InField 2010 Institute on Field Linguistics and Language Documentation University of Oregon

Transcription

1 InField 2010 Institute on Field Linguistics and Language Documentation University of Oregon Workshop Coursepack: ELAN 1: Aligning Text to Audio and Video Using ELAN Instructors: Andrea Berez & Christopher Cox June 22, 23, 24, 28, :30-3:15 PM Knight Library, Room 267B Table of Contents: Syllabus 1 Lecture Slides 3 Quick Start Guide 29 Exercise 1: Creating a single language transcript with one speaker 30 Exercise 2: Creating a single language transcript with multiple speakers 33 Exercise 3: Creating a two language basic transcript 36 Exercise 4: Creating a two language transcript with full IGT 38 Exercise 5: Working with video 43 Exercise 6: Some tricks for automating transcription 44

2 InField 2010 ELAN: Syllabus Syllabus About ELAN: ELAN is a professional tool for the creation of complex annotations of audio and video resources. Creating transcripts in ELAN has several advantages for the language documentation workflow. First, it creates an archival XML document that links your annotations (text) to the timeline of the media in a way that is long-lasting and not reliant on proprietary software for recovery (meaning that your transcription will be available well into the future). Second, ELAN is flexible enough to be used when you have a recording of one speaker, or five speakers, or several languages at once, etc. Third, ELAN allows import from and export to a range of other popular linguistic software and format (like Transcriber, Toolbox, CHILDES, etc.). Fourth, ELAN files can be used in the creation of pedagogical and presentation products for language maintenance and revtalization. ELAN is highly specialized software, and it can take a while to learn how to set up your files. This workshop will help you climb over the learning curve, and youʼll see that ELAN isnʼt difficult once you know how to use it! **The Level 1 workshop assumes students have very little or no experience with ELAN. The Level 2 workshop assumes students have at least a little experience with ELAN (but the Level 1 workshop is not required).** Course overview: In this class we will learn how to use the ELAN software to link transcriptions to audio and video media. We will talk extensively about how to conceptualize tiers (layers of text, linguistic analysis, and translations) so that you can create annotations that are flexible enough for a range of configurations of speakers and languages. Day-by-day schedule for Level 1 workshop: Day 1 (Tuesday June 22) Introduction: a tour of ELAN Starting up, file naming, file storage Creating a single-language annotation file with one speaker Creating a single-language annotation file with two speakers Day 2 (Wednesday June 23) Review of yesterday Creating a two-language simple annotation (sentence-level) Day 3 (Thursday June 24) Creating a two-language complex annotation (full interlinearized glossed text) Day 4 (Monday June 28) Working with videos Some advanced features of ELAN: Working with templates, gaps, silence Andrea Berez & Chris Cox 1

3 InField 2010 ELAN: Syllabus Beyond ELAN: Importing from and exporting to other useful file types Brainstorming for your own projects File storage during the class: There is no central shared server space where you can store your files, so you will need to keep everything on your thumb drive. Please be sure to save all your files and keep them, because you may want to see them later. Contact information: Andrea: Chris: Andrea Berez & Chris Cox 2

4 !"#$%&'%#()*+)+*%,-./%/0 #12)0%3+2%4)2-0%15)+*%!"#$ 9:(+&;;<9:(+&;=>&;?!? G2* *-G68D:,FC+*5,-G,! K6*L4267&M6,F4 N(&5234&O6*L4267&O+&,3E&56&7*6H3)+&,&2,()4<6( 3(5*6):GP6(&56&"#$%&J6*&C+Q3((+*4&CIR B26O3(Q&I6:&O2,5&"#$%&34&,()&O2I&6(+&O6:F)&O,(5&56&:4+ 35 S*6H3)3(Q&,(&6H+*H3+O&6J&52+&G6(G+754&I6:&O3FF&(++)&56&C+ 4:GG+44J:F&O352&"#$% M3H3(Q&I6:&52+&6776*5:(35I&56&7*,GPG+&O352&,&*,(Q+&6J&),5, 45*:G5:*+&5I7+4 K+&,3E&56&7*+7,*+&I6:&J6* T6:*&6O(&7*6U+G54 V2+&+8G3P(Q&"#$%&;&O6*L4267&(+85&O++LW ;

5 BIFF,C:4R&Y,I&! V:+4),I&9:(+&;; N(5*6):GP6(R&K2,5&34&"#$%&,()&O2I&)6&O+&O,(5 56&:4+&35Z %:54&,()&C6F54R&)6O(F6,)[3(45,FF,P6(>&\F+&456*,Q+ 16(G+75:,F3/3(Q&5*,(4G*3754 "8+*G34+4&!0;R&1*+,P(Q&,&43(QF+&F,(Q:,Q+&PE+<,F3Q(+)&5*,(4G*375&3(&"#$% ](+&47+,L+* ^:FP7F+&47+,L+*4 X BIFF,C:4R&Y,I&; K+)(+4),I&9:(+&;X `+H3+O&6J&),I&! $C6:5&7,*+(5&0&G23F)&P+*4>&F3(Q:34PG&a45+*+6b5I7+4>,()&P+*&C+2,H36*4 "8+*G34+&XR&1*+,P(Q&,&43E7F+&5O6<F,(Q:,Q+> 4+(5+(G+<F+H+F&PE+<,F3Q(+)&5*,(4G*375 _

6 BIFF,C:4R&Y,I&X V2:*4),I&9:(+&;_ `+H3+O&6J&),I4&!&0&; ^6*+&,C6:5&7,*+(5&0&G23F)&P+*4>&F3(Q:34PG a45+*+6b5i7+4>&,()&p+*&c+2,h36*4 "8+*G34+&_R&1*+,P(Q&,&5O6<F,(Q:,Q+&PE+<,F3Q(+) 5*,(4G*375&O352&J:FF&3(5+*F3(+,*3/+)&QF644+)&5+85 anmvb c BIFF,C:4R&Y,I&_ ^6(),I&9:(+&;= `+H3+O&6J&),I4&!<X K6*L3(Q&O352&H3)+6 B6E+&652+*&J+,5:*+4&6J&"#$%R V+E7F,5+4 M,74 B3F+(G+.*,3(456*E3(Q&J6*&I6:*&7*6U+G54R&26O&O3FF&I6: 45*:G5:*+&I6:*&"#$%&\F+4Z d

7 #66L3(Q&,2+,)R V2+&#+H+F&;&O6*L4267&(+85&O++L %6O&52,5&O+&L(6O&"#$%>&O2,5&G,(&O+&)6 O352&35ZR $:56E,PG&4+QE+(5,P6(&6J&*+G6*)3(Q4 ^:FPF3(Q:,F&4:CP5F+)&YfY4 1,7P6(+)&T6:V:C+&H3)+64 1F3GL,CF+&456*3+4&a3(5+*,GPH+>&O+C<C,4+) 7*+4+(5,P6(4&:43(Q&1:S"Yb e %:54&0&.6F54 =

8 Y6O(F6,)3(Q&0&N(45,FF3(Q H H 2h7ROOO-F,5<E73-+:[566F4[+F,( a"#$%&34&,f*+,)i&3(45,ff+)&6(&52+4+&f,c&g6e7:5+*4b g Y6O(F6,)3(Q&0&N(45,FF3(Q K3()6O4R Y6:CF+<GF3GL&6(&52+ &&&"#$%ix<g<?io3(i3(45,ff-+8+ &&&,()&J6FF6O&52+&3(45,FF+*&3(45*:GP6(4- ^,GR Y6:CF+<GF3GL&6(&52+ &&&"#$%ix<g<?ie,gi3(45,ff-/37&a56&+85*,g5b>&52+( &&&Y6:CF+&GF3GL&6(&52+ &&&"#$%ix<g<?ie,gi3(45,ff &&&,()&J6FF6O&52+&3(45,FF+*&3(45*:GP6(4-!?

