• 沒有找到結果。

第五章 結論與未來展望

5.2 未來展望

本研究從第一階段詞綴構詞開始到第二階段加入韻律模型其 word lattice 愈長愈龐大,對測試 語料重新評分其所需時間也就愈長,未來必須盡可能限制 word lattice 的大小,縮短辨識所需時間。

從本研究可以延伸出三項議題值得未來進一步探討。第一,目前辨識時間過長,未來必須

47

大幅縮短辨識時間,盡量縮小辨認時的搜尋路徑;第二,從辨識結果可以發現,OOVs 大部分 屬於人名的部分,未來在第一階段中的語言模型需考慮人名這項因素,解決人名造成的錯誤率;

第三,目前本研究採用 TCC300 語料庫,屬於麥克風朗讀語音,未來若能將此辨識系統延展到 更貼近生活化的自發性語音,相信語音辨認可以更廣泛地應用於生活中。

48

參考文獻

【1】 Peter F. Brown, Vincent J. DellaPietra, Peter V. deSouza, Jennifer C. Lai, and Robert L.

Mercer. “Class-based N-gram models of natural language,” Computational Linguistics, vol. 18, no. 4, pp. 467–479, 1992.

【2】 Chien-Pang Chou, “Improvement on Language Modeling for Large-Vocabulary Mandarin Speech Recognition,” NCTU Speech Processing Lab, 2009

【3】 Yun-Shu Yang, “Large-Vocabulary Mandarin Speech Recognition using Hierarchical Language Model,” NCTU Speech Processing Lab, 2010

【4】 Matthew A. Siegler and Richard M. Stern “On The Effects of Speech Rate in Large Vocabulary Speech Recognition Systems”

【5】 F. Martinez, D. Tapias and J. Alvarez “Towards Speech Rate Independence in Large Vocabulary Continuous Speech Recognition”

【6】 T. Pfau, R.Faltlhauser, and G. Ruske “A Combination of Speaker Normalization and Speech Rate Normalization forAutomatic Speech Recognition”

【7】 C.-Y. Chiang, S.-H. Chen, H.-M. Yu, and Y.-R. Wang, “Unsupervised joint prosody labeling and modeling for Mandarin speech,” Journal of the Acoustic Society of America, vol.

125, no. 2, pp.1164-1183, Feb. 2009.

【8】 Z. Sheng, J.-H. Tao, and D.-L. Jiang,“Chinese prosodic phrasing with extended features,”Proceedings of the IEEE ICASSP 2003, Vol. 1, pp.492-495, 2008

【9】 C.-Y. Tseng, S.-H. Pin, Y.-L. Lee. H.-M. Wang, and Y.-C Chen, “Fluent speech prosody:Framwork and modeling,”Speech Commun. Special issue on quantitive prosody modeling for natural speech description and generation, 46, 284-309 2005

【10】 S.-H. Chen and Y.-R. Wang, “Vector quantization of pitch information in Mandarin speech,”

IEEE Transactions on Communications, vol. 38, no. 9, pp. 1317-1320, September 1990.

49

【11】 J. A. Bilmes and K. Kirchhoff, “Factor language models and generalized parallel backoff,”

in Proc. of HLT/NACCL, 2003, pp. 4-6.

【12】 A. Stolcke, “SRILM – An extensible language modeling toolkit,” in Proc. ICSLP, 2002.

【13】 P. Beyerlein, “Discriminative model combination,” in Proc. ICASSP 1998, pp. 481-484.

【14】 Ming-Chieh Liu, “An Implementation of Prosody-Assisted Mandarin Speech Recognition System,” NCTU Speech Processing Lab, 2011

【15】 C.-Y. Chiang, J.-H. Yang, M.-C. Liu, Y.-R. Wang, Y.-F. Liao, and S.-H. Chen, “A New Model-based Mandarin-speech Coding System,” Proc. of INTERSPEECH-2011, Florence, Italy, pp. 2561-2564, Aug., 2011.

50

附錄:決策樹之問題集

The question set used to construct the decision trees for building the break syntax model ( n| )n

P B l and P pd ed( n, n,pj dl dfn, n, n|B ln, )n is listed below:

' Is the inter-syllable location an utterance boundary?' ' Is the inter-syllable location an interword?'

' Does a PM exist at the inter-syllable location'

' Does a Major PM exist at the inter-syllable location ' ' Does a。exist at the inter-syllable location '

' Does a,exist at the inter-syllable location ' ' Does a、exist at the inter-syllable location ' ' Does a.exist at the inter-syllable location ' ' Does a;exist at the inter-syllable location ' ' Does a:exist at the inter-syllable location ' ' Does a?exist at the inter-syllable location ' ' Does a!exist at the inter-syllable location ' ' Does a(exist at the inter-syllable location ' ' Does a)exist at the inter-syllable location '

' Is the the preceding special prefix words + special 1-syllable words: Ng, Ncd, Di, DE, I, T' ' Is the POS of the preceding word A'

' Is the POS of the preceding word C' ' Is the POS of the preceding word D' ' Is the POS of the preceding word N' ' Is the POS of the preceding word I or T' ' Is the POS of the preceding word P'

51

' Is the POS of the preceding word V' ' Is the POS of the preceding word DE' ' Is the POS of the preceding word SHI' ' Is the POS of the preceding word FW' ' Is the POS of the preceding word DM'

' Is the POS of the preceding word Da Di Dk D' ' Is the POS of the preceding word Dfa'

' Is the POS of the preceding word Dfb'

' Is the POS of the preceding word Na Nb Nc Nv' ' Is the POS of the preceding word Nd'

