We report a multi-month engineering program to build Sophea, a production bilingual Greek-English automatic speech recognition system. We evaluate the system against nine production gates covering Greek and English word error rate, language identification, and hallucinations on non-speech audio. Across twenty-three training iterations and two model architectures, no training-data composition passed all nine gates simultaneously. Meeting the Greek noisy-environment target required about 1,500 steps of dense domain exposure, while preserving English language identification tolerated only about 250 steps, or about 1,250 with a rebalanced mix that reduced Greek accuracy. We describe a six-stage data pipeline in which calibrating an audio-quality filter against in-domain anchors reduced the discarded share of scored Greek audio from 98.7 percent to 10.6 percent. A pre-registered ablation isolated a hallucination defect to one training-data package. A three-model ROVER ensemble increased gate coverage from 4-7 of 9 for individual models to 9 of 9 and reduced overlapping-speech WER from 53.35 percent to 37.87 percent, a 29 percent relative improvement. A separate learned per-clip arbiter over two models is listed as sophea/asr-k1 (preview) on the public Open ASR Leaderboard, with 4.26 percent average WER across eight public English test sets, and reaches 25.88 percent WER on live Greek noisy-environment traffic. We also document five cases in which a measurement tool produced a plausible but incorrect result and seven substantial approaches that were evaluated but not shipped. No model weights or training data are released; we report methodology and quantitative results only.
Figures & tables
Figure 1: Required versus tolerated training-step budgets for Greek noisy-environment exposure. The Greek noisy-environment gate needs s≥1500 steps; the English language-identification gate tolerates s≤250 (baseline mix) or s≤1250 (rebalanced mix, at the cost of stalling Greek accuracy short of its own gate). The two intervals never overlap.
Figure 2: Share of scored Greek training rows an audio-quality filter would discard: a conventional absolute threshold against one calibrated to known-clean, in-domain reference recordings scored under the same model.
Figure 3: Production gates passed (of nine) across the compact model’s seven iterations (C1 to C7). C6 replaced machine-transcribed Greek labels with human-verified consensus labels and changed nothing else.
Gate
Threshold
Value
Result
Anchor WER
≤ 3.09
3.03
PASS
Code-switch (cs)
≤ 7.72
7.42
PASS
Meeting (ITN)
≤ 14.85
14.79
PASS
Audio-LONG (ITN)
≤ 12.77
10.42
PASS
Boilerplate (BP)
≤ 18
26
FAIL
English WER (W-EN)
≤ 2.63
2.49
PASS
Table 1: L4’s internal ship-decision gate battery, distinct from the nine production gates in Table 3 . Anchor WER is a frozen read-Greek regression probe; cs is a code-switching probe specific to this campaign, not the code-switched benchmark reported elsewhere; meeting and Audio-LONG are ITN-scored internal probes, the latter on content in the domain of the data source identified as an anchor toxin above; BP is a raw boilerplate hit count and the sole gate this checkpoint failed; W-EN is English word error rate on a 500-utterance LibriSpeech test-clean probe. None of these surfaces, including “meeting,” matches a benchmark used elsewhere in this paper: this table’s Meeting (ITN) value is not the “Greek, business meetings” row of Table 3 .
Model
FLEURS
FLEURS
LibriSpeech
Real-world
Community-
Parliamentary
(English)
(Greek)
(clean)
Greek, agg.
sourced Greek
Greek
Large model
10.22%
9.13%
2.67%
15.97%
1.73%
20.81%
Turbo tier
13.02%
11.53%
2.46%
20.65%
21.61%
24.98%
Table 2: Word error rate, large production model versus the untrained turbo tier, across six evaluation surfaces ( n=200 for the three real-world Greek columns).
Figure 4: Word error rate by benchmark, large model versus the turbo tier. Turbo runs on stock weights plus decode-time tuning only; both fine-tuning attempts regressed code-switching and were not shipped.
Benchmark
Ours
Canary
Whisper
Two-model
Three-model
FT
FT
vote
vote
Greek, clean read speech (FLEURS)
8.99
8.63
5.23
6.96
3.98
Greek, TEDx talks
13.29
6.44
6.73
6.99
6.01
Greek, community-sourced speech
8.70
8.70
1.63
7.07
2.72
Greek, informal conversational speech
9.64
8.58
7.85
7.70
6.93
Greek, business meetings
20.53
18.27
15.73
17.55
16.59
Table 3: Word error rate (lower is better) and language-identification accuracy (higher is better, in percent) across three independently trained recognizers and their confusion-network combinations, on the nine production benchmarks plus five additional public and internal evaluation surfaces. “Ours” is the production bilingual model of Section 5.1 ; “Canary FT” and “Whisper FT” are Greek fine-tunes of NVIDIA Canary and Whisper large-v3 respectively, used only as ensemble members here, not as production models in their own right. Bold marks the best of the five columns shown per row. All entries are single-run point estimates with no confidence intervals; differences between columns smaller than roughly one WER point should not be treated as established (Section 11 ).
Figure 5: Sophea ASR K1 against its two individual members, average word error rate over the seven cleaned test sets of Table 4 , scored with the public Open ASR Leaderboard’s evaluation harness (K1 listed on the live board as sophea/asr-k1 (preview) since 11 September 2026).
Test set
Sophea ASR K1
Qwen far-field variant alone
Canary FT alone
AMI
8.50
8.19
13.86
AMI, close-talk
7.29
7.27
11.63
Earnings-22 (cleaned, AA-chunked)
6.03
8.11
9.64
GigaSpeech
7.69
7.85
8.08
GigaSpeech, clean
7.65
7.80
7.96
LibriSpeech, test-clean
1.18
1.38
1.25
Table 4: Word error rate on the public Open ASR Leaderboard’s standard English test sets, using the leaderboard’s own evaluation harness, forced-English decoding; submitted as PR #201 to the leaderboard’s public repository and listed on the live board as sophea/asr-k1 (preview) since its 11 September 2026 release. “Canary FT alone” is the same fine-tune as Table 3 ; the “Qwen far-field variant” is a separate LoRA repair of the compact model (Section 5.1 ), not the C6 checkpoint used elsewhere in this paper. “Sophea ASR K1” is the learned-arbiter combination of the two, described above. Both averages use the same seven test sets for all three systems, with Earnings-22 entering as its cleaned AA-chunked split throughout, since that is the only Earnings-22 split scored for the two members (K1’s raw Earnings-22 from our own run is 8.22). The canonical seven-set average uses the raw AMI, GigaSpeech, and VoxPopuli splits, the leaderboard’s original composite; the cleaned seven-set average substitutes each dataset’s close-talk, cleaned, or AA-chunked split where the leaderboard provides one. The live board’s listed 4.26 for K1 is the board’s own eight-set average, which adds Voice Arena Monsoon (3.66), a set the two members were not scored on. Bold marks where Sophea ASR K1 leads both individual members by more than the roughly one-WER-point noise floor discussed below; entries within that floor (e.g. VoxPopuli, clean-accent) are left unmarked even where the raw numbers nominally favor one side. All entries are single-run point estimates with no confidence intervals; differences smaller than roughly one WER point should not be treated as established (Section 11 ). The AMI and AMI, close-talk rows additionally carry the Canary fine-tune’s confirmed AMI-pretraining exposure noted in Section 6.3 , capping their external validity as an unseen-data comparison.
Figure 6: Word error rate on overlap-only speech segments from real meeting recordings, single model versus two-model confusion-network voting.
Modern TTS systems approach human quality for high-resource languages but degrade when clean speech data is scarce. Modern Greek exemplifies this, lacking the curated corpora behind state-of-the-art synthesis. We propose a data curation recipe that transforms audiobook recordings into TTS-ready data via WhisperX alignment and filtering. Then we fine-tune Parler-TTS (880M), a prompt-based multilingual model whose pre-training encodes phonetic priors transferable to Greek. During development, we find that LLM-generated style prompts introduce speaker drift at inference. Replacing them with deterministic prompts resolves this, and a speaker-specific LoRA stage trained on 3.5 h of single-speaker data anchors identity while updating ~5% of parameters. Our system achieves WER 10.7% (2.9 above the ASR floor), MOS-I 4.00 (vs. 4.36 human speech), and near-human speaker consistency (MOS-C 4.24 vs. 4.30), showing that robust single-speaker Greek TTS is achievable with limited curated data.
Institute for Language and Speech Processing, Athena R.C., Greece · Department of Digital Medicine, University of Bern, Switzerland · School of Electrical and Computer Engineering, National Technical University of Athens, Greece
Speech recognition in real time in far-field conversational settings still incurs significant errors. Identifying potential error spans during streaming ASR permits a recourse mechanism where the speaker can be asked to repeat the affected content under cleaner acoustic conditions. We propose PARDON, a new framework for online error localization, that identifies spans likely to benefit from this intervention. Our supervision uses clean audio to estimate how the recognizer's output might change under improved acoustic conditions. We probabilistically align noisy and clean decodes using a pair-HMM to obtain soft supervision. These soft targets train a counterfactual gain scorer by combining a suite of acoustic and language features to produce token-level scores, that are further grouped into spans for targeted recourse. An LLM uses the ASR transcript of the repetition to revise only the selected span. At a precision of 70%, PARDON achieves the highest recall by identifying up to 52% more erroneous tokens compared to the strongest baseline. Targeted repetition yields relative WER reductions of up to 24% under a budget that accounts for both interruption frequency and the lengths of re-uttered spans.
Effective retrieval-augmented generation across bilingual Greek--English applications requires embedding models capable of capturing both domain-specific semantic relationships and cross-lingual semantic alignment. Existing multilingual embedding models distribute their representational capacity across numerous languages, limiting their optimization for Greek and failing to encode the morphological complexity and domain-specific terminological structures inherent in Greek text. In this work, we propose ORPHEAS, a specialized Greek--English embedding model for bilingual retrieval-augmented generation. ORPHEAS is trained with a high quality dataset generated by a knowledge graph-based fine-tuning methodology which is applied to a diverse multi-domain corpus, which enables language-agnostic semantic representations. The numerical experiments across monolingual and cross-lingual retrieval benchmarks reveal that ORPHEAS outperforms state-of-the-art multilingual embedding models, demonstrating that domain-specialized fine-tuning on morphologically complex languages does not compromise cross-lingual retrieval capability.
Ioannis E. Livieris, Athanasios Koursaris, Alexandra Apostolopoulou +2
Novelcore, Athens, GR 10436 · Department of Business Administration & Organization Administration, University of Peloponnese, Kalamata„ GR 24100