9 j3f+&b56*,q+&y:*3(q&1f,44 $FF&52+&G6:*4+&\F+4>&3(GF:)3(QR,&SYj&G6:*4+7,GL&O352&52+&F+G5:*+&4F3)+4&,()&+8+*G34+ )+4G*37P6(4 $:)36>&H3)+6>&,()&5+85&\F+4&J6*&+8+*G34+4,*+&,H,3F,CF+&6(&52+&GF,44&O+C435+R 2h7R[[F6Q64-:6*+Q6(-+):[3(\+F);?!?[O6*L42674[,F3Q(3(Q<5+85< +F,(![3() SF+,4+&4,H+&+H+*I523(Q&56&I6:*&52:EC&)*3H+>&3(GF:)3(Q 52+&O6*L&I6:&)6&):*3(Q&GF,44<<I6:&O3FF&(++)&52+E F,5+*W!! N(5*6):GP6(!;

10 K2,5&34&,&PE+<,F3Q(+) 5*,(4G*375Z V3E+<,F3Q(+)&5*,(4G*3754&E,5G2&C354&6J&5+85 a!""#$!%#"&b&56&4+gp6(4&6j&,(&,:)36&6*&h3)+6 *+G6*)3(Q&a'()*!bR!X (thanks Chris!) K2,5&34&,&PE+<,F3Q(+) 5*,(4G*375Z V3E+<,F3Q(+)&5*,(4G*3754&E,5G2&C354&6J&5+85 a!""#$!%#"&b&56&4+gp6(4&6j&,(&,:)36&6*&h3)+6 *+G6*)3(Q&a'()*!bR!_ (thanks Chris!)

11 K2,5&34&,&PE+<,F3Q(+) 5*,(4G*375Z "8,E7F+4&6J&PE+<,F3Q(+)&5*,(4G*3754&,*+&(65 2,*)&56&\()R (thanks Chris!) ]7+*,&4:*P5F+4 ^6H3+&4:CP5F+4!c 1F64+)&G,7P6(3(Q K2I&C652+*Z T6:*&F,(Q:,Q+&*+46:*G+4&O3FF&C+&J,*&*3G2+*-&T6: O3FF&C+&,CF+&56 B+,*G2 f+*3ji&gf,3e4 `+3(J6*G+&I6:*&F,(Q:,Q+&,4&,&F3H3(Q&F,(Q:,Q+ ^,5G2&6*526Q*,72I&56&7*6(:(G3,P6(&J6*&F35+*,GI V*,(4J6*E&I6:*&*+G6*)4&3(56&)3H+*4+&)347F,I J6*E,54 ---!d

12 K2,5&34&"#$%Z "#$%&34&,&566F&J6*&7*6):G3(Q&PE+<,F3Q(+)&5*,(4G*3754 6J&,:)36H3):,F&*+G6*)3(Q4- Y+H+F67+)&,()&E,3(5,3(+)&CI&52+&^,8&SF,(GL N(4P5:5+&3(&%3UE+Q+(- k"ayn1]&#3(q:34pg &&&&&$((65,56*l&<&O2+*+ &&&&k"ayn1]l&34&52+ &&&&&":*67+,(&Y345*3C:5+) &&&&&16*7:4l&7*6U+G5- (thanks Chris!)!e K2I&:4+&"#$%Z "#$%&34&)345*3C:5+)&J*++&6J&G2,*Q+&,()&,4&67+(< 46:*G+&46mO,*+- "#$%&34&G*644<7F,n6*E-&N5&*:(4&:()+*&K3()6O4>&^,G>,()&#3(:8- N5&34&)+H+F67+)&CI&F3(Q:3454&J6*&F3(Q:34PG&:4+- Y+H+F67+*4&O6*L&,GPH+FI&O352&:4+*4&56&G6(P(:,FFI 3E7*6H+&52+&46mO,*+&56&E++5&52+&(++)4&6J&F,(Q:,Q+ O6*L+*4- V2+*+&34&,&Q*6O3(Q&4:776*5&G6EE:(35I>&,GG+443CF+ 52*6:Q2&52+&6(F3(+&J6*:E-!=

13 K2I&:4+&"#$%Z "#$%&7*6):G+4&(6(<7*67*3+5,*I>&45,(),*)4< G6E7F3,(5&6:57:5R A(3G6)+>&46&G2,*,G5+*&)347F,I&34&(65&,&7*6CF+E- o^#>&46&i6:*&)6g:e+(5&34&f6(q<f,4p(q-&n5&g6(j6*e4&56&o+ff )6G:E+(5+)&o^#&4G2+E,&,()&34&2:E,(&+I+C,FF<*+,),CF+- ]:57:5&G,(&C+&5*,(4J6*E+)&3(56&,&*,(Q+&6J&:4+J:F 7*6):G54&J6*&,&O3)+&,:)3+(G+- B6E+&,*+&C:3F5<3(&+876*54>&46E+&G,(&C+&G*+,5+)&O352&523*)< 7,*5I&7*6):G54&a,()&46E+&,*+&I+5&56&C+&)+H+F67+)Wb-!g K2,5&3J&N&U:45&O,(5&,&7*3(5+) 5*,(4G*375ZW V2+*+&,*+&E,(I&*+,46(4&56&:4+&,&7*3(5+)&5*,(4G*375- N(&52+&7,45>&52+&7*3(5+)&5*,(4G*375&O,4&G6(43)+*+)&56 C+&7*3E,*IR K2,5+H+*&52+&7*3(5&G6:F)&(65&G,75:*+&O,4&6m+(&(65 3(GF:)+)&3(&52+&)34G:4436(- T6:&O+*+&F:GLI&56&Q+5&,(&,GG6E7,(I3(Q&*+G6*)3(Q>&C:5&5234 O,4&U:45&G6(43)+*+)&Q*,HI- N(&"#$%&52+&5+85&,()&52+&*+G6*)3(Q&,*+&:(3\+)> 67+(3(Q&52+&)66*&56&E,(I&76443C3F3P+4&J6*&*+4+,*G2,()&F+,*(3(Q- V2+&7*3(5+)&5*,(4G*375&34&U:45&6(+&6J&E,(I&76443CF+ +876*54- ;?

14 K2,5&"#$%&34&p(65p "#$%&34&(65&,&E+)3,&+)356*R&35&G,(l5&G2,(Q+&I6:*,:)36&6*&H3)+6&\F+- T6:&(++)&56&)6&5234&3(&46E+&652+*&46mO,*+- NJ&I6:&+)35&I6:*&E+)3,&\F+>&I6:lFF&,F46&(++)&56&G2,(Q+&I6:* "#$%&5*,(4G*375&56&E,5G2&52+&(+O&PE+F3(+- "#$%&34&(65&,&5+85&+)356*R&354&,((65,P6(4&,()&6:57:54,*+&3(&7F,3(&5+85- NJ&I6:&O,(5&56&,))&+EC+FF342E+(54&F3L+&:()+*F3(3(Q> 60(2)+*>&6*&*$!+*,*-!%#">&I6:lFF&(++)&56&)6&35&3(&,&5+85&+)356*,m+*&I6:&+876*5&I6:*&7*3(5+)&5*,(4G*375- ;! $&56:*&6J&"#$% #+5l4&5,L+&,&F66L&,5&52+&"#$%&O3()6O>&O23G2 G6(5,3(4R ^+(:&C,* f3)+6&,()&o,h+j6*e&h3+o+*4 ^+)3,&7F,I+*&G6(5*6F4R&G6,*4+>&\(+>&,()&4+F+GP6( $((65,P6(&)+(435I&H3+O+* ^:FP<P+*&H3+O+* $*+,&J6*&Q*3)>&5+85>&4:CP5F+4>&,:)36[H3)+6 *+G6Q(3/+*4>&E+5,),5,>&,()&G6(5*6F&H3+O+*4- ;;