' Is the POS of the preceding word Neu Nes Nep Neqa Neqb Nf' ' Is the POS of the preceding word Ng Ncd'

' Is the POS of the preceding word Nh'

' Is the POS of the preceding word VA VAC VG'

' Is the POS of the preceding word VB VC VCL VD VE VF VJ VK VL' ' Is the POS of the preceding word VH VHC VI'

' Is the POS of the preceding word V_2' ' Is the POS of the preceding word Caa' ' Is the POS of the preceding word Cab' ' Is the POS of the preceding word Cba' ' Is the POS of the preceding word Cbb' ' Is the POS of the preceding word Da' ' Is the POS of the preceding word Di' ' Is the POS of the preceding word Dk' ' Is the POS of the preceding word D'

52

' Is the POS of the preceding word Na' ' Is the POS of the preceding word Nb' ' Is the POS of the preceding word Nc' ' Is the POS of the preceding word Ncd' ' Is the POS of the preceding word Neu' ' Is the POS of the preceding word Nes' ' Is the POS of the preceding word Nep' ' Is the POS of the preceding word Neqa' ' Is the POS of the preceding word Neqb' ' Is the POS of the preceding word Nf' ' Is the POS of the preceding word Ng' ' Is the POS of the preceding word Nv' ' Is the POS of the preceding word I' ' Is the POS of the preceding word T' ' Is the POS of the preceding word VA' ' Is the POS of the preceding word VAC' ' Is the POS of the preceding word VB' ' Is the POS of the preceding word VC' ' Is the POS of the preceding word VCL' ' Is the POS of the preceding word VD' ' Is the POS of the preceding word VE' ' Is the POS of the preceding word VF' ' Is the POS of the preceding word VG' ' Is the POS of the preceding word VH' ' Is the POS of the preceding word VHC'

53

' Is the POS of the preceding word VI' ' Is the POS of the preceding word VJ' ' Is the POS of the preceding word VK' ' Is the POS of the preceding word VL' ' Is the length of the preceding word 1' ' Is the length of the preceding word 2' ' Is the length of the preceding word 3' ' Is the length of the preceding word 4' ' Is the length of the preceding word 5' ' Is the length of the preceding word 6'

' Is the length of the preceding word less than 2' ' Is the length of the preceding word less than 3' ' Is the length of the preceding word less than 4' ' Is the length of the preceding word less than 5' ' Is the length of the preceding word less than 6'

' Is the following special 1-syllable words: Ng, Ncd, Di, DE, I, T + special suffix words' ' Is the POS of the following word A'

' Is the POS of the following word C' ' Is the POS of the following word D' ' Is the POS of the following word N' ' Is the POS of the following word I or T' ' Is the POS of the following word P' ' Is the POS of the following word V' ' Is the POS of the following word DE' ' Is the POS of the following word SHI'

54

' Is the POS of the following word FW' ' Is the POS of the following word DM'

' Is the POS of the following word Da Di Dk D' ' Is the POS of the following word Dfa'

' Is the POS of the following word Dfb'

' Is the POS of the following word Na Nb Nc Nv' ' Is the POS of the following word Nd'

' Is the POS of the following word Neu Nes Nep Neqa Neqb Nf' ' Is the POS of the following word Ng Ncd'

' Is the POS of the following word Nh'

' Is the POS of the following word VA VAC VG'

' Is the POS of the following word VB VC VCL VD VE VF VJ VK VL' ' Is the POS of the following word VH VHC VI'

' Is the POS of the following word V_2' ' Is the POS of the following word Caa' ' Is the POS of the following word Cab' ' Is the POS of the following word Cba' ' Is the POS of the following word Cbb' ' Is the POS of the following word Da' ' Is the POS of the following word Di' ' Is the POS of the following word Dk' ' Is the POS of the following word D' ' Is the POS of the following word Na' ' Is the POS of the following word Nb' ' Is the POS of the following word Nc'

55

' Is the POS of the following word Ncd' ' Is the POS of the following word Neu' ' Is the POS of the following word Nes' ' Is the POS of the following word Nep' ' Is the POS of the following word Neqa' ' Is the POS of the following word Neqb' ' Is the POS of the following word Nf' ' Is the POS of the following word Ng' ' Is the POS of the following word Nv' ' Is the POS of the following word I' ' Is the POS of the following word T' ' Is the POS of the following word VA' ' Is the POS of the following word VAC' ' Is the POS of the following word VB' ' Is the POS of the following word VC' ' Is the POS of the following word VCL' ' Is the POS of the following word VD' ' Is the POS of the following word VE' ' Is the POS of the following word VF' ' Is the POS of the following word VG' ' Is the POS of the following word VH' ' Is the POS of the following word VHC' ' Is the POS of the following word VI' ' Is the POS of the following word VJ' ' Is the POS of the following word VK'

56

' Is the POS of the following word VL' ' Is the length of the following word 1' ' Is the length of the following word 2' ' Is the length of the following word 3' ' Is the length of the following word 4' ' Is the length of the following word 5' ' Is the length of the following word 6'

' Is the length of the following word less than 2' ' Is the length of the following word less than 3' ' Is the length of the following word less than 4' ' Is the length of the following word less than 5' ' Is the length of the following word less than 6'

Is the initial of the following syllable a null one or in { m, n, l, r}?

Is the initial of the following syllable a null one or in { b, d, g}?

Is the initial of the following syllable a null one or in { f, s, sh, h}?

Is the initial of the following syllable a null one or in { c, ch, q}?

Is the initial of the following syllable a null one or in { p, t, k}?

Is the initial of the following syllable a null one or in { z, zh, j}?

相關文件