Adapting Multilingual Embedding Models to Turkish via Cross-Lingual Tokenizer Surgery and Offline Distillation
For Turkish NLP practitioners, this provides an efficient and cost-effective method to adapt multilingual embedding models to a specific language, outperforming larger models while reducing computational requirements.
The paper introduces embeddingmagibu-200m, a Turkish sentence embedding model with 768-dimensional vectors and an 8,192-token context window, achieving Pearson/Spearman correlations of 77.55%/77.45% on STSbTR, surpassing the teacher model's 73.84%/72.92%, and a mean score of 63.9% on TR-MTEB (7th out of 26 models) with 33% fewer parameters.
Sentence embeddings are a foundational component for semantic search, clustering, classification, and retrieval-augmented generation. This paper presents embeddingmagibu-200m, a Turkish-focused sentence embedding model that produces 768-dimensional L2-normalized vectors and supports an 8,192-token context window, far exceeding the 512-token limit of earlier BERT-based Turkish encoders. Instead of full pretraining, an efficient three-stage adaptation pipeline is introduced: (1) construct a Turkish-optimized multilingual tokenizer with a 131,072 vocabulary by pruning redundant tokens from the teacher's vocabulary and incorporating multilingual tokens via frequency analysis on a 40-language corpus, (2) clone a teacher embedding model while preserving transformer backbone weights and initializing a compatible embedding table for the new vocabulary via mean-composition token mapping, and (3) perform offline embedding distillation from precomputed teacher vectors using a cosine similarity objective over a balanced 40-language Wikipedia corpus. The resulting student model contains approximately 200M parameters and trains in roughly four hours on a single GPU by avoiding online teacher inference during training, at a total cost of $5-$20. Empirically, Pearson/Spearman correlations of 77.55%/77.45% are obtained on STSbTR, surpassing the 300M-parameter teacher model (73.84%/72.92%). On TR-MTEB (26 tasks), a mean score of 63.9% is achieved (7th out of 26 models), providing a competitive cost-quality trade-off with 33% fewer parameters than the teacher. To facilitate reproducibility and downstream use, all artifacts are released including model weights, tokenizer files, precomputed embedding datasets, and open-source cloning and distillation tooling.