FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W33873003> ?p ?o ?g. }

• W33873003 endingPage "560" @default.
• W33873003 startingPage "543" @default.
• W33873003 abstract "In many successful character recognition systems, a character is first normalised (e.g. aligned in position), then preprocessed, (e.g. by feature extraction), and then classified. To say that a three-stage machine of this sort could recognise unconstrained hand-printed 0, ‥,9,A, ‥,Z is to state an hypothesis. In 1971 this is an hypothesis and not a proven fact. Scientifically, this hypothesis is not allowable, since it can never be disproved. To disprove it we would have to try out all possible normalisation schemes, all possible preprocessors and all possible classifiers, which is of course impossible. Therefore, it is possible that the usual methodology in which we concentrate either on normalisation, or on preprocessing, or on classification, will never lead to a machine which reliably recognises unconstrained hand-printed 0, ‥9, A, ‥Z and if this is true we shall never find out that this is true. This consideration, and the fact that much work has already been based on the three-stage hypothesis, has prompted a search for an alternative approach to character recognition, wherein we do not divide the recognition machine a priori into functional blocks whose details are to be worked out later. Instead, system development is to be broken down into manageable steps by other means. According to the following suggestion, the recognition problem is idealised, as a first step. Besides varying in position, size, orientation, and to a limited extent in general perspective, hand-printed characters are subject to variations which we cannot exactly specify a priori. Because of the limitations of a human's grasp of complexity, we may not wish to provide a machine with a specification of these details by means of intuitive trial and error, but instead by means of an automatic process based on training sets. If this automatic process is to be successful when characters are subject to non-trivial variations, then it should work also when characters are subject only to variations in position. It is suggested that we should not use our knowledge of variation in position in designing the machine, but instead, as a preliminary exercise, we should make the machine learn to recognise variations in position in the same way that it will (we hope) learn non-trivial variations which we are unable to specify. In character recognition it is usual to eliminate variations in position by normalisation (alignment), and then to concentrate on non-trivial variations. The present paper suggests the reverse of this: that as a first step we eliminate non-trivial variations by considering an hypothetical idealised problem in which characters belonging to any given recognition class vary only in position. This suggestion has not yet lead to good practical recognition performance, but it has contributed towards the achievement of a very limited experimental demonstration of transference of learning between recognition classes. If training sets of the letters A, ‥Z are used to improve the recognition of numerals 0, ‥9, this exemplifies transference of learning." @default.
• W33873003 created "2016-06-24" @default.
• W33873003 creator A5033454591 @default.
• W33873003 date "1972-01-01" @default.
• W33873003 modified "2023-10-16" @default.
• W33873003 title "IDEALISED PATTERN RECOGNITION" @default.
• W33873003 cites W113049248 @default.
• W33873003 cites W1995426156 @default.
• W33873003 cites W2000263793 @default.
• W33873003 cites W2005080020 @default.
• W33873003 cites W2024533723 @default.
• W33873003 cites W2027875806 @default.
• W33873003 cites W2067958620 @default.
• W33873003 cites W2068639469 @default.
• W33873003 cites W2079690395 @default.
• W33873003 cites W2091451843 @default.
• W33873003 cites W2115739330 @default.
• W33873003 cites W1496270511 @default.
• W33873003 doi "https://doi.org/10.1016/b978-0-12-737140-5.50031-2" @default.
• W33873003 hasPublicationYear "1972" @default.
• W33873003 type Work @default.
• W33873003 sameAs 33873003 @default.
• W33873003 citedByCount "3" @default.
• W33873003 crossrefType "book-chapter" @default.
• W33873003 hasAuthorship W33873003A5033454591 @default.
• W33873003 hasConcept C10138342 @default.
• W33873003 hasConcept C111472728 @default.
• W33873003 hasConcept C119857082 @default.
• W33873003 hasConcept C12713177 @default.
• W33873003 hasConcept C138885662 @default.
• W33873003 hasConcept C153180895 @default.
• W33873003 hasConcept C154945302 @default.
• W33873003 hasConcept C161191863 @default.
• W33873003 hasConcept C162324750 @default.
• W33873003 hasConcept C198082294 @default.
• W33873003 hasConcept C23123220 @default.
• W33873003 hasConcept C2524010 @default.
• W33873003 hasConcept C2776401178 @default.
• W33873003 hasConcept C2777855551 @default.
• W33873003 hasConcept C2780861071 @default.
• W33873003 hasConcept C33923547 @default.
• W33873003 hasConcept C34736171 @default.
• W33873003 hasConcept C41008148 @default.
• W33873003 hasConcept C41895202 @default.
• W33873003 hasConcept C75553542 @default.
• W33873003 hasConcept C88548561 @default.
• W33873003 hasConceptScore W33873003C10138342 @default.
• W33873003 hasConceptScore W33873003C111472728 @default.
• W33873003 hasConceptScore W33873003C119857082 @default.
• W33873003 hasConceptScore W33873003C12713177 @default.
• W33873003 hasConceptScore W33873003C138885662 @default.
• W33873003 hasConceptScore W33873003C153180895 @default.
• W33873003 hasConceptScore W33873003C154945302 @default.
• W33873003 hasConceptScore W33873003C161191863 @default.
• W33873003 hasConceptScore W33873003C162324750 @default.
• W33873003 hasConceptScore W33873003C198082294 @default.
• W33873003 hasConceptScore W33873003C23123220 @default.
• W33873003 hasConceptScore W33873003C2524010 @default.
• W33873003 hasConceptScore W33873003C2776401178 @default.
• W33873003 hasConceptScore W33873003C2777855551 @default.
• W33873003 hasConceptScore W33873003C2780861071 @default.
• W33873003 hasConceptScore W33873003C33923547 @default.
• W33873003 hasConceptScore W33873003C34736171 @default.
• W33873003 hasConceptScore W33873003C41008148 @default.
• W33873003 hasConceptScore W33873003C41895202 @default.
• W33873003 hasConceptScore W33873003C75553542 @default.
• W33873003 hasConceptScore W33873003C88548561 @default.
• W33873003 hasLocation W338730031 @default.
• W33873003 hasOpenAccess W33873003 @default.
• W33873003 hasPrimaryLocation W338730031 @default.
• W33873003 hasRelatedWork W1502614025 @default.
• W33873003 hasRelatedWork W2066259560 @default.
• W33873003 hasRelatedWork W2262783296 @default.
• W33873003 hasRelatedWork W2380927352 @default.
• W33873003 hasRelatedWork W2391959412 @default.
• W33873003 hasRelatedWork W2728578317 @default.
• W33873003 hasRelatedWork W3108285770 @default.
• W33873003 hasRelatedWork W3119026898 @default.
• W33873003 hasRelatedWork W3197541072 @default.
• W33873003 hasRelatedWork W823110065 @default.
• W33873003 isParatext "false" @default.
• W33873003 isRetracted "false" @default.
• W33873003 magId "33873003" @default.
• W33873003 workType "book-chapter" @default.