The semantics of the lemon model
One of the most important aspects of a lemon model is the connection between the semantics contained within the lexicon, and the semantics in the ontology. Before we can hope to handle this difficult task we divide the tasks that are intended to be performed into three main groups
- Lexical: These are tasks that belong purely to the lexical layer of lemon and can be processed without strong semantic knowledge
- Tokenization: Identifying words and components
- Chunking: Identifying multi-word elements
- POS-Tagging: Identifying labels for each token/chunk.
- Lemmatization: Decomposing terms into inflectional components.
- Parsing: Building a parse tree over a sentence.
- Coreference: Deducing which elements in a sentence share a reference. (N.B., this task can often be considered to include weak semantic knowledge as well).
- Inflection: Inverse of lemmatization
- Grammar Generation: Inverse of parsing.
- Correspondence: Involves tasks that primarily focus on the task of connecting lexical realizations to conceptual objects.
- Annotation/Word Sense Disambiguation: Converting the lexical representation to a semantic (ontological) relation
- Selection: Inverse of grounding
- Reasoning: Making higher level deductions based on an ontological representation without lexical information.