Documentation - Linguistic Data Consortium
Contents
1. the cursor Set SU type backchannel for current segment Set SU type incomplete for current segment Set SU type question for current segment Set SU type statement for current segment Unset SU type for current segment Insert lt background gt lt background gt in text Insert lt breath gt in text Insert lt cough gt in text Insert lt foreign language gt lt foreign gt in text Insert lt Iname gt in text Insert lt laugh gt lt laugh gt in text Insert lt lipsmack gt in text Insert lt sneeze gt in text Appendix 3 Sample Transcript Output Meeting recording excerpt This excerpt shows a tdf version of a meeting recording transcript Meetings typically contain 3 8 speakers this particular recording included 4 speakers This file is from the Rich Transcription 2007 Meeting Recognition Evaluation training dataset file NIST_20060223 1416 file unicode channel int start float end float speaker unicode speakerType unicode speakerDialect unicode transcript unicode section int turn int segment int sectionType unicode suType unicode 7 MM sectionTypes None None 77MM audioBinding NIST_20060223 1416_HM 01_Subj 159_16K16b 0 NIST_20060223 1416_HM 01_Subj 159_16K16b 0 NIST 20060223 1416 HM 02 Subj 161 16K16D 0 NIST 20060223 1416 HM 02 Subj 161 16K16b 0 NIST 20060223 1416 HM O03 Subj 162 16K16b 0 NIST 20060223 1416 HM 03 Subj 162 16K16b 0 NIST 20060223 1416 H2. upper right This will set subsequent segments to this speaker eliminating the need for you to input information for this speaker again As soon as you encounter speech from another speaker you can clear the selected speaker by clicking CLR in the widget panel This will force you to assign a new speakerlD for the next segment you create Set a speaker for further segmentation 5 2 Editing SpeakerlDs There are several ways to modify speaker information 5 2 1 Correcting Speaker Information To change speaker information for a single segment you can right click on the speakerlD box for that segment and select Assign new speakerlD single The speakerlD dialog window will appear and you can modify speaker information You can also change speaker information for a single segment by first placing your cursor within that segment then hitting the NSI new speaker information button on the widget panel To change speaker information for all segments from a given speaker you can right click on the speakerlD box for one of the segments and select Edit speaker information global or you can click the ESlg edit speaker information global button on the widget panel 5 2 2 Merging Two Speakers To reassign a speakerlD to another existing speaker for a single segment you can right click on the speakerlD box for that segmen
3. you hover over the widget XTrans will display a longer description of each button s function e MRG Merge Reassign speaker to another existing speaker for just the current segment e MRGg Global merge Merge two speakerlDs throughout entire file e VOS View one speaker Display only segments for selected speaker e VAS View all speakers Display segments for all speakers e SRT Sort Sort speakers listed on the speaker panel toggling between chronological and alphabetical display helpful for doing speaker ID verification and double checking spelling of soeaker names e CLR Clear Clear the check boxes on the speaker panel e NSI New speaker information Assign new speakerlD for just the current segment e ESIg Edit speaker information global Change speakerlD information for the selected speaker throughout the whole file e LRS Listen random segment Listen to a random segment from this speaker helpful for speakerlD verification e LAS Listen all segments Listen to all segments from this speaker helpful for speakerlD verification e LAG Listen all gaps Listen to the unsegmented audio gaps helpful for doing quality control to catch unsegmented speech 5 The Audio Panel This region includes the waveform display and audio control buttons The audio control functions at the upper righthand portion of the audio panel include from left to right play repeat pause restart and stop Next to the audio control functions is a ho
4. 01_Subj 159_16K16b 0 32 1949612403 38 7150387597 Sub 3 159 male native What we want to try to achieve in this meeting is your personal satisfaction 0 9 9 IST_20060223 1416_HM 01_Subj 159_16K16b 0 38 8349612403 46 3450387597 Sub j 159 male native Is because I want you to be comfortable with a with asking me any of the financial questions that you have concerns about 0 10 10 NIST_20060223 1416_HM 01_Subj 159_16K16b 0 46 49 48 17 Subj 159 male native If I m unable to answer them 0 14 14 NIST_20060223 1416_HM 01_Subj 159_16K16b 0 48 52958159 49 8606276151 Sub 3 159 male native I will tell you so 0 15 15 12 For more information about LDC s meeting recording transcription efforts see http projects ldc upenn edu Transcription NISTMeet NIST_20060223 1416_HM 01_Subj 159_16K16b 0 50 26 53 92 Sub j 159 male native As a registered investment adviser the difference that I make 0 16 16 NIST_20060223 1416_HM 01_Subj 159_16K16b 0 53 92 57 97 Subj 159 male native to you is that I am liable for fully disclosing 0 17 17 NIST_20060223 1416_HM 01_Subj 159_16K16b 0
