A Hierarchical Evaluation Methodology in Speech Recognition
Gespeichert in:
| Titel: | A Hierarchical Evaluation Methodology in Speech Recognition |
|---|---|
| Autoren: | Gábor Gosztolya, András Kocsor |
| Weitere Verfasser: | The Pennsylvania State University CiteSeerX Archives |
| Quelle: | http://www.inf.u-szeged.hu/~kocsor/publications/Papers/2005/folyoirat/Paper-2005-Acta-GG/Web/GKo05.pdf. |
| Bestand: | CiteSeerX |
| Schlagwörter: | search methods, multi-stack decoding |
| Beschreibung: | In speech recognition vast hypothesis spaces are generated, so the search methods used and their speedup techniques are both of great importance. One way of getting a speedup gain is to search in multiple steps. In this multipass search technique the first steps use only a rough estimate, while the latter steps apply the results of the previous ones. To construct these raw tests we use simplified phoneme groups which are based on some distance function defined over phonemes. The tests we performed show that this technique could significantly speed up the recognition process. |
| Publikationsart: | text |
| Dateibeschreibung: | application/pdf |
| Sprache: | English |
| Relation: | http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.95.6170 |
| Verfügbarkeit: | http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.95.6170 http://www.inf.u-szeged.hu/~kocsor/publications/Papers/2005/folyoirat/Paper-2005-Acta-GG/Web/GKo05.pdf |
| Rights: | Metadata may be used without restrictions as long as the oai identifier remains attached to it. |
| Dokumentencode: | edsbas.8032A4BC |
| Datenbank: | BASE |
Nájsť tento článok vo Web of Science | Abstract: | In speech recognition vast hypothesis spaces are generated, so the search methods used and their speedup techniques are both of great importance. One way of getting a speedup gain is to search in multiple steps. In this multipass search technique the first steps use only a rough estimate, while the latter steps apply the results of the previous ones. To construct these raw tests we use simplified phoneme groups which are based on some distance function defined over phonemes. The tests we performed show that this technique could significantly speed up the recognition process. |
|---|