"RIV/49777513:23520/07:00000266" . . "3"^^ . . "2007-01-01+01:00"^^ . . "P(LC536)" . "M\u00FCller, Lud\u011Bk" . "449-454" . . . "Adaptace jazykov\u00FDch model\u016F s vyu\u017Eit\u00EDm r\u016Fzn\u00FDch t\u0159\u00EDdov\u00FDch model\u016F"@cs . "SPECOM 2007 Proceedings" . "Language Model Adaptation Using Different Class-Based Models"@en . "430520" . . "Pra\u017E\u00E1k, Ale\u0161" . . . . . . "Language Model Adaptation Using Different Class-Based Models" . "\u010Cl\u00E1nek popisuje dv\u011B rozd\u00EDln\u00E9 metody pro p\u0159id\u00E1v\u00E1n\u00ED nevid\u011Bn\u00FDch slov do LVCSR syst\u00E9mu. Ob\u011B metody pou\u017E\u00EDvaj\u00ED principy t\u0159\u00EDdov\u00FDch jazykov\u00FDch model\u016F - prvn\u00ED metoda vyu\u017E\u00EDv\u00E1 znalosti z\u00E1visl\u00E9 na \u00FAloze, druh\u00E1 metoda je pln\u011B automatick\u00E1 a na \u00FAloze nez\u00E1visl\u00E1. Rozs\u00E1hl\u00E9 experimenty navr\u017Een\u00FDch jazykov\u00FDch model\u016F na ASR syst\u00E9mu pracuj\u00EDc\u00EDm v re\u00E1ln\u00E9m \u010Dase ukazuj\u00ED, \u017Ee ob\u011B techniky poskytuj\u00ED zlep\u0161en\u00ED p\u0159esnosti rozpozn\u00E1v\u00E1n\u00ED. Nav\u00EDc p\u0159\u00EDsp\u011Bvky obou dvou metod lze kombinovat, co\u017E vede k celkov\u00E9mu absolutn\u00EDmu zlep\u0161en\u00ED a\u017E 2 %."@cs . "Moscow" . "23520" . "Language Model Adaptation Using Different Class-Based Models"@en . "ASR; LVCSR; HMM; real-time; class-based language model; morphological tag"@en . . . "Language Model Adaptation Using Different Class-Based Models" . "6"^^ . . . . "The paper presents two different methods for adding previously unseen words into the LVCSR system. Both methods employ the principles of class-based language modeling – the first one exploits task-specific knowledge, the second one is fully automatic and task independent. Extensive test of the proposed language models in the real-time ASR system showed that both techniques provide a consistent improvement in terms of recognition accuracy. Moreover, the contributions from both methods appear to be additive, yielding a total improvement of up to 2 % absolute." . "Moskow State Linguistic University" . "5-7452-0110-X" . "RIV/49777513:23520/07:00000266!RIV08-MSM-23520___" . "Moscow, Russia" . . "3"^^ . . . "Adaptace jazykov\u00FDch model\u016F s vyu\u017Eit\u00EDm r\u016Fzn\u00FDch t\u0159\u00EDdov\u00FDch model\u016F"@cs . "Ircing, Pavel" . "The paper presents two different methods for adding previously unseen words into the LVCSR system. Both methods employ the principles of class-based language modeling – the first one exploits task-specific knowledge, the second one is fully automatic and task independent. Extensive test of the proposed language models in the real-time ASR system showed that both techniques provide a consistent improvement in terms of recognition accuracy. Moreover, the contributions from both methods appear to be additive, yielding a total improvement of up to 2 % absolute."@en . . "[2D5875EAF188]" . .