5. Keybindings The default XTrans keybindings are shown in the table below XTrans keybindings are user configurable by going to the Tools gt Edit key bindings menu or hitting ait T ait z A dialog box will open Place your cursor on the second column of the row containing the keybinding you want to change Double click on a cell to edit it Erase the existing text type the double quoted string of key combinations you want then press Enter Click on the ox button to commit the change If you undo the modification before committing it use the Reset button The Reset a11 button will cancel all modifications from the current editing session The Default button restores the default setting for the current keybinding and the Default A11 button does the same for all keybindings Use the cance1 button to cancel all changes and close the dialog box It is also possible to save the changes as a configuration file using the Tools gt Save configuration file menu or ait T Ait s To load the saved configuration file use Tools gt Load configuration file menu or ait T Alt L Category Keybinding Description Toggle between two actions play begin_selection vs Alt M stop_playing finalize_selection Alt S Stop playback Alt P Pause resume playback Alt Space Play current region Alt Return Play from the right edge of current selection Alt Play just ending part of current selection Alt Play from the beginning of current waveform wi
6. M O04 Subj 160 16K16b 0 NIST 20060223 1416 HM O04 Subj 160 16K16b 0 MM sectionBoundaries 0 0 9999999 0 NIST_20060223 1416_HM 01_Subj 159_16K16b 0 0 538744769874 5 62125523013 Subj 159 male native So I d like to thank you all for coming this afternoon to this financial literacy workshop 0 0 0 IST_20060223 1416_HM 01_Subj 159_16K16b 0 5 65937238494 10 8506276151 Subj 159 male native What I want to do is try to answer and address all of your concerns that you have 0 1 1 IST_20060223 1416_HM 01_Subj 159_16K16b 0 10 9093723849 12 8306276151 Sub j 159 male native about money and money matters 0 2 2 IST_20060223 1416_HM 01_Subj 159_16K16b 0 13 0593723849 16 3806276151 Sub j 159 male native I would be glad to speak to you about all types of 0 3 3 IST_20060223 1416_HM 01_Subj 159_16K16b 0 16 9949612403 20 22 Subj 159 male native products you can use to accumulate money and wealth 0 4 4 IST 20060223 1416 HM O01 Subj 159 16K16b 0 20 61 23 52 Subj 159 male native For your retirement for your education concerns 0 6 6 IST_20060223 1416_HM 01_Subj 159_16K16b 0 23 86 26 83 Sub 3 159 male native for buying a home reducing for reducing debt 0 7 7 IST 20060223 1416 HM O01 Subj 159 16K16b 0 26 9949612403 32 0100775194 Subj 159 male native Whatever matters that uh whatever financial matters that you want to address 0 8 8 IST_20060223 1416_HM
7. Using XTrans for Broadcast Transcription A User Manual Version 3 0 Linguistic Data Consortium October 11 2007 Developer Haejoong Lee lt haejoong ldc upenn edu gt Manager Stephanie Strassel sstrassel ldc upenn edu http www idc upenn edu Projects Transcription 1 WEF O CU CTHONY oe se Ee ee A E ee ee 3 2 oo VERWERK ee eee tree Ee AG WEE el eie 3 2 Getting Slane dopriate AE E AE EAE EA EOE A e 6 al Open Audio File oos ie ae Pe es ee N Te oe Ee 6 3 2 Open Transcript Fles Rd ed AG ee de EE EES 7 3 3 Associate Transcript and Audio File RENEE EER ER EE WK ede 7 4 Audio Segmentation is ESE ERK E EKE K R ER RENEE KERR N Wese RR K ER D KE Ge SR 7 4 1 Mark a Region and Insert a Segment ccccceeeeeeeeeeeeeeeeeeeeeeeeeeeees 7 4 2 Modifying Segments see ee ee ee ee ee ee 8 43 Overlapping Segments see RR ee 8 5 Assign Speaker Information se KERE NEE EER EE EG RE RE ee gee ee 8 5 1 Creating Additional Segments for Existing Speaker ss see 10 52 Editing SpeakerilDS see ii Ee EE EE Ee DE EG ES Ee 10 5 2 1 Correcting Speaker Information ccceeeeeeeesseeeeeceeeeeeeeeees 10 5 2 2 Merging TWO Speakers cccccceeeeeeeeeeecneeeeeeeeeeeeeeecnneeeeeeeeeees 11 6 Overlapping SSS CI castes EE n EG N EG SE Die ennt 11 7 Assigning Sections Story Boundary Labeling sssssssessssssssssrrrrnrsessssesee 11 8 Assigning SUs Sentence Unit Labeling cc cceceeeeeeeeeee
