[ Skip to the content ]

Institute of Formal and Applied Linguistics Wiki

[ Back to the navigation ]

You are here: start » draft

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision 		Previous revision
Next revision Both sides next revision
draft [2009/07/15 13:54]
ptacek 		draft [2009/07/15 15:26]
ptacek
Line 1: Line 1:
 ====== Progress Report ====== ====== Progress Report ======
  
- +[[Progress Report]] - dal jsem to na zvlastni stranku, abysme si nelezli do zeli
-Hi Marc, +
- +
-... +
- +
-Re: progress: there is progress in the following: +
- +
-- evaluation of the ASR performance using the WoZ data (WP5.1) +
-- language model re-training for the collected dialogue data (using also data sources external to COMPANIONS) (WP5.1) +
-- implementation of the real-time speaker adaptation (WP5.1) +
-- additional dialogue transription for ASR is ongoing (WP52.? T5.2.1) +
-- DM has been transferred from USFD to Prague (WP5.3) +
- being extensively tested +
-- DAF editor transfer is complete (WP5.3) +
-- Sample dialogues (specifically aimed at the demo) +
- are ready - issues are being resolved between CU/ZCU +
-- DAFs are being prepared for the SC-CZ scenario AND +
- the sample dialogues +
-- DA set is being prepared, also based on the sample dialogues (WP5.2) +
-- preliminary DA tagger (on std DAMSL-SWBD tagset) working (~35% error rate) (WP5.2) +
--  +
-- integration work is ongoing (CU/ZCU, internally at CU) +
- but no functioning full demo yet (beyond what we've presented in Madrid) +
- +
-I hope this is OK for the progress report. Pavel (I.) might add more specifics regarding the ASR and especially TTS progress. +
- +
-Best, +
- +
--- Jan +
- +
- +
  
  
Line 38: Line 7:
 ====== Description of Czech Companion November Prototype ====== ====== Description of Czech Companion November Prototype ======
  
-The Czech version of Companion deals with the Reminiscing about User's Photos scenario taking advantage of date recorded in first phase of the project. The basic architecture is same as of the English version, i.e. set of modules communicating through the Inamode Relayer (TID) backbone;+The Czech version of Companion deals with the Reminiscing about User's Photos scenario taking advantage of data recorded in first phase of the project. The basic architecture is same as of the English version, i.e. set of modules communicating through the Inamode Relayer (TID) backbone;
 however the set of modules differs (see Figure 1). Regarding the physical settings: the Czech version runs on two notebook computers connected by local network; one can be seen as a Speech Client, running modules dealing with ASR,TTS and ECA, second as an NLP Server. however the set of modules differs (see Figure 1). Regarding the physical settings: the Czech version runs on two notebook computers connected by local network; one can be seen as a Speech Client, running modules dealing with ASR,TTS and ECA, second as an NLP Server.
  
 photopal domena, nahranej korpus, ze na to sou dafy (reusing SHEFF DM intergrated through Inamode Relayer (TID)) vhodny, moreover reusable for expected pomdp DM from UOX (reuse states, let pomdp's do the topology and handle the states transitions, post november work), photopal domena, nahranej korpus, ze na to sou dafy (reusing SHEFF DM intergrated through Inamode Relayer (TID)) vhodny, moreover reusable for expected pomdp DM from UOX (reuse states, let pomdp's do the topology and handle the states transitions, post november work),
-typy odpovedi a zpusob jejich implementace, rucne vyrobene dafy pro nasledujici topics: Person_Retired, Person_in_productive_age, Child, Husband, Wife, Wedding, Christmas, Handle_stalled_dialog+typy odpovedi a zpusob jejich implementace, 
 NLP server s tectomt, ASR/TTS/SR client, connected over network NLP server s tectomt, ASR/TTS/SR client, connected over network
 XXX JPta XXX JPta
Line 57: Line 26:
 features: improved language models, real-time speaker adaptation features: improved language models, real-time speaker adaptation
 performance indicator: WER performance indicator: WER
 +
 +
 +
  
