Named Entity Discovery Demo Instructions

1. Make sure [Connected to server] is displayed on the bottom of the applet screen.
   • If the connection failed, the server component of the demo isn't running.
2. Click [Load Corpus NEs] to load the English corpus Named Entities from the server.
• NEs displayed in blue were matched with a correct transliteration by a human.
• Numbers of occurences of each NE in the English corpus are displayed in brackets.
3. Either select a corpus NE from the list or write your own in the [Custom Named Entity] field.
4. Click [Find Transliterations] to have the trained transliteration model select the highest ranking list of transliterations found in the Russian side of the bilingual corpus.
• For original NEs appearing in blue, the candidate list may contain the correct transliteration, also colored blue.
5. Click [Align Transliterations] to re-rank transliterations according to how well temporally aligned they are with the original English NE.
   • We expect that temporal alignment will often help the correct transliteration to move up the list.
   • Note that re-ranking cannot be done for a custom NE, since it has no temporal information associated with it.

Back to the applet