8. d its type is report and that the second section starts at 425 3 seconds and its type is nontrans and this is the last section last ones are ignored These two lines are optional Besides these comments are allowed anywhere in the file except for the first line A comment line starts with and ignoring this line doesn t cause any problem to the data integrity with one exception that lines starting with MM might have meaningful information as shown in the example above 12 Saving and Exporting Transcripts To save your work type ait F Ait s or select Save or Save As from the File menu Files are saved automatically at one minute intervals to a backup file with a sav extension If XTrans crashes users can recover the file by renaming the backup file by removing the sav extension In normal conditions the backup file is removed when the user closes the file You can export your transcript file into two other common data formats trs used with the Transcriber tool and txt used with older LDC transcription tools To do this select the Export function from the File menu For instance telephone speech transcripts usually do not include section information 10 A third export format SGML Weblog is used for some specialized LDC annotation tasks and is not intended for general use Appendix 1 Recommended Strategy There are many different ways to interact with XTrans to create a time aligned transcript The
9. dicate time in minutes and thousandths of seconds We are currently investigating existing technology for adjusting playback speed without changing sample rate The image below summarizes these features Pause Cursor position start time Restart end time segment duration si Repeat Playback Speed Reset Playback Speed Mono Stereo Selector eo Fielative Volume i no effect on 1 channel audio Waveform Amplitude Slider to control location in audio file Current selection Before you begin transcription you should also familiarize yourself with the most common playback and waveform display and manipulation functions of XTrans While all of these functions can be achieved using the mouse in the standard way transcribers will find it faster to use the keyboard only for most tasks Default keybindings for XTrans are described in Appendix 2 3 Getting Started 3 1 Open Audio File On the file menu pull down Open audio file Currently XTrans accepts sph and wav audio format After you load an audio file your display will look something like this LIES VOS VAS _SRT CLR NSI Esi Mee mead sod s0 I 00 lz ma 250 felt a i ie i To be precise XTrans supports any audio formats that libsndfile supports which includes more than dozen of file formats Unfortunately mp3 is not included however we pla
10. eeeteeeeeeeeeeees 12 9 12102101 DUI se ene E Ai eer er Pore A ere er E AAEE 13 9 1 Language Input FUNEUGNS sees se ei ERA se ee EN EG de Ee 13 9 1 1 Changing Input for Windows Systems i iese sees see ee ee 13 9212 Changing Input for Unix SY StOM NS es ER SEK Ee Ke Ie 14 92 Transcript Markup Shortcuts users sAS Rae ES RES REEKS K ER se es We KERE es ede 15 ESE ie EE EE EA N 15 10 Other NM oe ee cere ea De ie EA ie ea ee AA ee ee de 15 11 BEEN TT Oe 15 12 Saving a d Exporting TrarserupiE RE RE ERKEN WE EA ER WE ER ea 16 Appendix 1 Recommended Strategy sesse ee ee ee nn 17 Appendix 2 KeybindingS EE EE ES EG GE RR EG DO Se ie 18 Appendix 3 Sample Transcript OUIDUI si sk GER N Be RR EK kw Aa ie be 20 1 Introduction To support the demand for efficient careful and consistent transcription LDC has created a next generation speech annotation toolkit XTrans to support a full range of speech annotation tasks including quick and careful and rich transcription and MDE annotation of broadcast audio telephone speech and meetings XTrans is a multi lingual multi platform transcription toolkit Powered by Qt s international language support it can be used for transcription tasks in many different languages Based on Qt it can be easily ported to most UNIX derivatives Microsoft Windows and Mac OS X XTrans consists of several re usable components such as text pane and waveform display Most of the components are written in Py
11. esponds to this region of the program now has a color coded section value assigned whereas regions without section assignments remain grey A horizontal line also appears in the section column at the segment corresponding to the start of this section Note that all section boundaries must correspond to the beginning of a speaker segment 8 Assigning SUs Sentence Unit Labeling SU assignment is very simple if you follow the practice of creating one segment per sentence during the segmentation task To label the SU type for an existing segment simply enter one of the following commands while your cursor is in the segment either on the waveform or in the corresponding transcript region 7 Alternatively you can right click on the box in the section column to assign or re assign section for a region You can also type Ctrl E Ctrl S to edit type information for an existing section EtEIHL Ctrl U CEELTS Set statement SU CtritI Ctrl U Ctrit9 Set question SU Cerl 1 Ctrl U Ctrl4 1 Set incomplete SU Cerl Ll Cerl U CESLB Set backchannel SU Ctrl D Ctr1 U Unset SU markup The transcript itself will not change but you will notice that the box in the SU column that corresponds to this segment now has a color coded SU value assigned whereas segments without SU assignments remain dark grey 9 Transcription There are a handful of features in XTrans to facilitate transcription itself The basic transcription process
