• 検索結果がありません。

[招待講演] 音声言語処理研究の進展と今後

N/A
N/A
Info
ダウンロード
Protected

Academic year: 2021

シェア "[招待講演] 音声言語処理研究の進展と今後"

Copied!
6
0
0

読み込み中.... (全文を見る)

今ダウンロードする ( 6 ページ )

全文

(1)2006-SLP-64 (5) 2006/12/21. IPSJ SIG Technical Reports. E-mail:. t [email protected],. ,. HMM,. Progress and Future Problems on Spoken Language Processing SeiichiNAKAGAWAr T Toyohashi University of Technorogy E-mail: Abstract. 1-1, Hibarigaoka, Tenpaku-cho, Toyohashi, Aichi, 441-8580 Japan T [email protected],. The technology on spoken language processing has progressed in decades and has been applied to practical. fields. However, the usage did not spread unlike we expected. This fact is not caused by only the incompleteness of user-interface and the lack of killer applications, but also the poor robustness on speech recognition technology. In this paper, we look back over our past studies and discuss future problems from a viewpoint of a researcher.. Keyword. Spoken Language Processing, Speech Recognition, DP Matching, HMM, Acoustic Model, Language Model. 1. I :ro. 30. io *u. LfeiE tv^i:. -25-.

(2) ^ a* ^ ^° ^ - v. 2. of®. 5fls* ARPA (Df^PmfeMfftyvi?*? b. (ATIS) J. ATR. —* (asj, jnas), c. i}>\z.te'zitzkl*s X *9 i^ A^ yci >9 >. DP. m (D it VXD dp. 1971 *£frb 5¥M**$tLfeTp< y ^O APRA (S. HMMIi, DP-^yf-i. t> <£> T* & 5 o. ^ttlcJ:^^. % DARPA. h :. -26-.

(3) 10. hmm. Deng. 9w<h^"C#6o. ^flh(OZ.bt^^o, HMM. 4. —J. , #Jx.{4\ 50%. &fr>tl&0 ^®L%k$U$tib 60—80% 50% c. it^<. f4 2000f§N^M¥{i(7)/No — -fU^- i/^r >i 14 40 b i~^>o. ^CfcMdJ:^ 2000 Mft (/^—7" l^^f'>x-f 2000) ■7s- 4 ifi 40 (-. 40 (D 8 ^. (=65536X10 O 8 ^). 8 ^(=6561 <D 10 <. 40. -27-. <D 0 h, 6553.6X10.

(4) A(Wc)>A(We)"C#> 3 Rli,. L(Wc). Lowerre. (9. A(Wc)<A(We)T* L(Wc)> L(We) «!r. 1976. (IE. A (ftW). >A (««), L. <L. Lowerre tOitV^T'fc 60 1976. tp 15a. Reddy (Lowerre <D»«««) *S(S^PI C*ife«: f-A L!£. t ^ 9 ig^jjs &> o fe t fflV^T ^So. tt. t ^ 5 - ^"Cttft < /' open problem. Nakagawa:. Speech. 1976 ¥. Understanding. It. £■?&£. System. -. (Sakai,. LITHAN-. and Some Applications, Proc. 3rd Int. Joint Conf. Pattern. Recognition) o O D P ^ v ^ 1/ if \z. J; 5 17 - K ^ 1£ v 7" >f >-. 85%,. ,. Bridle |:iot 1973. D P. 10. 17.9. Piiit 1978 ^^y- k. 87.5%. 17.5. :. HMM. Word. Spotting. , 1983.3),. , 1976. HMM. (fl. HMMhmm,. HMM ,. 1986 ¥3 flU. HMM. (i (. Nakagawa:. Recognition. 5.. 5.1.. y *f-. Dynamic. in. Time. 1986 ¥. 10. Speaker-independent Continuous Warping. Speech. Method,. by. Proc.. 8X*hZ> Consonant. a 8th. Stochastic Int.. Conf.. Pattern Recognition) 1/ h&{\LXt)D. 1988 ¥ 3. fc. -28-. («*.