  
  
 ===== Speech Reconstruction (WP 5.1 ???) ===== ===== Speech Reconstruction (WP 5.1 ???) =====
-features: omit filler phrases, irrelevant speech events, false starts, repetitions, corrections, syntax smoothing (WO,  +features: omit filler phrases, remove irrelevant speech events, handle false starts, repetitions, and corrections, polish word ordering 
-imlementation(zahrnout tuhle info?): moses natrenovany na korpusu +performance indicator: BLEU score between actual output and manually reconstructed sentences from corpora (T5.2.1), baseline: Moses with default settings 
-performance indicator: BLEU score (overall scoring of all features) to annotated corpora from T5.2.1., nejaka baseline + 
-XXX Mirek+ 
 + 
 + 
  
 ===== Morphology Analyzer and POS tagging (WP 5.2) ===== ===== Morphology Analyzer and POS tagging (WP 5.2) =====
-features: XXX Mirek/Johanka +features: coverage of photo-pal domain, domain adapted tagger (XXX prida nam Jarka OOV slova co najdeme, bude PDTSC rucne oznackovane - do listopadu?) 
-performance indicator: accuracy+performance indicator: OOV rate, accuracy 
  
 ===== Syntactic Parsing (WP 5.2) ===== ===== Syntactic Parsing (WP 5.2) =====
 features: induce dependencies and labels features: induce dependencies and labels
-performance indicator: f-measure+performance indicator: accuracy (correctly induced edges, labels)
 v tipu je natrenovat MacDonnalda na dialog datech, ten task je do M42, ted ne. v tipu je natrenovat MacDonnalda na dialog datech, ten task je do M42, ted ne.
 +
  
  
  
 ===== Semantic Parsing (WP 5.2) ===== ===== Semantic Parsing (WP 5.2) =====
-features: meaning representation with semantic roles (69 roles), coordinations, argument structure, partial ellipsis resolution, pronominal anaphora resolution, +features: meaning representation with semantic roles (69 roles), coordinations, argument structure, partial ellipsis resolution, pronominal anaphora resolution, post parsing detection of ungrammatical edges (caused by long utterances) 
-performance indicator: f-measure+performance indicator: accuracy
  
 ===== Information Extraction (WP 5.2) ===== ===== Information Extraction (WP 5.2) =====
Line 99: Line 76:
 features: za tohle bych vydaval klasifikator, co rozhoduje ,jestli se rekne 'To je smutné/veselé'. Tem adjektivum rucne priradim negative/positive sentiment. features: za tohle bych vydaval klasifikator, co rozhoduje ,jestli se rekne 'To je smutné/veselé'. Tem adjektivum rucne priradim negative/positive sentiment.
 performance indicator: f-measure performance indicator: f-measure
 +
 +
 +
  
  
Line 104: Line 84:
 ===== Complete System Evaluation ===== ===== Complete System Evaluation =====
 T5.2.7 tohle zminuje, nick webb to pro nas asi neudela T5.2.7 tohle zminuje, nick webb to pro nas asi neudela
-performance indicator: pocet slov ve vypovedich uzivatele, dotazniky+performance indicator: number of tokens in user reply utterances, post-session questionare
  
  
Line 111: Line 91:
 ===== Dialog Manager (WP 5.3) ===== ===== Dialog Manager (WP 5.3) =====
 features: reply types, using (language independed) predicates (prakticky to znamena, ze pojmenuju testy na prechodech v dafech anglicky) features: reply types, using (language independed) predicates (prakticky to znamena, ze pojmenuju testy na prechodech v dafech anglicky)
 +Handmade DAF covering following topics: Person_Retired, Person_in_productive_age, Child, Husband, Wife, Wedding, Christmas, Handle_stalled_dialog
 performance indicator: rucni hodnoceni prijatelnosti vybrane akce performance indicator: rucni hodnoceni prijatelnosti vybrane akce
  

[ Back to the navigation ] [ Back to the content ]