12. following is a synopsis of LDC s recommended strategy for creating broadcast transcripts with XTrans Note that most of these functions are keyboard rather than mouse based commands For quick transcription it is strongly recommended that transcribers choose keyboard over mouse based functions as much as possible This takes a little getting used to but you will find it much faster and easier to use the keyboard only rather than switching between keyboard and mouse and it s easier on your wrists Quick Guide for Quick Transcription 1 open audio file File gt Open audio file 2 open new transcript file File gt New 3 associate audio and transcript Edit gt Blindly associate transcript to audio 4 begin playback and mark segment start Alt M 5 stop playback and mark segment end Alt M 6 insert segment Ctrl N CtritInsert on nix 7 assign speaker information dialog box use tab amp arrow keys to select options 8 create next segment repeat 4 7 To create segment for same speaker first select speaker in speaker panel then repeat steps 4 6 9 assign section boundary Ctrl I Ctrits 10 assign SU type Ctrl I Ctrl U Ctrl __ 11 transcribe the segment 12 save your work frequently Alt F Alt S 13 repeat steps 4 12 14 save and exit 11 Some transcribers prefer to fully segment the file the go back and transcribe it while others prefer to transcribe as they segment Appendix 2
13. gn or re assign SU for a segment AR for Arabic The user can set their desired input language by selecting from that blue box Windows2000 To enable the input of Arabic text on Windows2000 1 Go to the control panel 2 Select Regional options 3 Select Input locales 4 Hit the change button 5 Add required languages Note that it may be necessary to reboot your machine for changes to take effect 9 1 2 Changing Input for Unix Systems There are multiple ways to change the language input mode on a NIX system Method 1 Keyboard Mapping With this method the user sets the keyboard mapping via a separate terminal window xterm For instance Change input mode to English is a shell prompt setxkbmap us Change input mode to Arabic is a shell prompt setxkbmap ar Note that the input mode is set to English first then Arabic because once the input mode has been changed to Arabic you will be unable to type any English command on the terminal In bash GNU Bourne Again Shell if you set it to English first even though you cannot type you can still use the up arrow key to recall the command line that was used to set the input mode to English when you need to type in English Method 2 SCIM Another option is to use an input method like SCIM Smart Common Input Method SCIM provides a user friendly full featured input method user interface for POSIX style operating systems including Linux FreeBSD and other U
14. involves typing the words that correspond to an existing segment Recall that the transcript and waveform display are linked in that you can click on a segment in the waveform and XTrans will find that segment in the transcript and vice versa 9 1 Language Input Functions By design XTrans supports multilingual text input The default input mode is English 9 1 1 Changing Input for Windows Systems You can change input methods on a Windows system by accessing the Language Bar functions of Windows WindowsXP For instance to enable the input of Arabic text on Windows XP 1 Make sure you have an XP installation disk 2 Go to Regional and Language Options 3 Select the Language tab 4 Select Install files for complex script and right to left languages This will copy files from somewhere to somewhere else and add Arabic to the lists of available languages to add 5 Hit Apply and reboot the machine at the prompt 6 Go back to Regional and Language Options 7 Go to the Advanced tab and select Arabic from the Languages for non unicode programs list 8 Go back to the Language tab and make sure that Arabic is listed as an installed service under Details and Text Services and Input Languages The result of this process should be a little blue box in the lower right corner of the screen that displays EN when English is the selected input language and Alternatively you can right click on the box in the SU column to assi
