Slinket - SLINK Extraction

Slinket (Slink Events in Text) is a partial modal parser which identifies subordination constructions introducing modality information in text; essentially, infinitival and 'that' clauses embedded by factive predicates ('regret'), perception predicates ('see'), predicates referring to events of attempting ('try'), volition ('want'), command ('order'), among many others. SlinkET annotates these subordination contexts as SLINKs (subordination links), the TimeML tag used to represent these kind of modal contexts. SLINKs can be of different type, depending on the modality information brought about by the relation between the embedding and the embedded predicates:

• factive: the event denoted by the embedded predicate is presupposed or entailed to be true ('John managed to leave the party.').
• counter-factive: the event denoted by the embedded predicate is presupposed or entailed to be false ('John was unable to leave the party.').
• evidential: the subordinating predicate denotes a reporting or perception event ('Mary saw/told that John left the party.').
• negative evidential: the subordinating predicate denotes a reporting event which conveys negative polarity ('Mary denied that John left the party').
• modal: the subordinating predicate introduces a reference to possible worlds ('John wanted to leave the party').

SLINK annotation follows version 1.2.1 of the TimeML specifications. An example of Slinket output is shown below, the added SLINK is in bold type.

The input to Slinket is chunked text with part-of-speech tags and event tags (chunks and pos tags are not displayed in the example above).