1 / 9

Presentation Outline

Presentation Outline. Introduction to Language Recognizers Background Work Project Objective Discussion on language models on other approaches on language Id based tools Future work & Conclusion. Introduction to Language Recognizers.

toyah
Download Presentation

Presentation Outline

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Presentation Outline • Introduction to Language Recognizers • Background Work • Project Objective • Discussion • on language models • on other approaches • on language Id based tools • Future work & Conclusion

  2. Introduction to Language Recognizers • Language Identification is process of identifying the language being spoken from a sample of speech by an unknown speaker • Applications based on language ID: • Call routing at customer centers • Automatic Translation Systems • International Airport • Rescue Activities • Data Verification

  3. Background Work(1/3) • Language Characteristics • Phonemes • Finite set of phonemes, with different languages having overlapping sets. • 56 phonemes. • Phonotactic Rules • Governs the way different phonemes are combined. • Language specific. • Language IDs try to exploit these characteristics.

  4. Background Work(2/3) • Language IDs are made of three sub-systems • Front-end processing. • Converts speech into an intermediate form like feature vectors. • Phoneme recognizers. • Generates phoneme sequences from feature vectors. • HMM models are used. • Language Models. • Does probabilistic analysis of phoneme sequences for identifying the language being spoken.

  5. Background Work(3/3) • Example: Phonetic Recognition/ Language Modeling (PRLM) • Example: Parallel-PRLM

  6. Project Objectives • An algorithm for more efficient language IDs, in field of language models. • Some language independent method for front-end processing and phoneme recognition. • A tools based on above algorithms for call routing and Content Verification System(CVS)

  7. Discussion-on language models (1/X) • Probabilistic models. • Input: Phoneme Sequences. • Output: Language being spoken. • Phases: • Training Phase • Tuning Phase • Testing Phase • N-Gram approach • P( ph1 ph2 ph3 | X) = P( ph1 | X) . P( ph2 | X) . P( ph3 | X)…….unigram • = P( ph2 | X, ph1) . P( ph3 | X, ph2)………....bigram • = P( ph3 | X, ph1, ph2)……………………...…trigram

  8. Discussion-on language models (2/X) • Language Models – Training phase • Unigram probability • Bigram probability • Trigram probability • Uni_Prob(ph| X) = No. of time phoneme ‘ph’ occur in the training data • -------------------------------------------------------------------- • Count of total no of phonemes in the training data for ‘X’ • Bi_Prob(ph2 | X, ph1) = No. of time phoneme ph1 is followed by phoneme ph2 in the data • ----------------------------------------------------------------------------------------- • No of times phonemes ph1 occurs in the training data of language ‘X’ • Tri_Prob(ph3 | X, ph1, ph2) = No. of time phoneme ph1 is followed by ‘ph2 ph3’ in the data • -------------------------------------------------------------------- No. of time phoneme ph1 is followed by ph2 in the training data of language ‘X’

More Related