15 B6E+&3E76*5,(5&G6(G+754 V6&:4+&"#$%&+q+GPH+FI>&I6:&E:45&:()+*45,()R T6:*&6O(&Q6,F4&J6*&52+&45*:G5:*+&6J&I6:*&5*,(4G*375 "#$%l4&c:3f5<3(&g6(g+754&,c6:5&)3q+*+(5&l3()4&6j,((65,p6(4 r6o&i6:*&q6,f4&e,7&6(56&"#$%l4&g6(g+754 #+5l4&45,*5&O352&52+&\*45&6(+R K2,5&L3()&6J&3(J6*E,P6(&)6&N&O,(5 3(&EI&PE+<,F3Q(+)&5*,(4G*375Z #+5l4&F66L&,5&46E+&J,E3F3,*&45*:G5:*+4&J6* 7*3(5+)&5*,(4G*3754- ;X B6E+&3E76*5,(5&G6(G+754R V2+&45*:G5:*+&6J&I6:*&5*,(4G*375 "8,E7F+&!R&B3(QF+&F,(Q:,Q+>&6(+&47+,L+*- B3E7F+&5*,(4G*37P6(&4I45+E- ;_

16 aa.#/!+*$0>&7*6h3)+)&ci&`,)36#,c 2h7R[[OOO-O(IG-6*Q[426O4[*,)36F,Cbb 9642R&sr6O&)6&7+67F+&E,L+&5234&U:)QE+(5Z&j6*Q+5 O2+52+*&6*&(65&52+4+&U:)QE+(54&,*+&*3Q25&6*&O*6(Q> U:45>&O2,5l4&Q63(Q&6(&3(&52+&C*,3(&52,5&E,L+4&7+67F+ )34P(Q:342&46&(,5:*,FFI&,()&3(5:3PH+FI&C+5O++(& O6&G,4+4Z&K23G2>&J*6E&,(&,G5:,*3,F&763(5&6J&H3+O>,*+&H+*I&H+*I&H+*I&43E3F,*&3J&(65&3)+(PG,F---t ;c B6E+&3E76*5,(5&G6(G+754R V2+&45*:G5:*+&6J&I6:*&5*,(4G*375 "8,E7F+&;R&B3(QF+&F,(Q:,Q+>&E:FP7F+&47+,L+*4- ^,IC+&I6:&,*+&:43(Q&,&E6*+&3(J6*E,PH+ 5*,(4G*37P6(&4I45+ER S*646)3G&72*,43(Q B7+,L+*&6H+*F,7 S,:4+&F+(Q52 #,:Q25+*&,()&652+*&(6(<F,(Q:,Q+&6*,F&(634+4 S2*,4+&,GG+(5[45*+44 av2+4+&,*+&52+i&4iec6f4&3(&52+&a1b.&5*,(4g*37p6( 4I45+E&3(&52+&(+85&4F3)+b ;d

17 !;c $#fn%u&&i+,2> &aa1(+(2*&*#">&"(qf342bb!;d &N&v2,H+(w5&<<!;e &--&F3L+&v,)E3h+)FI>!;= &N&v2,H+(w5>!;g &a?-;b!x? &ab%njjb!x! &a?-xb!x; &N5w4&vJ:((I>!XX &N&x2,H+(w5&xO,5G2+)&5264+&3(&vI+,*4>!X_ &I6:&yL(6O>!Xc S"V"`u&&&&&&&&&&&&&&yT+,2-z!Xd $#fn%u&&&&&&&&&&&&&&&&&&&&&&&&4z6&vnwh+&526:q25>!xe &a?-eb!x= &NwH+&v526:Q25&&{>!Xg &35&E3Q25&C+&vJ:(&56&v4++&52+E&,Q,3(-!_? &v.:5>!_! &a?-_b!_; &yae>z!_x S"V"`u&&&yx]*z&v(6R5&y;&46&J:z(-! $#fn%u&&&&&&&&&&&&&&&&&&&&&&&&y;a1]amrbz!_c &&&--&T+,2>&yDDDE,I<>!_d S"V"`u&&&&&&&&&&&&&&&&&&&&&&yDDDDDDDDDz!_e $#fn%u&&&^,ic+&35w4&vc+45&u:45&56&*+e+ezc+*&52+&v7f+,4,(5&y;xg23f)266)&xe+e6zyx*3+4-z!_= $##NB]%u&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&y;DD&V2+&O,Iz&&&&&&&&&&yX35&vO,4>z!_g #"$u&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&yxd&t+,2-z ;e B6E+&3E76*5,(5&G6(G+754R V2+&45*:G5:*+&6J&I6:*&5*,(4G*375 "8,E7F+&XR&VO6&F,(Q:,Q+4R $&4:CU+G5&F,(Q:,Q+ $&5*,(4F,P6(&F,(Q:,Q+ B+(5+(G+&F+H+F&5*,(4F,P6(4 ;=

18 aa3+!,*(/45!$(/>&y+(,l3(,-&b7+,l+*r&$()*+o&.,ff:5,bb! %%%% A7&,5&52+&2+,)&6J&#,L+&1F,*L>&:7&3(&52,5&H,FF+I>&52+*+&,*+&7,44+4&3(&52+&$F,4L, `,(Q+&3(&C652&)3*+GP6(4- ; B*4<7*"*4*)<7!+-(C4,7;=4D;($"=4<<=0!4:;!")!-)+("47!4$;*C4>*4$!;!4"+!"47!- %%% K2+(&52+&QF,G3+*4&45,*5&E+FP(Q>&,FF&52+&O,5+*& 6O4&3(56&52+&*3H+*- X 37=74:;!")!-)+("4,7;=4E7=:=$("<7(7$"=4)!7D!)*+!C47!40(7)*42("4()*+!CA %%% $()&52+(&35&J6*E4&3(56&,&F,L+F,*L- _& F*4$!;!4<7*"*G40*4()*+!C4,7;=:;=46*-78(749("!4D=;=4()*+!CA %% V2,5&QF,G3+*&O,5+*>&35&J6*E4&}3/2U+2&f+(,&y#,L+&1F,*Lz- c& B*4<7*"*4()*+!C4,7;=4H=")!+%"49("!4D*:;=4()*+!CA &&& $()&52+(&35& 6O4&3(56&%:(),FP(&f+(,&y%6(),F56(&#,L+z-l ;g B6E+&3E76*5,(5&G6(G+754R V2+&45*:G5:*+&6J&I6:*&5*,(4G*375 "8,E7F+&_R&VO6&F,(Q:,Q+4&O352&J:FF&NMVR $&4:CU+G5&F,(Q:,Q+ $&5*,(4F,P6(&F,(Q:,Q+ S*646)3G&72*,4+4 ^6*72+E+4 ^6*72+E+&QF644+4 B+(5+(G+&F+H+F&5*,(4F,P6(4 X?

19 KAN Sentence: ((Pear Story, Kannada. Speaker: Keshava Subramanya)) Ii kathe obba, ha!!u maaliya bagge, mattu, obba hu"ugana bagge ide. Intonation Unit: Ii kathe obba, Morphemes: ii kathe obba Gloss: this story one Intonation Unit: ha!!u maaliya bagge, Morphemes: ha!!u maali -ya bagge Gloss: fruit gardener- SRC about Intonation Unit: Morphemes: Gloss: mattu, mattu and Intonation Unit: obba hu"ugana bagge ide. Morphemes: obba hu"uga -na bagge ide Gloss: someone.m boy -GEN about be.3sm Free Translation:!This story is about a fruit farmer and a boy. X! B6E+&3E76*5,(5&G6(G+754R V2+&45*:G5:*+&6J&I6:*&5*,(4G*375 "8,E7F+&cR&%6(<6*,F&E6),F35IR V*,(4F,P6(&3(&Y:5G2 V*,(4F,P6(&3(&"(QF342 `3Q25&2,()&QF644 #+m&2,()&qf644 `3Q25&2,()&)3*+GP6(&,()&F6G,P6( #+m&2,()&)3*+gp6(&,()&f6g,p6(.*6o4>&+i+&q,/+>&2+,)>&--- X;

