Search results

Further evaluation of USFM marker placement algorithm

researchResearch topicsResearch topics

Status: Open.

#728 In sillsdev/silnlp;

· isaac091 opened on May 13, 2025

Lora no longer works in silnlp

bugSomething isn't workingSomething isn't working

pipeline 4: trainIssue related to training a model.Issue related to training a model.

pipeline 6: inferIssue related to using a trained model to translate.Issue related to using a trained model to translate.

researchResearch topicsResearch topics

Status: Open.

#688 In sillsdev/silnlp;

· laura-burdick-sil opened on Mar 25, 2025

Use LLM to revise back translations

researchResearch topicsResearch topics

Status: Open.

#616 In sillsdev/silnlp;

· johnml1135 opened on Jan 3, 2025

Can we reliably improve the output of Back Translations by sending the drafts to an LLM for correction.

researchResearch topicsResearch topics

Status: Open.

#576 In sillsdev/silnlp;

· davidbaines opened on Oct 29, 2024

Parse front and back matter and extra books according to their stylesheets.

researchResearch topicsResearch topics

usfmUSFM parsing issueUSFM parsing issue

Status: Open.

#483 In sillsdev/silnlp;

· davidbaines opened on Aug 14, 2024

Investigate mapping token embeddings from source to target

researchResearch topicsResearch topics

Status: Open.

#481 In sillsdev/silnlp;

· mshannon-sil opened on Aug 13, 2024

Research methods of adding tags during training and using them during translation.

researchResearch topicsResearch topics

Status: Open.

#423 In sillsdev/silnlp;

· davidbaines opened on Jun 19, 2024

Provide a method for training models where the text is not already known by NLLB.

researchResearch topicsResearch topics

Status: Open.

#422 In sillsdev/silnlp;

· davidbaines opened on Jun 19, 2024

Automated aligned translation candidates

researchResearch topicsResearch topics

Status: Open.

#402 In sillsdev/silnlp;

· jcuenod opened on Jun 4, 2024

Compare NLLB and MADLAD-400

researchResearch topicsResearch topics

Status: Open.

#398 In sillsdev/silnlp;

· isaac091 opened on May 31, 2024

Research default probability for adding missing characters to unigram tokenizers

pipeline 3: preprocessIssue related to preprocessing.Issue related to preprocessing.

researchResearch topicsResearch topics

Status: Open.

#348 In sillsdev/silnlp;

· isaac091 opened on Mar 22, 2024

Preserve Spanish « open and closing » double quote marks in the training data.

researchResearch topicsResearch topics

Status: Open.

#297 In sillsdev/silnlp;

· davidbaines opened on Jan 12, 2024