbs called FORM. Right now I have forms base and imp. For regular verbs, I added the appropriate feature at the time of the inflection. For irregular verbs, I created a verb type for irregular base verbs. I did not include irregular imperative verbs because I do not have any in my test suite. I then changed declarative/interrogative clause to only accept verbs of form base. I then created an imperative clause that will only accept verb of form imp. The imperative clause phrase gives it the command semantics. This allowed my grammar to correctly parse one more sentence. Most of my examples included negatives, 3rd person or argument optionality. It did give it a lot more parses that I guess are grammatical, but I have never thought of. For example it gave the sentence 'canta la donna' which normally means 'the women sings' the interpretation 'Sing the woman' as if the woman were a thing that could be sung (maybe the title of a song). I also got parses that I think are ungrammatical, but not sure what to do about it. For example, for the sentence 'Maria dorme e canta' I got the parse "Mary eats and (she) sings" and the parse "Mary eats and sing (imp)." I thought about requiring that the messages be the same in order to coordinate, but I don't know if that would be a good rule in general. Although this grammar is not dealing with coordination of embedded clauses, embedded clauses can coordinate different messages, for example "so se tu piaccia la ragazza e che la ragazza te piace" (I know whether you please the girl and that the girl pleases you.) is grammatical, so it doesn't seem wise to stipulate the messages on the coordination. Because of this confusion, my grammar overgenerates with coordination.