Neosophie
Contact

Blog

Articles about Neosophie's product development and technology.

#asr#benchmark#books#dataset#diarization#finetune#llm#metrics#oss#rag#tts
Are WER and CER Not Enough? Re-evaluating Japanese ASR with Noun and Proper Noun F1 Scores og preview
asrbenchmark

Are WER and CER Not Enough? Re-evaluating Japanese ASR with Noun and Proper Noun F1 Scores

Japanese speech recognition is more challenging than many other languages due to the wide variety of kanji and proper noun representations. This primary benchmark article implements a morphological-analysis-based F1 score evaluation using Sudachi to quantify "kanji and proper noun recognition accuracy" that WER and CER struggle to capture, and re-evaluates 9 open-source ASR models.

Neosophie Team

Building a High-Quality ASR Training Data Pipeline with Multiple Models — Design Philosophy and Implementation of Whisper × Qwen3 × LLM og preview
asrdataset

Building a High-Quality ASR Training Data Pipeline with Multiple Models — Design Philosophy and Implementation of Whisper × Qwen3 × LLM

A deep-dive into a two-stage pipeline that integrates Whisper and Qwen3-ASR transcriptions via LLM diff-merging to generate ASR training data that surpasses what any single model can achieve. The pipeline combines Whisper's strength in kanji and proper nouns with Qwen3's lower overall CER, then applies Forced Alignment to assign word-level timestamps. The downstream chunk-splitting stage automates natural 20–30 second chunk generation using sentence-end scoring, followed by Whisper re-transcription for CER-based quality filtering.

Neosophie Team

How I Finetuned IBM Granite Speech 1B on Japanese Audio and Improved CER from 0.37 to 0.14 og preview
asrfinetune

How I Finetuned IBM Granite Speech 1B on Japanese Audio and Improved CER from 0.37 to 0.14

I finetuned IBM Granite Speech (`granite-4.0-1b-speech`) on 100 hours of Japanese speech data and reduced CER from 0.37 to 0.14. The official script's Projector+LoRA-only training has a ceiling on accuracy gains. The key breakthrough was additionally training `lm_head` and the last 8 layers of the Language Model. The result matches Qwen3-ASR-1.7B (CER 0.14) with only 1B parameters.

Neosophie Team

Replaced Embedding RAG with PageIndex on My Own Site — Here's What Actually Happened og preview
rag

Replaced Embedding RAG with PageIndex on My Own Site — Here's What Actually Happened

After hitting a wall with hybrid BM25 + Embedding + Rerank search (wrong retrievals, high infra cost, Japanese embedding model limitations), I implemented **PageIndex** — a vector-free RAG architecture — as the AI assistant for my company site. The result: no VectorDB, no embedding model, no Reranker needed. Two-stage LLM-based filtering via a URL tree navigated to the correct pages for queries like *"What products do you have?"* and *"What ASR-related blog posts exist?"* — zero vector math involved.

Neosophie Team