...[Bac86].2.1
More information on the importance of the part-of relation can be found in [Sim87]. For events and parthood, see also [Pia96a], [Pia96b].
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... on.2.2
It will not be possible, here, to go into any depth on these questions, so we refer the reader to the available literature, among which: [Kri89], [Dow79], [Hig97], etc.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... languages.2.3
For a discussion of aspectuality, see [Gio97].
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... lack.2.4
See [Dow79] and [Lan92] for a discussion of the intensional properties of the progressive.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... contribution.2.5
For more information, see [Hig97].
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
....2.6
Microplanning is the process by which more fine grained syntactic structures for the sentence are produced from the results of discourse and text planning activities.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... storeroom2.7
Although this is not really clear from Talmy's discussion, 'motion' is used in this case to refer to what he elsewhere names translational motion, i.e. change of position, rather than to contained motion, i.e. motion along an unbounded path.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... proto-agent-cause-move-manner2.8
cf. 2.3.2, 3.10.3 for a description of the treatment chosen to encode information on thematic relations within Acquilex
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... wordnets3.1
Note that it is possible to derive the hierarchical structuring of any wordnet, including WordNet1.5, for a particular set of ILI-records using a so-called Projection function in the EuroWordNet database
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... senses3.2
WORDNET 1.5 in fact includes an experiment in representing related senses, called cousins. The scope of this experiment is however small and its results are very preliminary
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... types3.3
The basic type `tme' (time) occurs twice, grouping together two subdomains of WORDNET as one in CORELEX.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... way5.1
We have to point out here that in a fully automatic approach (compared to a manual), issues of compromise between the precision and recall come at play.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... terms5.2
Nested terms are sub-strings of other terms.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... modification5.3
By variation we understand an exchange or syntactic re-structuring of components; by modification we understand the addition of modifying words to the MWL.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... Centre5.4
For a more detailed description of the formalism see [Kar93,Tap94,Seg95].
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... matin5.5
For a related treatment of French dates see [Mau93].
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... problem5.6
A problem is AI-complete if its solution requires a solution to the general AI problems of reasoning about arbitrary world knowledge.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... disambiguation5.7
In particular the bag of words approaches mentioned below
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... expressions5.8
Multi-words expressions including: idiomatic expression such as to sweep something under the rug, phrasal verbs to spit up, compounds warning light.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... tags5.9
Or some other derived tag set.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... indications5.10
Especially considering that WordNet provides only two senses of stick together: S35 and S41.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... object5.11
The accuracy of the system for Subject assignment is 93% success for newspapers and 97% success for technical texts.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... framework5.12
There are, of course numerous exceptions: [Roc94a] used a categorial grammar parser for a text categorisation and information extraction project; [Lin95] uses a principle and parameters parser for information extraction work
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
EAGLES Central Secretariat eagles@ilc.cnr.it