Linguistic pattern recognition
- 1.5k Downloads
In general, the pattern recognition task is defined as one where an infinite, continuous set of inputs is associated with a finite variety of outputs. A typical example is face recognition, where the goal is to identify the face as belonging to the same person in spite of changes in viewing angle, distance, light, makeup and hairdo, facial expression, etc. We speak of linguistic pattern recognition when the set of outputs is structured linguistically. This means both that the output units of linguistic significance follow each other in discrete time (e.g. a temporal succession of letters, words, or sentences) and that these units themselves come from a finite (or finitely generated) set. We could stretch the definition to include data that lack temporal organization. For example, the recognition of isolated characters is considered by many to be a linguistic pattern recognition task, especially in the case of Han and Hangul characters, which can be decomposed spatially though not necessarily temporally (see Sproat 2000). However, no amount of stretching the definition will allow for face or fingerprint recognition, as the output in these domains can be made finite only by imposing some artificial cutoff or limitation on the system.
KeywordsHide Markov Model Speech Recognition Content Word Function Word Topic Detection
Unable to display preview. Download preview PDF.