(5) 2005 ¥ 8. Spoken Language Systems (Ohmsha, IOS Press)(C^g^> L. "PJH :. f-JII. TVN6. fc\ IBM<D Brown. (Nakagawa: Overview)o. 1987 {d. (i27 l^-. h. ?£ (-IBM te 9 7 u-. 100. 5.2.. ICASSP)O. 1991. ¥. 10. M. X* fo. & (Nakagawa,. Ono,. 12011,. Hirata:. Dimensionality Reduction of Dynamical Patterns Using a. LITHAN(t»±»^:, 1973-1976). Neural Network, Proc. IEEE Workshop Neural Networks for Signal Processing). (»«IDP>1975), 1(1976) P (1983),. O(n)DP (One pass DP, 1982),. 5/ ^ D P(1986),. 7 U—. Btt¥ • Perplexity •. -"14(1987),. (HMMK,. 1991), *. m ft, 1998.3 ,. SMR-Perplexity. 2*). ,. VQIIHMM. GMM. , 1998.7).. (d-distant n-gram, 1990),. 1989 ¥ 9. M X h Z> (Nakagawa:. Continuous. Speech. An. Evaluation. Recognition. Systems,. Method Proc.. for. ESCA. Workshop (Speech I/O & Database)o. DSP. ,. 1988),. A( 1995),. 1999.12,. ^ «t. -29-. X A(1980),.

(6) (1996), }Jlffi*^#|g'>^7iA(2002)% JSIjI • $4 I V. 6. A (2004). (1997), — -----■. 9 "". ^f # ^ *t*([email protected]>)o 3ft « DP(1980), Left-to-right M £ Island-driven M HMM(D. __. _. ... x. £. [1]. ^. „.,,,„, ^^.u,. """. f,. L SLP32-5,. 2000.. 1989.. /j^^f. [2]. i£^(1983), 7 l^ —A±ff (O^iS^jE^{t;(1996)s *2i. ^ ""%7 /U ^ ^ r* ^ *~ ■. (2000), GMUkHMU(D{^m(2003), >T 7°X K 7 ^ IE. t3l [4]. Web ^(D^. l&fllW^a, [5]. SLP6-8,. 1995.5.. tin I? — , f:^i"ififf*^f3i<7)^F^Eti^J ^^-= f&tt«*0fa, S L P 1-7, 1994. •. #^^I#^A,P 1998.PD CAI(1995), iiTny h^^(19987),. [7]. 4UUH-, #^^»W%^»J1% i1^^^^, P99-93,. (2002), B^^^i1^! (2005). S. 1999.. [8] 4»JI|lg —, g ^^iiic^Mi1^^!^, ^^Sfe«t PI#, li^if Slii^ff ^^j^-g-^:^, DP-4-2, 2006.. [9]. ^3(2002), ^QA(2003), mi&V<T *<» ^ ^ "f V V \\L. (2006). ^". nTni. tfJ'IM —, 'HcPI&Wifc bUtP'mMMM^ #^#lfe, 1996.. [10]<=PJHIi-, ^f >„ SLP-50-4, 2004.. "". ... .^.J,. ^(1986), ^ii^SlJ^C^> S^Kf^S^(1987), 0 #!§;£. 1994.. #(1991) , -f V K* ^>T S&iC X 5 0 ^§§7"^^ h^^. 0(n)DP fe(D .. (2005). S82-7,. 1982. — ,. 4 .. •If",. iti/|0\ 1988), ^^,—. B^e'^^^lfe, :,. -30-. 0. 2001. SLP-63-9-3,. 2006.. R^"m.

(7)

参照

今ダウンロードする ( PDF - 6 ページ - 557.89 KB )

関連したドキュメント

Foundation; the language of the idealistic filtration

(Construction of the strand of in- variants through enlargements (modifications ) of an idealistic filtration, and without using restriction to a hypersurface of maximal contact.) At

On the solvability of nonlocal pluriparabolic problems ∗

The aim of this paper is to prove existence, uniqueness, and continu- ous dependence upon the data of solutions to mixed problems for pluri- parabolic equations with nonlocal

ON Semiconductor Packaging and Labeling Guidelines

(4S) Package ID Vendor ID and packing list number (K) Transit ID Customer's purchase order number (P) Customer Prod ID Customer Part Number. (1P)

言語獲得 言語獲得 言語獲得

Guasti, Maria Teresa, and Luigi Rizzi (1996) &#34;Null aux and the acquisition of residual V2,&#34; In Proceedings of the 20th annual Boston University Conference on Language

(1)手話言語研究センター講話会

関西学院大学手話言語研究センターの研究員をしております松岡と申します。よろ

手話言語研究センター講話会

手話言語研究センター講話会.

有 岡 千 帆

Among other languages spoken in the country, there are Vedda, an indigenous language, Tamil, another official language, a few Creoles and English. However, in recent years, Vedda,