Semantic interpretation

For NLP purposes, semantic interpretation can be viewed as the task of translating a natural language into a formal meaning representation language (MRL).

$$\mbox{NL} \Rightarrow \mbox{MRL}$$

Minimally, we require of any MRL that:

1.

it have a well defined semantics,

2.

it be unambiguous (in contrast to NLs),

3.

it support inference.

Examples of potentially suitable MRLs include:

• a formal logic (e.g., the predicate calculus),
• a database query language (e.g., SQL),
• a semantic network language (e.g., KL-ONE).

Ambiguous NL phrases map to multiple distinct representations:

NatCorp export nuts

1.

export(natcorp,nuts1)

2.

export(natcorp,nuts2)

Rules of inference apply:

$$\mbox{\em employs\/}(\mbox{\em NatCorp\/},\mbox{\em John\/})$$

$$\forall x. \mbox{\em employs\/}(\mbox{\em NatCorp\/},x) \rightarrow \mbox{\em located\/}(x,\mbox{\em London\/})$$ ------------------------------

$$\mbox{\em located\/}(\mbox{\em John\/}, \mbox{\em London\/})$$