20 V*,(4F,P6(&Y:5G2&&&"+(&26()&F3+7&67&/3U(&Q+E,L-&V6+(&23U&C3U&++(&4F,Q+*4/,,L&LO,E>&/,Q&23U&++( C65&F3QQ+(- V*,(4F,P6(&"(QF342&$&)6Q&O,4&O,FL3(Q&,*6:()-&K2+(&2+&G,E+&,G*644&,&C:5G2+*>&2+&4,O&,&C6(+- MF644&`r&"(QF342&&&& Y]M MF644&#r&"(QF342&&&&&&Y]M MF644&`r&&&&&&&&&&&&&& r]%y MF644&#r&&&&&&&&&&&&&& r]%y "I+&,7+*5:*+&&&&&&&&& &C ^6:52&&&&&&&&&&&&&&&& &w26()w "I+&M,/+&&&&&&&&&&&& &&C2 r+,)&&&&&&&&&&&&&&&&& &( "I+&M,/+&&&&&&&&&&&&&&&&&&&&G "I+&,7+*5:*+&&&&&&& &C "I+&M,/+&&&&&&&&&&&& &&* `6F+&&&&&&&&&&&&&&&&& &)6Q ^6:52&&&&&&&&&&&&&&&&& GF64+)>&J6*O,*) ^6:52&B"&&&&&&&&&&&&& &[SA`B"Y[ MF644&`r&"(QF342&&&&&&a7<b&O,FL3(Q&)6Q MF644&`r&&&&&&&&&&&&&& a7<b&f67+(&26() aa"1r]&;??_>&7*6h3)+)&ci&52+&a(3h+*435i&6j MF644&#r&"(QF342&&&&&&a7<b&O,FL3(Q&)6Q&a<2b %3UE+Q+(>&2h7R[[F+5-L:(-(F[43Q(<F,(Q[+G26bbXX MF644&#r&&&&&&&&&&&&& &a7<b&f67+(&26()&a<2b B6E+&NE76*5,(5&16(G+754R V3+*4&0&VI7+4&a3(5*6b N(&43E7F+&5+*E4>&,&P+*&34&+H+*I523(Q&J*6E&, 5*,(4G*375&52,5&34&52+&4,E+&L3()&6J 3(J6*E,P6(R "H+*I523(Q&$FH3(&4,I4&3(&"8-&;&a6*&#+,>&S+5+*> $FF346(b "H+*I523(Q&F,C+F+)&kY+(,l3(,&5+85l&3(&"8-&X "H+*I523(Q&F,C+F+)&kN(56(,P6(&A(35l&a6* k^6*72+e+4l&6*&kmf644lb&3(&"8-&_ "5G--- X_

21 B6E+&NE76*5,(5&16(G+754R V3+*4&0&VI7+4&a3(5*6b NE,Q3(+&I6:&2,H+&,&H3)+6&6J&,&5O6<G2,*,G5+* 7:77+5&426O&)+43Q(+)&56&5+,G2&%6*52&B,,E3 4+(5+(G+4&56&47+,L+*4&6J&j3((342-&K2,5&,*+ I6:*&P+*4Z r+3)+ r+q6( Xc B6E+&NE76*5,(5&16(G+754R V3+*4&0&VI7+4&a3(5*6b r+3)+ r+q6( $FF&6J&r+3)+l4&:h+*,(G+4&a3(&%6*52&B,,E3b V*,(4F,P6(4&6J&,FF&6J&r+3)+l4&:h+*,(G+4&3(56&j3((342 $FF&6J&r+Q6(l4&:h+*,(G+4&a3(&%6*52&B,,E3b Xd V*,(4F,P6(4&6J&,FF&6J&r+3)+l4&:h+*,(G+4&3(56&j3((342-

22 B6E+&NE76*5,(5&16(G+754R V3+*4&0&VI7+4&a3(5*6b K352&7*3(5+)&5*,(4G*3754>&52+&\8+)&O3)52&6J 52+&7,7+*&J6*G+4&:4&56&7:5&52+&)3q+*+(5&7,*54 6J&,&43(QF+&P+*&6(&)3q+*+(5&F3(+4- Tier 1... Tier 2... Tier 3... Tier 1... Tier 2... Tier 3... Tier 1... Tier 2... Tier 3... Tier 1... Tier 2... Tier 3... Tier 1... Tier 2... Tier 3... Tier 1... Tier 2... Tier 3... Xe B6E+&NE76*5,(5&16(G+754R V3+*4&0&VI7+4&a3(5*6b K352&,&PE+<C,4+)&566F&F3L+&"#$%>&O+&G,( G6(G+3H+&6J&+,G2&P+*&,4&+834P(Q&,4&, G6(P(:6:4&45*+,E&6(&,&43(QF+&F3(+>&G6(PQ:6:4 56&52+&PE+F3(+&6J&52+&E+)3,R V3+*&! V3+*&; V3+*&X X=

23 B6E+&NE76*5,(5&16(G+754R V3+*4&0&VI7+4&a3(5*6b K352&E:FP7F+&G6(P(:6:4&P+*4>&O+&G,( G,75:*+&6H+*F,7&C+5O++(&47+,L+*4R V3E+ Xg B6E+&NE76*5,(5&16(G+754R V3+*4&0&VI7+4&a3(5*6b.I&G6(G+3H3(Q&6J&P+*4&,4&7*6Q*+443(Q&J6*O,*)&52*6:Q2 PE+>&O+&G,(&C+Q3(&56&+45,CF342&46E+&*+F,P6(42374 C+5O++(&)3q+*+(5&P+*4- V2+&7,*+(5<G23F)&*+F,P6(4237&6J&P+*4R S,*+(54R&P+*4&52,5&G6(5*6F&52+&C+2,H36*&6J&652+*&P+*4&a52+3* G23F)*+(b&3(&46E+&O,I 123F)*+(R&P+*4&52,5&,*+&,446G3,5+)&O352&,()&)+7+()+(5 :76(&652+*&P+*4&a52+3*&7,*+(54b- 1+*5,3(&G2,(Q+4&56&7,*+(5&P+*4&O3FF&,q+G5&G23F)&P+*4- _?

24 B6E+&NE76*5,(5&16(G+754R V3+*4&0&VI7+4&a3(5*6b ^645&P+*4&G,(&C+&C652&,&G23F)&56&6(+&P+*&,(),&7,*+(5&56&,(652+*>&C:5&52+&567<F+H+F&7,*+(5 P+*a4b&34[,*+&52+&6(+a4b&52,5&G6(5,3(a4b&,&)3*+G5 *+7*+4+(5,P6(&6J&52+&G6(5+(5&6J&52+&E+)3,- V2+4+&,*+&F3(L+)&)3*+G5FI&56&52+&*+G6*)3(Q- "H+*I&"#$%&5*,(4G*375&2,4&,5&F+,45&5234&6(+&P+*- V3+*&!R&$&5+85&*+7*+4+(5,P6(&6J&52+&:h+*,(G+4&6(&52+&*+G6*)3(Q- V3+*&;&a67P6(,Fb V3+*&X&a67P6(,Fb--- _! B6E+&NE76*5,(5&16(G+754R V3+*4&0&VI7+4&a3(5*6b V2+&7*+G34+&*+F,P6(4237&C+5O++(&,&7,*+(5&,()&, G23F)&34&)+\(+)&3(&"#$%&CI&sF3(Q:34PG&5I7+4t- $&F3(Q:34PG&5I7+&34&)+5+*E3(+)&CI&52+&L3()&6J&),5,&, P+*&G6(5,3(4R Ah+*+)&O6*)4 V*,(4F,P6(4 #3(Q:34PG&,(,FI434>&*+4+,*G2+*&G6EE+(54>&+5G- ](FI&G+*5,3(&F3(Q:34PG&5I7+4&G,(&C+&G23F)*+(&56&652+* F3(Q:34PG&5I7+4 K+&O3FF&47+()&,&F65&6J&PE+&F+,*(3(Q&,C6:5&"#$%l4 F3(Q:34PG&5I7+4&3(&5234&O6*L4267>&46--- _;

25 #+5l4&Q+5&45,*5+)W _X "8+*G34+&!R ](+&F,(Q:,Q+>&6(+&47+,L+* N(&5234&+8+*G34+&O+&O3FF&F+,*(R r6o&56&g*+,5+&,&(+o&"#$%&\f+&j*6e&,&*+g6*)3(q r6o&56&+(5+*&pe+<,f3q(+)&5+85&6(&52+&)+j,:f5&p+* r6o&56&+)35&pe+<,f3q(+)&5+85 r6o&56&4,h+&,(&"#$%&\f+- V:*(&56&"8+*G34+&!&3(&I6:*&G6:*4+7,GL- j3f+4&j6*&5234&+8+*g34+&,*+&j6:()&3(&52+&j6f)+* G,FF+)&+8+*G34+!iF,(Q:,Q+i673(36(-

