A novel weighting technique for combining likelihood scores in language identification systems

Fusion is one of the key research issues in modern Language Identification (LID) systems. In this paper we compare existing fusion techniques for LID systems and propose an alternative. By directly utilizing language-dependent performance information, a novel Language-Dependent Weighting (LDW) approach is introduced and implemented. The language-dependent weighting coefficients are directly derived from pair-wise LID performances on development dataset. The advantage of using language-dependent weighting over language-independent weighting is illustrated using a Language- Dependent Performance Map (LDP-MAP). Experiments on NIST LRE 2003 task and OGI database demonstrate that the proposed fusion technique outperforms other recent fusion techniques when the amount of available development data is limited.