Getting started[]
- Video instructions etc.
- Complete treebanking guidelines
- The Aeneid at nlp.perseus.tufts.edu (the server where treebanking is enabled)
- Note that you will only see the syntax links if you browse the text by "card" (as opposed to simply entering the line number into the box).
- Our Perseus usernames: tw (Tarik), jsedlak (James), aheydrick (Aubree), nguyena (Thao), cuomos (Shelby), dunnej (Jameson), cgarcia (Christian), smithk (Katie), mangelosanto (Matt), wajdamarygrace (MaryGrace), gribbell (Laura), hollyhavel (Holly), Meshach (Meshach)
Some initial examples from Aeneid 1.195-219[]
Tips[]
- Please don't forget to choose a lemma (which dictionary-entry each word is derived from) and a morphological identification for each word before beginning the graphical editing process. (This is covered in the ten-minute introductory video.)
- Tree banking is all about the syntax (function) of the words. Morphology is secondary to this. Syntactic functions may be filled by cases and parts of speech different from what you expect (nominative SBJ, accusative OBJ, etc.). For example:
- an accusative can be SBJ (as in accusative + infinitive);
- an ablative can be SBJ (in the ablative absolute his dictis, his is the subject of dictis);
- a relative clause's head-word may serve to specify (ATR), may provide the SBJ or OBJ of another verb, or may serve an ADV function (quam ob rem...);
- a participle may be attributive (ATR) or circumstantial (ADV);
- the "predicate nominative" function (PNOM) can even be filled by a genitive (guidelines, p. 25: es bonae voluntatis, "You are of good will").
- One thing treebanking requires us to think about is whether expressions used with verbs "are particular to the one verb they appear with" (OBJ) or "can really apply to almost any verb" (ADV). See David Bamman on OBJ vs. ADV, where the phrases in asterisks (*with wine*, etc.) are OBJ and "yesterday" is ADV. Our more memorable in-class example of an "always optional" (Guidelines, p. 14) ADV was "in bed."
- Why do we annotate the punctuation? David Bamman provides this answer: "Even if punctation marks don't exist orthographically, they still signal the presence of real syntactic phenomena. Veni vidi vici has exactly the same syntactic structure as veni et vidi et vici and if we don't have commas (or overt coordinators like et or -que) to have specific tokens on which the coordination can depend, then we have to resort to using ellipsis (annotating the hidden structure) which is much more complicated (though equally valid). Annotating the punctuation helps simplify the annotation."
- How to tag an ellipsis (external dependency). To get these results, I used the table editor and entered the head and relation manually for the words dependent on the word left out of the Latin. (See my archive, "O qui res hominum...")
- Participles in a nutshell: attributive participles are ATR dependent on nouns; circumstantial participles are ADV dependent on verbs. The latter rule means that the a circumstantial participle and the (pro)noun it agrees with may not be linked in the diagram. For example, in Aeneid 1.240, insequitur (PRED) has both viros (OBJ) and actos (ADV) depending on it; actos does not depend on viros because it is adverbial/circumstantial, not adjectival/attributive.
- The guidelines do not clearly address adjectives used with genitive or dative, as Aeneid 1.239, fatis contraria fata. From pp. 16-17 of the guidelines, you might think that ATR is the only choice. However, we seem to have been right to say that fatis is OBJ of contraria. (I confirmed this to my satisfaction from how Mambrini annotates Aeschylus Supp. 497, ὅμοιον Ἰνάχῳ.) (It is a bit confusing that objective genitives like metas rerum shouldn't be treated in the same way, but Guidelines, p. 16, clearly specify ATR. I suggest you understand this by considering that "limits" could be of many things, and the objective genitive specifies much as any other ATR.)
- With Aeneid 1.259f., sublimemque feres ad sidera...Aenean, a sort of OBJ-vs.-ADV question comes up, translating into OCOMP-vs.-ATV. I suggest that ATV is correct for sublimem. It may seem similar to facio omnia nova (OCOMP) (Guidelines, p. 25), but the difference is that "I make all things..." necessarily raises the question, "What do you make them?" This tells us that an object of some kind (OBJ/OCOMP) is needed to complete the meaning. On the other hand, feres Aenean is complete by itself, and the ATV sublimem is but one of many ways we can add something adverbial.
Keeping track of our own and others' annotations[]
Each of our accounts is enabled to see all of the others' work, once it has been archived. Once you have annotated a sentence and archived it, it will show up at http://nlp.perseus.tufts.edu/hopper/annotate/courses/all.jsp.
When more than one of us have created diagrams of the same sentence, they can be viewed together for "reconciliation" at http://nlp.perseus.tufts.edu/hopper/annotate/courses/reconcile.jsp. Here you will see an editable annotation, and also a comparison showing the differences between the versions. Once your work is done on the editable version, you can save it to your list like any other tree (it will remain in your reconcile.jsp until you archive it from your own list). (Unfortunately, it seems that sometimes reconciliation may not immediately show anything when there are multiple archived annotations of the same sentence by our group. So you may have to bypass use of the reconciliation feature. Remember that you can continue to edit and re-save your own archived contributions. Anyway, reconciliation does not offer graphical editing, so you may well prefer the graphical environment.)
http://nlp.perseus.tufts.edu/hopper/annotate/index.jsp is the URL for viewing your own list.
Practice student treebanking: Aeneid 1. 223-246[]
These are viewable by students in the class in our shared archive.