26 "8+*G34+&;R ](+&F,(Q:,Q+>&E:FP7F+&47+,L+*4 N(&5234&+8+*G34+&O+&O3FF&F+,*(R r6o&56&g*+,5+&,))3p6(,f&p+*4 r6o&56&o6*l&o352&6h+*f,773(q&47++g2 $&F3hF+&C35&,C6:5&F3(Q:34PG&5I7+4 V:*(&56&"8+*G34+&;&3(&I6:*&G6:*4+7,GL- j3f+4&j6*&5234&+8+*g34+&,*+&j6:()&3(&52+&j6f)+* G,FF+)&+8+*G34+;i5+F+H3436(- _c "8+*G34+&XR VO6&F,(Q:,Q+4>&C,43G&5*,(4G*375 N(&5234&+8+*G34+&O+&O3FF&F+,*(R r6o&56&g6(g+3h+&6j&,()&g*+,5+&,))3p6(,f&f3(q:34pg 5I7+4&,()&P+*4 r6o&56&o6*l&o352&5o6&f,(q:,q+4r $&4:CU+G5&F,(Q:,Q+ $&5*,(4F,P6(&F,(Q:,Q+ V:*(&56&"8+*G34+&X&3(&I6:*&G6:*4+7,GL- j3f+4&j6*&5234&+8+*g34+&,*+&j6:()&3(&52+&j6f)+* G,FF+)&+8+*G34+Xi56*PFF,iE,L3(Q- _d

27 "8+*G34+&_R VO6&F,(Q:,Q+4>&J:FF&NMV N(&5234&+8+*G34+&O+&O3FF&F+,*(R $C6:5&,))3P6(,F&F3(Q:34PG&5I7+4&52,5&,*+&(++)+) J6*&NMV r6o&56&o6*l&o352&(:e+*6:4&p+*4 V:*(&56&"8+*G34+&_&3(&I6:*&G6:*4+7,GL- j3f+4&j6*&5234&+8+*g34+&,*+&j6:()&3(&52+&j6f)+* G,FF+)&+8+*G34+_i7+,*i456*I- _e "8+*G34+&cR K6*L3(Q&O352&H3)+6 N(&5234&+8+*G34+&O+&O3FF&F+,*(&,C6:5&O6*L3(Q O352&H3)+6&E+)3,>&3(GF:)3(Q&45,*P(Q&,&(+O E:5F3<H3)+6&5*,(4G*375- V:*(&56&"8+*G34+&c&3(&I6:*&G6:*4+7,GL- j3f+4&j6*&5234&+8+*g34+&,*+&j6:()&3(&52+&j6f)+* G,FF+)&+8+*G34+ciE6*(3(Q- _=

28 "8+*G34+&dR $&J+O&5*3GL4&J6*&,:56E,P6( N(&5234&+8+*G34+&O+&O3FF&F+,*(&5O6&O,I4&56,:56E,5+&I6:*&5*,(4G*37P6(&7*6G+44&,&C35R V2+&$:)36&`+G6Q(3/+* K6*L3(Q&O352&Q,74 V:*(&56&"8+*G34+&d&3(&I6:*&G6:*4+7,GL- j3f+4&j6*&5234&+8+*g34+&,*+&j6:()&3(&52+&j6f)+* G,FF+)&+8+*G34+di*6,)5*37- _g T6:*&7*6U+G54R #+5l4&C*,3(456*EW V23(L&,C6:5&7*6U+G54&I6:&2,H+&)6(+&6*&O,(5 56&)6- K2,5&L3()&6J&),5,&)6&I6:&2,H+Z $:)36&[&H3)+6&*+G6*)3(Q4Z V*,(4G*3754Z K2,5&,*+&I6:*&Q6,F4Z r6o&e:g2&)+5,3f&g,(&i6:&*+,46(,cfi&3(gf:)+z 1,(&I6:&,))&E6*+&)+5,3F&F,5+*Z r6o&o3ff&i6:&45*:g5:*+&i6:*&"#$%&\f+4z c?

29 InField 2010 ELAN: Quick Start Guide ELAN Quick Start Guide Downloading and installing ELAN ELAN is free software developed by the Max Planck Institute for Psycholinguistics in Nijmegen, The Netherlands. The ELAN website is found here: From this page, you can download ELAN and the manual, as well as visit their support forum. Once you have downloaded the latest version of ELAN, open the installer and follow the instructions. Starting and working with a new file: 1. Start ELAN. 2. To start a new file, go to File > New. 3. Navigate to your media file (.mpg,.mpeg,.wav, etc.), then click the double arrow (>>) button to move your file into the right-hand window. Click OK. (note: there are occasions when you would want to put more than one file into the right-hand window, for example if you had a video stream, an audio stream, and a predefined template). 4. Save your file right away: File > Save. The ELAN file extension is.eaf. See On file naming below. 5. Set your Automatic backup to 1 minute. File > Automatic Backup > 1 minute. 6. Set up Linguistic Types and Tiers. 7. To select areas for annotation, use the mouse to highlight a certain section of the waveform. The selected areas will be blue. 8. To add an annotation, double-click inside the blue area. Type your annotation. 9. To make the annotation stick, use ctrl-enter (or cmnd-enter). 10. To delete an annotation, click the annotation so that the horizontal line is blue. Then use Edit>Delete Annotation. Be sure to save your file. If you later move your audio file into another location, ELAN will lose it and ask you where it is. Youʼll have to renavigate to it. On file naming: You can name your ELAN file anything you want youʼll be able to identify it later by the file extension.eaf. However, it makes sense to give it a name that you can easily associate with the media file too. For example, if my audio file of the Kannada language is called KAN KS-PearStory.wav, then I will name my ELAN file KAN KS-PearStory.eaf. Also, I will keep them both in the same folder on my computer (which Iʼll also name KAN KS-PearStory). Pick a naming convention that works for you and stick with it. If you are working with an ELAN template for a batch of similarly structured recordings, give your template a name that helps you know which project it belongs to for example, KannadaNarratives.etf. Andrea Berez & Chris Cox 29

30 InField 2010 ELAN: Exercise 1 Exercise 1 Creating a single-language transcript with 1 speaker Goals: In this exercise we will make our first ELAN file using a recording of an English monologue. A one-speaker, one-language project like this one is easiest kind of ELAN transcript to create, since we can use almost all of ELANʼs default settings. We will also learn how to work with the features of the ELAN workspace. Todayʼs files are found in the folder called ʻexercise1_language_opinionʼ. 1 Step one: Create a new ELAN file. Open the ELAN software. Go to File > New. Navigate to the audio file weʼre using today: language_opinion.wav Highlight the audio file in the left hand box by clicking on it, then click the double arrow (>>) to move it over to the right hand box, under Selected Files. Click OK. The ELAN window will initialize for a moment and then display the waveform. Now save your file immediately. Go to File > Save. Give your new ELAN file a name that makes sense (e.g., the same name as the audio file minus the.wav; here ʻlanguage_opinionʼ), and click Save. Now set your automatic backup to save your file every minute so that you donʼt lose your work in case of a crash. Go to File > Automatic Backup > 1 minute. Step two: Add some descriptive metadata to the default tier. In the Multi-tier Viewer area, you will see a tier already present, labeled ʻdefault.ʼ ELAN provides this tier already configured to contain a text representation of the contents of the media stream. However, ʻdefaultʼ isnʼt a very helpful name, and we should add some more information (descriptive metadata) that is specific to this tier. Right click on the tier label ʻdefaultʼ and select ʻChange Attributes of default.ʼ This opens up the ʻChange Tier Attributesʼ dialog box. Youʼll see several things in this box: an upper area that contains a list of all the tiers that are present in this file four tabs: Add, Change, Delete, Import (Youʼre currently looking at the Change tab) Some drop down menus and input boxes. In the lower half of the ʻChange Tier Attributesʼ dialog box, do the following: Change the tier name to something a little more helpful, like ʻNatasha Warner Text. 1 The audio is from a lecture entitled ʻLanguage Opinionʼ from a course entitled ʻLanguage: Individuals and Societies by Dr. Natasha Warner at the University of Arizona and is available on itunes U. Andrea Berez & Chris Cox 30

