TUKE System for MediaEval 2014 QUESST
Two approaches to QbE (Query-by-Example) retrieving system, proposed by the Technical University of Kosice (TUKE)for the query by example search on speech task (QUESST), are presented in this paper. Our main interest was focused on building such QbE system, which is able to retrieve all given queries with and without using any external speech resources. Therefore we developed posteriorgram-based keyword matching system, which utilizes a novel weighted fast sequential variant of DTW (WFS-DTW) algorithm in order to detect occurrences of each query within the particular utterance file, using two GMM-based acoustic units modeling approaches. The first one, referred as low-resource approach, employs language-dependent phonetic decoders to convert queries and utterances into posteriorgrams. The second one, defined as zero-resource approach, implements combination of unsupervised segmentation and clustering techniques by using only provided utterance files.
PDFTasks
Datasets
| Task | Dataset | Model | Metric Name | Metric Value | Global Rank | Benchmark |
|---|---|---|---|---|---|---|
| Keyword Spotting | QUESST | TUKE g-zero late submission(for the development set) | Cnxe | 0.971 | # 41 | |
| MinCnxe | 0.922 | # 46 | ||||
| ATWV | 0.106 | # 17 | ||||
| MTWV | 0.107 | # 18 | ||||
| Keyword Spotting | QUESST | TUKE p-low late submission (for the development set) | Cnxe | 0.948 | # 34 | |
| MinCnxe | 0.854 | # 34 | ||||
| ATWV | 0.191 | # 12 | ||||
| MTWV | 0.191 | # 13 | ||||
| Keyword Spotting | QUESST | TUKE g-zero(for the development set) | Cnxe | 0.974 | # 44 | |
| MinCnxe | 0.934 | # 50 | ||||
| ATWV | 0.091 | # 18 | ||||
| MTWV | 0.091 | # 19 | ||||
| Keyword Spotting | QUESST | TUKE p-low(for the development set) | Cnxe | 0.960 | # 36 | |
| MinCnxe | 0.892 | # 41 | ||||
| ATWV | 0.161 | # 13 | ||||
| MTWV | 0.162 | # 14 |
