BABEL is an online Semantic Role Labeling Platform. It tags semantic roles in the English and Italian language, according respectively to the FrameNet data and the FLaIT data. BABEL uses Structured Learning approaches to tag both Frames (i.e. the intended event in a sentence) and Semantic Roles (i.e. the arguments of each Frame). Frames are classified starting from a dictionary of Lexical Units, with a multi-classification approach (i.e. by using SVM-Multiclass). For what concern the Semantic Roles, BABEL exploits sequential labeling approach based on the SVM-HMM paradigm. In particular, a sentence is seen as a sequence of tokens, whose labels are decided by a discriminative classification function that relies on the Viterbi algorithm for the final sequence labeling.
Its architecture and underlying approaches are more deeply described in:
- (2012): Structured learning for semantic role labeling. In: Intelligenza Artificiale, 6 (2), pp. 163-176, 2012.
It can tag semantic roles according to the Frame Semantics both for the Italian and English languages. It exploits a Structured Learning paradigm to label a sentence with a hybrid discriminative-generative approach.
For more information about the system and the used approaches in it feel free to contact: