Update Premove ITN model metadata

README.md

@@ -1,19 +1,32 @@
 ---
 library_name: premove-itn
+language:
+  - en
+base_model:
+  - microsoft/deberta-v3-large
 tags:
   - inverse-text-normalization
+  - text-normalization
+  - speech-processing
   - voice-agents
+  - structured-prediction
+  - candidate-ranking
+  - safetensors
   - custom-code
 license: mit
 ---
 
-# Premove ITN Contextual v0.1.0
+# Premove ITN v0.1.0
 
-This is the first frozen, inference-only artifact for Premove ITN's
-structured-value contextual scorer. It contains the complete trained model
-parameters in `model.safetensors`, the exact DeBERTa configuration, and the
-tokenizer files used by the selected production checkpoint. Optimizer state,
-schedulers, counters, and training datasets are not included.
+Premove ITN is an English inverse text normalization system for voice-agent
+utterances. This release is the first frozen, inference-only artifact for its
+structured-value contextual scorer. It ranks deterministic Rust candidates
+using sentence context, then applies exact maximum-score decoding.
+
+The artifact contains the complete trained parameters in `model.safetensors`,
+the exact DeBERTa configuration, and the tokenizer files used by the selected
+production checkpoint. Optimizer state, scheduler state, training counters,
+training data, and evaluation rows are not included.
 
 The artifact is loaded through the `premove-itn` package. It is not a generic
 Transformers model; use the package loader so candidate generation and scoring
@@ -28,22 +41,67 @@ tokenizer = artifact.tokenizer
 ```
 
 The model uses `microsoft/deberta-v3-large` at the exact revision recorded in
-`provenance.json`. The base model is referenced rather than redistributed.
-Microsoft's model card lists the base model and tokenizer under the MIT
-license; retain that attribution when redistributing this artifact.
+`provenance.json`. The artifact includes the tokenizer and base configuration;
+the base model is not separately redistributed. Microsoft lists the base model
+under the MIT license; retain that attribution when redistributing this
+artifact.
 
 This release is tagged `v0.1.0`. Do not move or overwrite that tag after
 publication. The artifact digest and source checkpoint lineage are recorded in
 `provenance.json`.
 
-## Scope
+## Intended use
+
+Use this artifact for English voice-agent text normalization where spoken
+numbers, dates, times, money, phone values, identifiers, URLs, and related
+structured values require context-sensitive disambiguation. The model receives
+only the utterance text at inference time. Candidate metadata is generated by
+the package's deterministic Rust layer.
+
+## Supported span kinds
+
+The supported span kinds are listed in `config.json`: `DIGIT_SEQUENCE`,
+`CARDINAL`, `TIME`, `DATE`, `MONEY`, `DECIMAL`, `PHONE`, `ELECTRONIC`,
+`MEASUREMENT`, `ORDINAL`, `PUNCTUATION`, `WHITELIST`, and `WORD`.
+
+## First Evaluation
+
+The retained First Evaluation used 1,500 frozen VoiceAgent ITN rows and three
+backends. The metrics below are semantic entity metrics unless marked strict.
+Formatting-only differences can reduce strict exact match while preserving the
+same semantic value.
 
-The supported span kinds are the values listed in `config.json`. This artifact
-selects deterministic Rust candidates with a contextual scorer. It does not
-contain training data or the frozen VoiceAgent benchmark dataset.
+| Backend | Overall semantic micro | Overall semantic macro | Voice-agent entities | Mean latency |
+|---|---:|---:|---:|---:|
+| Premove ITN | 89.70% | 90.71% | 398/400 (99.50%) | 56.49 ms |
+| Thutmose | 59.39% | 57.61% | 268/400 (67.00%) | 15.98 ms |
+| text-processing-rs | 55.79% | 55.05% | 273/400 (68.25%) | 0.14 ms |
+
+Premove leads the measured semantic accuracy overall and on the 400
+voice-agent-domain rows. It does not lead latency. This is a balanced synthetic
+stress benchmark, not an estimate of production traffic accuracy, and the run
+was not blind.
+
+## Limitations
+
+- This custom scorer is not directly loadable with `AutoModel.from_pretrained`.
+  Use `load_inference_artifact` and the matching `premove-itn` package version.
+- The artifact is English-only and depends on the package's candidate graph and
+  decoder for end-to-end normalization.
+- Benchmark results are not a guarantee for unseen domains or formatting
+  policies. Human gold adjudication and broader production-distribution testing
+  remain separate requirements.
 
 ## Release identity
 
 - Artifact version: `v0.1.0`
 - Architecture: `premove-candidate-scorer-v1`
 - Package version: `0.1.0`
+- Hub repository: `premove-itn/premove-itn`
+- Hub revision: `v0.1.0`
+- Source checkpoint SHA-256: `9021fa11a028faefb31ef67878170cbe29ed25e68a9a78999f37b120c2ad00d5`
+- Inference artifact SHA-256: `119c0f19767b61446e04da1f8f01a001edf97a47a66965e7146db2483b4937a1`
+
+## Licensing
+
+Premove ITN is MIT licensed. The pinned base model is `microsoft/deberta-v3-large`, whose model card lists an MIT license. Review and retain both notices when redistributing this artifact.