Multilingual Coref

[AngledLuffa]

multilingual coref!

• Added models which cover several different languages: one for combined Germanic and Romantic languages, one for the Slavic languages available in UDCoref Coref engineering #1406

new features

• streamlit visualizer for semgrex/ssurgeon Ssurgeon viz 3 #1396
• updates to the constituency parser ensemble Revamp the constituency parser ensemble #1387
• accuracy improvements to the IN_ORDER oracle Con oracle #1391
• Split-only MWT model - cannot possibly hallucinate, as sometimes happens for OOV words. Currently for EN and HE Apostrophe bug for Stanza English tokenizer model #1417 Mwt unks #1419
• download_method=None now turns off HF downloads as well, for use in instances with no access to internet Parameter for loading transformer & tokenizer with local_files_only #1408 Am I properly using stanza offline (coref English model - Electra Large)? #1399

new models

• Spanish combined models Incorrect Spanish verb decomposition #1395
• Add IACLT knesset to the HE combined models
• NER based on IACLT
• XCL (Classical Armenian) models with word vectors from Caval

bugfixes

• update tqdm usage to remove some duplicate code: Remove code duplication caused by not using tqdm disable option #1413 3de69ca
• long list of incorrectly tokenized Spanish words added directly to the combined Spanish training data to improve their tokenization: Single words tend to be over-segmented in Spanish, resulting in non-word tokens #1410
• Occasionally train the tokenizer with the sentence final punctuation of a batch removed. This helps the tokenizer avoid learning to tokenize the last character regardless of whether or not it is punctuation. This was also related to the Spanish tokenization issue 56350a0

---

This discussion was created from the release Multilingual Coref.