31 InField 2010 ELAN: Exercise 1 Insert the name of the speaker in the ʻParticipantʼ box (in this case, Natasha Warner). Insert your own name in the ʻAnnotatorʼ box. Leave everything else (ʻParent Tierʼ, ʻLinguistic Typeʼ, ʻDefault Languageʼ) alone for now. Weʼll learn more about these in future exercises. Click ʻChangeʼ, which will close the dialog box and change your tier name. Now when you hover over the new tier label ʻtextʼ, you should see a pop-up box with your new information. Step three: add annotations. Now comes the fun part, adding annotations. Use the audio controllers in ELAN to listen to the transcript. You can highlight a portion of the waveform by clicking and dragging with your mouse. When you find an area that you want to annotate, highlight it it will turn blue. Then, double click in the blue area across from the ʻtextʼ tier label. A white text box will open up there. Place your cursor in the box and type the words that you hear here. Now you have to make the annotation stick. Youʼll be tempted to just press Enter try it, and youʼll see that it doesnʼt make the annotation stick, it just moves the cursor down a line. To make your annotation stick, click control-enter (Mac cmd-return). Now go through the entire transcript and add annotations as you hear them. Step four: editing annotations. If you make a mistake, you donʼt have to delete the entire annotation (although you can). You can move the edges around in the timeline, shift an entire annotation earlier or later, or merge it with the next annotation. You should spend some time practicing these. 1. To move one edge of an annotation either forward or backward in time: Click on the annotation so that the horizontal line turns blue (it is now selected): While holding down the ALT key, place your cursor on the blue vertical bar at the starting or ending edge of the annotation. Click and hold them mouse, then drag the bar into the desired position. 2. To move an entire annotation either forward or backward in time: Like above, click on the annotation so that the horizontal line turns blue (it is now selected). While holding down the ALT key, place your cursor on the blue horizontal bar. Click and hold them mouse, then drag the entire annotation into the desired position. Andrea Berez & Chris Cox 31

32 InField 2010 ELAN: Exercise 1 3. To merge an annotation with the following one: Like above, click on the annotation so that the horizontal line turns blue (it is now selected). Right-click on the selected annotation, then choose ʻMerge with Next Annotation.ʼ It will automatically merge both the timeline and the text of the annotation. 4. To delete an annotation entirely: Like above, click on the annotation so that the horizontal line turns blue (it is now selected). Right-click on the selected annotation, then choose ʻDelete Annotation.ʼ Step five: Saving your file. Donʼt forget to save your file when youʼre done. ELAN will create a few different file types that you will see in your directory:.eaf: this is your actual ELAN transcript file. You can double-click on this file the next time you want to work on it and it will open automatically..psfx: this file contains your personal preferences about how the ELAN window should look when you open it. This file is not necessary, so if you lose it, donʼt worry. ELAN will create a new on automatically..eaf.001: this is your temporary backup file. Again, you donʼt need to have this file, so donʼt worry if itʼs missing. Andrea Berez & Chris Cox 32

33 InField 2010 ELAN: Exercise 2 Exercise 2 Creating a single-language transcript with multiple speakers Goals: In this exercise we will make an ELAN transcript of an audio clip of a English conversation between four people: Alvin, Lea, Peter, and Allison. We will learn how to create new tiers, and begin to think about linguistic types. Todayʼs files are found in the folder called ʻexercise2_television.ʼ Step one: Create a new ELAN file. Open the ELAN software. Go to File > New. Navigate to the audio file weʼre using today: television.wav Highlight the audio file in the left hand box by clicking on it, then click the double arrow (>>) to move it over to the right hand box, under ʻSelected Files.ʼ Click OK. The ELAN window will initialize for a moment and then display the waveform. Now save your file immediately. Go to File > Save. Give your new ELAN file a name that makes sense (e.g., the same name as the audio file minus the.wav; here ʻtelevision), and click Save. Your automatic backup should already be set from the previous exercise, but if not, set your automatic backup to save your file every minute so that you donʼt lose your work in case of a crash. Go to File > Automatic Backup > 1 minute. Step two: Define linguistic types. Because weʼre starting to work with multiple tiers, itʼs a good time to start thinking about ELANʼs ʻlinguistic types.ʼ This transcript will have four tiers, one for each of the four speakers, and they will all contain a textual representation of the actual words that are spoken on the recording. Remember that text representations of the contents of the recordings are considered to be top-level parent tiers, So, weʼll need a linguistic type that is associated directly with the timeline of the audio stream. Any tiers we make using this type will be used for dividing the audio stream into time-based portions. ELAN has a default type like this already assigned when you create a new file. Itʼs called ʻdefault-ltʼ, but letʼs give it a better name. Go to Type > Change Linguistic Type This opens up the ʻChange Typeʼ dialog box. Youʼll see several things in this box: an upper area that contains a list of all the types you have defined for this file four tabs: Add, Change, Delete, Import (Youʼre currently looking at the Change tab) Some drop down menus and input boxes. The first drop down menu contains a list of all the types that are defined for the file right now, it only has ʻdefault-ltʼ available. Andrea Berez & Chris Cox 33

34 InField 2010 ELAN: Exercise 2 Below that is a text box that contains the name of your current type, the one ELAN has named ʻdefault-lt.ʼ Give it a better name now: something like ʻtextʼ (or anything else you can easily remember). When picking a name for your type, remember that youʼre not giving a name to an actual tier here. For example, itʼs a bad idea to call this type ʻJoeʼs utterancesʼ because you can use this type for any tier that is associated directly to the audio timeline i.e., not just Joeʼs utterances, but also Maryʼs utterances, or even Maryʼs individual words. Save those more specific names for your tiers. The other items in the dialog box contain some information about the behavior of this linguistic type. Youʼll see that the stereotype is ʻNone.ʼ This is the correct stereotype for this linguistic type (weʼll talk about the other stereotypes later). The menu called ʻUse Controlled Vocabularyʼ can be ignored for now, as can the text box called ʻISO Data Categoryʼ and the ʻReferences to Graphics Allowedʼ check box. Notice that the ʻTime-alignableʼ check box is checked. This is correct for this type, since tiers made with this type are aligned directly to the audio timeline. Click ʻChangeʼ to accept the name change to this tier. Notice that the name changes in the window at the top of the dialog box. This is the only type we need for our single-language transcript, so click ʻCloseʼ to close the dialog box. Step three: Define tiers. Today we are going to make a transcript for the television.wav audio clip. If you like, you can follow along with the pre-typed transcript by opening television.txt in a text editor (wherein the stands for a pulse of laughter). Youʼll see that we have four speakers in this transcript, so we need to create four tiers: one each for Alvin, Lea, Peter and Allison. Each of these will be a top-level parent tier, and directly associated to the audio time line. Again, ELAN gives you a default tier of the correct configuration (but only one), so weʼll give it a better name and then create more tiers just like the first one for each speaker. Go to Tier > Change Tier Attributes A dialog box similar to the type dialog box will open up. You will see: an upper area that contains a list of all the tiers you have defined for this file four tabs: Add, Change, Delete, Import (Youʼre currently looking at the Change tab) Some drop down menus and input boxes. The first drop-down menu contains all the tiers you have set up for this file (it only contains one, called ʻdefaultʼ). Below that is an input box with the current tierʼs name. Letʼs give it a better name. Since we need a tier for each speaker, letʼs start by calling this tier ʻAlvinʼ, named after our first speaker. The next input box is labeled ʻParticipant.ʼ Here you can type the full name of the speaker if you wish (Alvin Smith, for example). In the box labeled ʻAnnotatorʼ you can type your own name if you wish (or the name of whoever annotated this tier for you). Andrea Berez & Chris Cox 34

