The TOSCA/ICE Grammar

A Quick Guide to the TOSCA/ICE Grammar used in our parsed corpora.

The TOSCA/ICE Grammar is the annotation scheme used in the parsed ICE-GB and DCPSE corpora.

It is a phrase structure grammar based on Quirk et al. (1985). Each node in the tree consists of three elements. By default, ICECUP displays these in the following locations.

Both corpus tree nodes and Fuzzy Tree Fragment nodes are arranged in the same way.

The figure above is of a single node FTF drawn left-to-right. Here is an example tree drawn in the same way.

What do the different labels mean? What is a ‘PU’ or an ‘OD’? What does ‘prd’ stand for? This page is a quick reference.

Much more information is available in the online Help file supplied with the corpus. Below is only a very brief outline.

Our ICECUP software is designed from the ground up around the problem that all users have to ‘learn the grammar in order to explore the grammar’. So: don’t be put off by the apparent complexity of the grammar. The software is very forgiving and you will learn by trying things out for yourself!

Outline

Individual terms form phrase structure trees. The scheme can be thought of as operating at two levels.

1. Part of speech tagging

Consider the sentence in ICE-GB/DCPSE above, I think that's fascinating (S1A-002/DI-A02 #28). This is tagged as follows:-

Part of speech tags, or ‘wordclass tags’, classify words into types. Here there are three types: pronoun, verb, and adjective. They may also include features. Here we have the following: personal singular pronoun, monotransitive present tense verb, demonstrative singular pronoun, copular present tense enclitic verb and -ing participle adjective. Note that ’s (= is) is treated as a separate word for the purposes of analysis.

Part of speech tagging doesn’t record the structure of the sentence.

We rectify this by fully parsing the sentence. The part of speech tags sit at the ‘leaves’ of the phrase structure tree. 

The illustration shows the phrase structure of this particular sentence. Leaf nodes annotating words contain the wordclass information. So, from bottom to top (right to left in the figure) —

• I (PRON(pers, sing)) is the head (NPHD) of the first noun phrase (NP). This is the subject of the main clause which is also the parsing unit (PU).
• think (V(montr, pres)) is the main verb (MVB) in the verb phrase (VP). This has the function label ‘verbal’ (VB), and is also found in the main clause. The features ‘monotransitive’ (montr) and ‘present’ (pres) are copied to (‘percolate up to’) the verb phrase and host clause.
• that (PRON(dem, sing) is the head of a second noun phrase (NP). This forms the subject of a dependent clause (CL(depend)) that is the direct object (OD) of the main clause.
• ’s (= is, V(cop, pres, encl)) is the main verb (MVB) in the second, copular, verb phrase (VP). The features ‘copular’ and ‘present’ percolate up to the verb phrase (VP) and host clause, i.e. the dependent clause.
• fascinating (ADJ(ingp)) is the head of an adjective phrase (AJP), which is a ‘predicative’ (prd) subject complement, i.e. it adds a predicate to the subject that. 
• The short pause ‘<,>’ is also tagged and attached to the tree.

Note that some of the features are hidden in this view for reasons of space. The small triangle in the feature sector indicates that some are not shown.