Differences
This shows you the differences between two versions of the page.
Both sides previous revision Previous revision Next revision | Previous revision Next revision Both sides next revision | ||
draft [2009/07/15 13:36] 147.228.47.142 |
draft [2009/09/30 20:48] ptacek |
||
---|---|---|---|
Line 1: | Line 1: | ||
- | ====== | + | ====== |
+ | The ASR module based on Hidden-Markov models transforms input speech into text, providing a front-end between the user and the Czech demonstrator. The ASR output is smoothed into a form close to standard written text using statistical machine translation in order to bridge the gap between dis-fluent spontaneous speech and a standard grammatical sentence. | ||
- | Hi Marc, | + | Results of the part-of-speech tagging are passed on to the Maximum Spanning Tree Syntactic Parsing module. A tectogrammatical representation of the utterance is constructed once the syntactic parse is available. Annotation of the meaning at tectogrammatical layer is more explicit than its syntactic parse and lends itself for information extraction. The Named Entity Recognition module then marks personal names and geographical locations. Afterwards, the dialog act classifier uses number of lexical and morphological features to assess the type of user utterance (such as question, acknowledgement, |
- | ... | + | The dialog is driven by a Dialog Manager component by USFD (originally developed for the English Senior Companion prototype). |
+ | CU has supplied the transition networks covering following topics: retired_person, | ||
+ | Dialogue Manager provides information about the appropriate communicative function along with the sentence that is to be generated to the NLG module. The TTS module integrated with the TID avatar transforms system responses from the text form into the speech and visual (face expressions, | ||
- | Re: progress: there is progress in the following: | ||
- | - language model re-training for the collected dialogue data | + | < |
- | - 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. | + | The Czech Companion follows |
+ | taking advantage of the data collected in the first phase of the project (using a Wizard-of-Oz setting). The full recorded corpora was transcribed, | ||
- | Best, | + | The architecture is the same as in the English version, i.e. a set of modules communicating through the Inamode (TID) backbone. However, the set of modules is different, see Figure 1. Regarding the physical settings, the Czech version runs on two notebook computers connected by a local network. One serves as a Speech Client, running modules dealing with ASR, TTS and ECA; the other one as an NLP Server. |
- | -- Jan | + | The NLU pipeline, DM, and NLG modules at the NLP Server are implemented using a CU's own TectoMT platform that provides access to a single in-memory data representation through a common API. This eliminates the overhead of a repeated serialization and XML parsing that an Inamode based solution would impose otherwise. |
+ | The Knowledge Base consists of objects (persons, events, photos) that model the information acquainted in the course of dialog. Those objects also provide a very basic reasoning (e.g. accounting for the link between date of birth and age properties). Each object' | ||
- | |||
- | |||
- | |||
- | |||
- | ====== 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; | ||
- | 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' | ||
- | typy odpovedi a zpusob jejich implementace, | ||
- | NLP server s tectomt, ASR/TTS/SR client, connected over network | ||
- | XXX JPta | ||
- | |||
- | advances in Czech NLU (on reconstructed spoken data): 300-500vet(? | ||
- | pos ? analyzovat, generovat a kontrolovat ' | ||
- | |||
- | |||
- | |||
- | |||
- | |||
- | ===== Automatic Speech Recognition ===== | ||
- | features: improved language models, real-time speaker adaptation | ||
- | performance indicator: WER | ||
- | |||
- | |||
- | ===== Speech Reconstruction ===== | ||
- | features: omit filler phrases, irrelevant speech events, false starts, repetitions, | ||
- | imlementation(zahrnout tuhle info?): moses natrenovany na korpusu | ||
- | 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 ===== | ||
- | features: XXX Mirek/ | ||
- | performance indicator: accuracy | ||
- | |||
- | ===== Syntactic Parsing ===== | ||
- | features: induce dependencies and labels | ||
- | performance indicator: f-measure | ||
- | v tipu je natrenovat MacDonnalda na dialog datech, ten task je do M42, ted ne. | ||
- | |||
- | |||
- | |||
- | ===== Semantic Parsing ===== | ||
- | features: meaning representation with semantic roles (69 roles), coordinations, | ||
- | performance indicator: f-measure | ||
- | |||
- | ===== Information Extraction ===== | ||
- | features: template based identification of predicates | ||
- | covering predicates from before-mentioned set of DAFs. | ||
- | performance indicator: accuracy | ||
- | |||
- | |||
- | ===== Named Entities Recognition ===== | ||
- | features: detect person names, geographical locations (organizations myslim nepotrebne) | ||
- | performance indicator: f-measure | ||
- | |||
- | ===== Dialog Act Tagging ===== | ||
- | features: tagset derived from DAMSL-SWBD, DA is a key feature driving the decision, what to say next. | ||
- | performance indicator: accuracy | ||
- | |||
- | |||
- | ===== Sentiment Analysis ===== | ||
- | features: za tohle bych vydaval klasifikator, | ||
- | performance indicator: f-measure | ||
- | |||
- | |||
- | |||
- | ===== Complete System Evaluation ===== | ||
- | T5.2.7 tohle zminuje, nick webb to pro nas asi neudela | ||
- | performance indicator: pocet slov ve vypovedich uzivatele, dotazniky | ||
- | |||
- | |||
- | |||
- | |||
- | ===== Dialog Manager ===== | ||
- | features: reply types, using (language independed) predicates (prakticky to znamena, ze pojmenuju testy na prechodech v dafech anglicky) | ||
- | performance indicator: rucni hodnoceni prijatelnosti vybrane akce | ||
- | |||
- | ===== Natural Language Generation ===== | ||
- | features: variations, underspecified input (dott format), emotional markup (natvrdo v dafech a templatech u hodnoticich vet) | ||
- | performance indicator: BLEU score |