35 InField 2010 ELAN: Exercise 2 The drop-down menu labeled ʻParent Tierʼ already contains ʻNone.ʼ This is correct remember, this is going to be a parent tier, so it has ʻnoneʼ for its parent. The menu labeled ʻLinguistic Typeʼ already has the correct type selected, the ʻtextʼ type that we previously defined. The ʻDefault Languageʼ menu allows you to pick a language for this tier. Click ʻChangeʼ to accept the renamed tier. Notice that the name has changed both in the upper portion of the dialog box and in the main ELAN window. Now we have to set up three more tiers just like this one for the other speakers in our audio file. Leave the dialog box open and click on the ʻAddʼ tab (or, if youʼve closed the box, go to Tier > Add New Tier). In the ʻTier Nameʼ box, type the name of the second speaker: Peter. Again you can type the speakerʼs full name in the ʻParticipantʼ box, and your own name in the ʻAnnotatorʼ box. Leave the ʻParent Tierʼ and ʻLinguistic Typeʼ settings as they are (ʻnoneʼ and ʻtextʼ respectively). When youʼve defined everything correctly, click ʻAddʼ to add this tier (if you make a mistake, click on the ʻChangeʼ tab and fix it!). Now add tiers just like this for the other two speakers, Allison and Lea. Click ʻCloseʼ to close the dialog box. Step four: add annotations. Now you can begin to add annotations for each speaker. Just like in the previous exercise, use the media controllers to listen to selected portions, and then you are ready to transcribe, double-click on a blue-highlighted area to open up a text box. Note that the multi-tier structure of ELAN allows you to overlap annotations in time whenever the recording contains overlapping speech. You may need to listen carefully for the overlaps! As you work, try to keep the edges of the annotations as close as possible to the actual beginnings and ends of the utterances you are transcribing. If you would like to cut-and-paste from the text version of the transcript, feel free. Keep in mind this is a good time-saving technique if you are working with files that have already been transcribed in a separate text editor, which you are now timealigning in ELAN for the first time. Step five: save your file. Andrea Berez & Chris Cox 35

36 InField 2010 ELAN: Exercise 3 Exercise 3 Creating a two-language basic transcript (sentence level) Goals: In this exercise, we will learn to make a very basic two-language ELAN file. By two languages, we mean a language spoken in the audio or video recording plus a translation language (as opposed to two different languages being spoken in the recording). By basic, we mean that utterances are given at the sentence level (as opposed to being broken down into smaller units or with any kind of linguistic or analytical information). Todayʼs files are found in the folder called ʻexercise3_tortilla_making.ʼ Step one: Create a new ELAN file. Follow the directions from the previous exercises for creating a new ELAN file, but use todayʼs audio file: tortilla_making.wav. Step two: Define linguistic types. For this file we now need two different linguistic types. The first type is just like the ones we used with the television.wav recording: a type for the top-level parent tier that is associated directly with the timeline. Follow the directions from the last exercise to rename ELANʼs default type to something a little nicer, like ʻtextʼ. This will be the type we use for tiers that contain the actual words the speaker in the recording is saying. Now we need another type that we havenʼt used before. We need a type that will be used for tiers that contain translations of the content of their parent tiers. That is, we need a type whose behavior says ʻIʼm not the actual words on the recording, but Iʼm a symbolic representation (i.e., not a literal representation) of the actual words (e.g., in another language) that are on the recording.ʼ Go to Type > Add New Linguistic Type This brings up the Add Type dialog box, which works like the other boxes weʼve seen. In the first input box in the lower half, insert the name of our new type, something like ʻtranslation.ʼ Notice that calling this type ʻtranslationʼ means we can use it for translation of units of any size, like the sentences in this lesson, or morpheme-level glosses in the next lesson. In the ʻStereotypeʼ drop-down menu, select ʻSymbolic Association.ʼ This tells ELAN essentially what was stated above, that tiers using this type are symbolic representations of the contents of their parent tiers (they are second-language translations of the actual words on the recording). Notice that the ʻTime-alignableʼ check-box becomes unchecked. This is correct, because this linguistic type is not associated directly to the time line of the audio stream. For now you can leave the other settings as they are (ʻUse Controlled Vocabularyʼ = ʻnoneʼ, ʻISO Data Categoryʼ = blank, ʻReferences to graphics Allowedʼ = unchecked). Click ʻAddʼ to add this type to our list of defined types, and then click ʻCloseʼ to close the dialog box. Andrea Berez & Chris Cox 36

37 InField 2010 ELAN: Exercise 3 Step three: Define tiers. For this exercise, weʼll need just two tiers: one parent tier, linked directly to the timeline to contain the original language transcription, and one child tier, linked to its parent, to contain the translation. Remember that ELANʼs default tier is a parent tier. Follow the directions from the last exercise to rename the default tier. You can choose to call it ʻOtomi textʼ and set the ʻParticipantʼ name to ʻAlejandra,ʼ or you can simply name the tier ʻAlejandra.ʼ Under the ʻDefault Languageʼ menu, for now you can leave ʻEnglishʼ selected, even though the contents of this tier are Otomi and not English. This menu is used for languages with non-roman orthographies, or for IPA symbols. Weʼll talk more about this menu later. Now you need to create a tier for the translation. Either click the ʻAddʼ tab or go to Tier > Add New Tier. In the ʻTier Nameʼ box, put in a name for your translation tier. I can be something like ʻEnglish Translationʼ or ʻEnglish Free Translationʼ. Now go to the ʻParent Tierʼ dropdown menu. Youʼll see the name of your parent tier there go ahead and select it. This tells ELAN that your ʻEnglish Translationʼ tier is a child of the tier called ʻOtomiʼ. Notice that if you filled in a Participant name for your parent tier, itʼs automatically filled in for you now. Also notice that ELAN has automatically selected ʻtranslationʼ as the Linguistic Type for this tier. That is correct. Click ʻAddʼ to add your tier, and then click ʻCloseʼ to close the dialog box. Step four: add annotations. Adding annotations is very similar to the previous exercises. Listen to the audio, and use the mouse to select/highlight a portion of the waveform, double-click in the blue highlighted portion in the area of your Otomi tier, and add your first line of annotation. Now try adding the English translation of the first line. Double click on the English Translation tier just below the first Otomi annotation. Even if that portion of the wave form is no longer selected (blue), a text box will open up that is directly under the parent annotation. It has the same starting and ending times as the parent. This is because the linguistic type we used for this tier tells ELAN to link it to its parent tier, rather than to the time line of the audio stream. Try as you might, you wonʼt be able to change the start and end times to be different from the parent annotation. Now go through the entire audio file and transcription, and add the Otomi and English annotations. Donʼt forget to save the file when youʼre done. Question to think about: What would you do if you had two (or more) Otomi speakers? How would you change or add new types and tiers? Andrea Berez & Chris Cox 37

