DOC PREVIEW
CMU CS 15492 - Speech Recognition Grammars Other ASR techniques

This preview shows page 1-2-19-20 out of 20 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 20 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 20 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 20 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 20 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 20 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

Speech Processing 15-492/18-492Speech RecognitionGrammarsOther ASR techniquesBut not just acoustics• But not all phones are equi-probable• Find word sequences that maximizes• Using Bayes’ Law• Combine models– Us HMMs to provide– Use language model to provideBeyond n-gramsTriTri--gram languages modelsgram languages modelsGood for general ASRGood for general ASRMore targeted models for dialog systemsMore targeted models for dialog systemsLook for more structureLook for more structureFormal Language TheoryChomsky HierarchyChomsky HierarchyFinite State MachinesFinite State MachinesContext Free GrammarsContext Free GrammarsContext Sensitive GrammarsContext Sensitive GrammarsGeneralized Rewrite Rules/Turing machinesGeneralized Rewrite Rules/Turing machinesAs LM or as Understanding mechanismAs LM or as Understanding mechanismFolded into the ASR or only ran on outputFolded into the ASR or only ran on outputFinite State MachinesTrigram is a word^2 FSMTrigram is a word^2 FSMFSM for greetingFSM for greetingHelloGoodMorningAfternoonFinite State GrammarSentences Sentences --> Start Greeting End> Start Greeting EndGreeting Greeting --> “Hello”> “Hello”Greeting Greeting --> “Good” TOD> “Good” TODTOD TOD --> Morning> MorningTOD TOD --> Afternoon> AfternoonContext Free GrammarX X --> Y Z> Y ZY Y --> “Terminal”> “Terminal”Y Y --> > NonTerminalNonTerminalNonTerminalNonTerminalJSGFSimple grammar formalism for ASRSimple grammar formalism for ASRStandard for writing ASR grammarsStandard for writing ASR grammarsActually finite stateActually finite statehttp://www.w3.org/TR/jsgfhttp://www.w3.org/TR/jsgfFinite State MachinesFinite State Machines:Finite State Machines:DeterministicDeterministicEach arc leaving a state has unique labelEach arc leaving a state has unique labelThere always exists a Deterministic machine There always exists a Deterministic machine representing a nonrepresenting a non--Deterministic oneDeterministic oneMiniminalMiniminalThere exists an FSM with less (or equal) states that There exists an FSM with less (or equal) states that accepts the same languageaccepts the same languageProbabilistic FSMsEach arc has a label and a probabilityEach arc has a label and a probabilityCollect probabilities from dataCollect probabilities from dataCan do smoothing like Can do smoothing like ngramsngramsNatural Language ProcessingProbably mildly context sensitiveProbably mildly context sensitivei.e. you need context sensitive rulesi.e. you need context sensitive rulesBut if we only accept context freeBut if we only accept context freeProbably OKProbably OKIf we only accept finite stateIf we only accept finite stateProbably OK too Probably OK tooWriting Grammars for SpeechWhat do people say?What do people say?No what do people *really* say!No what do people *really* say!Write examplesWrite examplesPlease, I’d like a flight to BostonPlease, I’d like a flight to BostonI want to fly to BostonI want to fly to BostonWhat do you have going to BostonWhat do you have going to BostonWhat about BostonWhat about BostonBostonBostonWrite rules grouping things togetherWrite rules grouping things togetherIgnore the unimportant thingsI’m terribly sorry but I would greatly I’m terribly sorry but I would greatly appreciate if you might be able to help me appreciate if you might be able to help me find an acceptable find an acceptable flight to Bostonflight to Boston..I, I I, I wannawannawant to go to want to go to ehmehmBoston.Boston.What do people really sayA: see who else will somebody else important all the A: see who else will somebody else important all the {mumble} the whole school are out for a week{mumble} the whole school are out for a weekB: oh reallyB: oh reallyA: {A: {lipsmacklipsmack} {breath} yeah} {breath} yeahB: okay {breath} well when are you going to come up thenB: okay {breath} well when are you going to come up thenA: um let’s see well I guess I I could come up actually A: um let’s see well I guess I I could come up actually anytimeanytimeB: okay well how about nowB: okay well how about nowA: nowA: nowB: yeah B: yeah A: have to work tonight A: have to work tonight ––laughlaugh--Class based language modelsConflate all words in same classConflate all words in same classCities, Names, numbers etcCities, Names, numbers etcCan be automatic or designedCan be automatic or designedAdaptive Language ModelsUpdate with new News storiesUpdate with new News storiesUpdate your language model every dayUpdate your language model every dayUpdate your language model with daily useUpdate your language model with daily useUsing user generated data (if ASR is good)Using user generated data (if ASR is good)Combining modelsUse “background” modelUse “background” modelGeneral triGeneral tri--gram modelgram modelUse specific modelUse specific modelGrammar based Grammar based Very localizedVery localizedCombineCombineInterpolated (just a weight factor)Interpolated (just a weight factor)More elaborate combinationsMore elaborate combinationsMaximum entropy modelsMaximum entropy modelsVocabulary sizeCommand and controlCommand and control< 100 words, grammar based< 100 words, grammar basedSimple dialogSimple dialog< 1000 words, grammar/tri< 1000 words, grammar/tri--gramgramComplex dialogComplex dialog< 10K words, tri< 10K words, tri--gram (some grammar for control)gram (some grammar for control)DictationDictation< 64K words, tri< 64K words, tri--gramgramBroadcast NewsBroadcast News256K plus, tri256K plus, tri--gram (and lots of other possibilitiesgram (and lots of other possibilitiesHomework 1Build a speech recognition systemBuild a speech recognition systemAn acoustic modelAn acoustic modelA pronunciation lexiconA pronunciation lexiconA language modelA language modelNote it takes time to buildNote it takes time to buildWhat is your initial WERWhat is your initial WERHow did you improve itHow did you improve


View Full Document
Download Speech Recognition Grammars Other ASR techniques
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Speech Recognition Grammars Other ASR techniques and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Speech Recognition Grammars Other ASR techniques 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?