Update README.md

README.md

@@ -15,4 +15,10 @@ tags:
 
 # LaBSE-Malach-Multilabel
 
-Multilabel text classification model fine-tuned on an English subset (Malach ASR) of the Visual History Archive. Given an input string, it predicts probablities for 1065 keyword IDs from the VHA ontology. The mapping from keyword IDs to labels is not yet provided.
+A multilabel text classification model fine-tuned on an English subset (Malach ASR) of the Visual History Archive.
+Input text segments consisted of ~350 words on average.
+
+Given an input string, the model predicts probablites for 1063 keyword IDs from the VHA ontology, sorted by probability.
+Typically, probabilities >= 0.5 are "True" if encoding them in a binary vector.
+
+The mapping from keyword IDs to labels will be added to the repository.