38 InField 2010 ELAN: Exercise 4 Exercise 4 Creating a two-language transcript with interlinearized glossed text (IGT) Goals: In this exercise, we will learn to make an ELAN file for IGT. This is the most complicated structure weʼve seen so far, but itʼs only a small conceptual leap away from the twolanguage basic annotation we created yesterday. We will use a recording of Kannada language with the following structure for our IGT (structures may vary for your own project): Tier 1: Kannada sentences (normative sentences) Tier 2: Kannada intonation units (prosodic phrases) Tier 3: Kannada words by morphemes (words with hyphens to show morpheme boundaries) Tier 4: English morpheme glosses Tier 5: English free translation of the entire Kannada sentence. Todayʼs files are found in the folder called ʻexercise3_pear_story.ʼ Step one: Create a new ELAN file. Follow the directions from the last exercise for creating a new ELAN file, but use todayʼs audio file: pear_story.wav. Before beginning to define linguistic types and tiers, letʼs take a moment to think about what types weʼll need for the tiers listed above. Tier 1: This will be a top-level parent tier, linked directly to the audio stream. It will contain a transcription of the actual utterances found in the recording, and they will be the size of a sentence. Tier 2: This tier will also contain transcriptions of the actual utterances in the recording, but they will be in smaller units than the first tier (i.e., subdivisions of the first tier). This tier is a child of Tier 1, because we want to show that the intonation units are subparts of the sentences. However, it will also be a parent tier to the tiers below it (because we will want to show that the items in the next tier, words and morphemes, are subparts of the intonation unit). This tier is also linked to the timeline is the sense that you can click on a single intonation unit and hear that part of the audio stream. Tier 3: This tier is not linked to the timeline (you couldnʼt click and hear just a single word or morpheme) 1 ; rather, itʼs a child of Tier 2. Its contents are subdivisions of Tier 2. Tier 4: This tier is a child of Tier 3, and contains a one-to-one translation of the items in its parent (a morpheme-by-morpheme gloss). Tier 5: This tier is a translation of, and thus a child of, Tier 1. Step two: Define linguistic types. For this file we need four linguistic types. You are already familiar with the first two: a type called ʻtextʼ for the top-level parent tier, the one containing the Kannada sentences; 1 Itʼs not that you could never do this with words or morphemes, itʼs just that here weʼre going to set it up so that you canʼt just hear a single morpheme by clicking on the annotation. You could certainly set it up that way if you wanted to! Andrea Berez & Chris Cox 38

39 InField 2010 ELAN: Exercise 4 and a type called ʻtranslationʼ for the two tiers that contain translations (both morpheme glosses and the sentence-level free translation). Notice that we can use the same type for two different tiers. Follow the instructions in the previous exercises to create these two types. You will also need two more types, and they are conceptually similar they are both for tiers that are subdivisions of their parents. However, ELAN handles them a little differently because one of these types is linked to the audio stream (so you can clickand-play), but the other is not. Go to Type > Add New Linguistic Type The first of these two types we will make will be called ʻtimed chunkʼ (go ahead and put that in the ʻType nameʼ box). In the ʻStereotypeʼ menu, select ʻTime Subdivision.ʼ This means that the type will be used for tiers that are a subdivision of some other tier, but are still linked to the timeline. Notice that the ʻTime-alignableʼ check box is checked. This is correct. Click ʻAddʼ to add this type. The second of these two types we will make is for tiers that are subdivisions of some other tier but are not linked to the timeline. Call this type ʻchunkʼ. 2 In the ʻStereotypeʼ menu, select ʻSymbolic subdivision.ʼ This is because we are not making timeline subdivisions, we are just creating a space to separate words and morpheme boundaries. Notice that the ʻTime-alignableʼ checkbox is now unchecked. This is correct. Click ʻAddʼ and then ʻCloseʼ. Step three: Define tiers. For this file we need five tiers. As before, we can use ELANʼs default tier as our top-level parent tier, but we should give it a better name, like ʻKannada Sentence.ʼ Please do so. Next, letʼs work our way down the list of tiers and set each up in turn (you could do this in any order, and the way you specify the parent/child relationships will turn out correct, but itʼs easier to just go from top to bottom). We need a tier for the intonation units. Add a new tier called ʻIntonation Unitsʼ (or ʻprosodic phrasesʼ or something equally descriptive), fill in the Participant and Annotator, and select this tierʼs parent, ʻKannada Sentence.ʼ Note that ELAN suggests a Linguistic Type in the drop down menu it doesnʼt always get it right so be sure to double check. For this tier, we want to set the ʻLinguistic Typeʼ to ʻtimed chunk.ʼ Click ʻAddʼ to create the tier. Now we need a tier for the words and morphemes. Create a new tier called ʻWords and Morphemesʼ3, and set its parent tier as ʻIntonation Units.ʼ Its proper Linguistic Type is ʻchunkʼ (not ʻtimed chunkʼ). Click ʻAdd.ʼ Next we need a tier for the glosses of words and morphemes. Make a new tier called ʻGlossʼ, whose parent is ʻWords and Morphemes.ʼ Remember that this is one of our two translation tiers, so set the Linguistic Type accordingly. Click ʻAdd.ʼ 2 Remember, these names can be anything you want, but names like ʻchunkʼ and ʻtimed chunkʼ are conceptually appropriate for what theyʼll be used for. 3 Do not use an ampersand (&)! ELAN sometimes has trouble with these as they can be part of the underlying programming language. Andrea Berez & Chris Cox 39

40 InField 2010 ELAN: Exercise 4 Our final tier can be called something like ʻEnglish Free Translation.ʼ This tier will contain a translation of the entire Kannada sentence (not just of words or of intonation units), so its parent is ʻKannada Sentence.ʼ Its linguistic type is, of course, ʻtranslationʼ. Once you have added all your tiers, itʼs a good idea to check to be sure that your parent/child relationships are correct. Close the ʻAdd Tierʼ dialog box, and then Go to View > Tier Dependencies You should see a tree structure something like this: This diagram shows the following information: Kannada Sentence is linked directly to the audio stream timeline Kannada Sentence has two immediate children: Intonation Units and English Free Translation Intonation Units has one immediate child: Words and Morphemes Words and Morphemes has one immediate child: Gloss Gloss and English Free Translation have no children If you donʼt see this, youʼve made a mistake somewhere. Go back and fix your tier relationships. Step four: add annotations. Now we are going to add our annotations, working with the transcription in todayʼs file, pear_story.txt. Start with the first Kannada sentence, and add the annotation by listening, selecting, and double clicking just like in the previous exercises. Now we want to divide this sentence into time-based intonation units. First double click in the Intonation Unit tier directly below the annotation for the first full sentence. A text box opens up that is just as long as its parent. Go ahead and type the words from just the first intonation unit into that box, and then type Control-Enter to make it stick. Now we want to keep adding intonation units there are five for this first sentence. So hereʼs what you will do for each new intonation unit: Andrea Berez & Chris Cox 40

41 InField 2010 ELAN: Exercise 4 1. click on the horizontal line containing your first intonation unit so that the line itself turns blue 2. right-click on that blue line, and select New Annotation After. A new text box will open up to the right of the first one. 3. Add the text for the second intonation unit and type Control-Enter. 4. Now you want to move the vertical separation line in between the units so that it lines up with the correct point in the wave form where the first unit ends and the second unit begins. Click either of the two intonation units so that the horizontal line turns blue. Now, press the ALT key, and while pressing it, use the mouse to pick up the vertical separation bar and drag-and-drop it where you want the new division to be. Repeat this process until you have added all the intonation units for your first Kannada sentence. Notice that by selecting a single intonation unit, you can press Play and hear just that unit. Next we want to add annotations to the Words and Morphemes tier (for the sake of this exercise, put each word with its morphemes into a single annotation; for example, the entire word hann-ina will all go into one annotation box). To add your first word, double click on the Words and Morphemes tier directly below the first intonation unit. A text box that is the same length as the intonation unit above will open up. Type in your first word, and then hit Control-Enter. To add additional words, you will do nearly the same thing you did when you added intonation units (selecting the word so that the horizontal line is blue, clicking New Annotation After, etc.). The difference is that you wonʼt be able to pick up the vertical separation line and move it. Instead, because we used the ʻSymbolic Subdivisionʼ type for this tier, ELAN divides the area into equally-sized units. Continue until you have added all the words for each intonation unit. Next youʼll add annotations to the Gloss tier by double clicking directly below each word. ELAN wonʼt allow you to Add Annotation After in this tier because of the properties of this tierʼs type. Continue to add glosses to all your words and morphemes. Finally, you can add annotations to the English Free Translation tier. Notice that when you double click, the text box is as long as its parent (Kannada Sentences), and not as long as the tier that is physically above it (Gloss). Now go through the entire file and add all the transcriptions. Youʼll get faster as you practice! If you finish early, read in the ELAN manual about the following things: there is one linguistic stereotype we havenʼt used itʼs called ʻIncluded In.ʼ What is the difference between this and ʻTime Subdivisionʼ? read about tokenizing tiers. How can this help you in annotating IGT? Questions to think about: You could go even further with your IGT. For example, you could have a tier that puts each underlying morpheme in a separate annotation (good for fusional languages). How would you do this? Andrea Berez & Chris Cox 41