15. is region displays three kinds of segments in three separate columns Each column contains boxes that correspond to segments that have already been created for that type of segment The left most column shows sections i e story boundaries e Reports e g news stories are in blue e Conversational e g interviews are in green e Non news e g commercials are in light yellow e Unassigned sections are in grey The middle column shows speaker segments one color per speaker The speakerlD is also displayed The right most column shows SUs sentence units e Statement SUs are shown as dark blue Question SUs are shown as red Incomplete SUs are shown as green Backchannel SUs are shown as yellow Unassigned SUs are shown as dark grey 2 The Transcript Panel This region shows transcribed text If you select a region of the transcript by clicking your mouse somewhere within that segment it will be highlighted in green and the corresponding audio segment will be highlighted in the waveform display with a blue box surrounding it 3 The Speaker Panel This region shows the list of all soeakers assigned to segments in the file Each name is color coded to match that speaker in the speaker column of the segment panel display 4 The Widget Panel This region features several shortcut buttons for doing speaker verification and other kinds of quality control Each button or widget has an abbreviation representing its functionality If
16. isting segment hit ctr1 p or on NIX ctr1 Delete To join or split two existing segments type the following commands Ctrlt Jd Join the current and following segments Ctrl Return Split segment into two at cursor point You can also correct or change the start or end boundary points of an existing segment in several ways With the mouse hold down the snift key then place your cursor on the start or end boundary you wish to modify and drag it to the correct location XTrans automatically adjusts the segment start and or end time to your new selection Segments can also be adjusted without the mouse using the standard XTrans keybindings described in the Appendix 4 3 Overlapping Segments XTrans prohibits the user from creating overlapping segments for a single speaker However segments from two different speakers can overlap See Section 6 for a full treatment of overlapping speech segmentation 5 Assign Speaker Information As soon as you insert a new segment XTrans prompts you for speaker information for that segment Note that this operation joins two consecutive segments for the same speaker vos vas ser ca NSI ESig MRG MRG LRS LAS LAG C ao EE EE Jal FAFA a o a za e wa sa m a 1a MPN HAAHI da ae a da di da a Seer th normal There are three fields to complete Speaker ID name o
17. n to add mp3 support in future versions Multiple time aligned audio files can be loaded e g for meeting or telephone speech transcription Note that XTrans will display audio file information file name path and sampling rate when you right click on the waveform display 3 2 Open Transcript File If you are starting a new transcript select New from the file menu If you are returning to an existing transcript select Open from the file menu Currently XTrans can directly open files in its native tdf format To open an existing transcript in Transcriber trs format select Import gt Transcriber from the file menu 3 3 Associate Transcript and Audio File Before you begin transcription you must first link the audio with your transcript To do this select Edit gt Blindly associate transcript to audio You cannot proceed without doing this step If you have opened an existing transcript you may also select Edit gt Blindly associate transcript to audio However it s also possible to use the Edit menu to manually Associate transcript to audio to specify which speaker is associated with which audio channel This is most useful when working with meeting recordings with more than two speakers 4 Audio Segmentation The first stage of transcription is segmentation which refers to the process of virtually chopping a long audio file into smaller meaningful units e g sections speaker t
18. ndow Alt Rewind one second Alt Fast forward one second Ctrl lt Fast grow current selection move left edge to the left by a couple of pixels Ctrl gt Fast shrink current selection move left edge to the right by a couple of pixels Alt gt Fast grow current selection move right edge to the right by a couple of pixels Alt lt Fast shrink current selection move right edge to the left by a couple of pixels Ctrl Grow current selection move left edge to the left by one pixel Ctrl Shrink current selection move left edge to the right by one pixel Alt Grow current selection move right edge to the right by one pixel Alt Shrink current selection move right edge to the left by one pixel Alt A Zoom in where the cursor is placed Alt Z Zoom out where the cursor is placed Alt X Make current selection fit the current waveform window Segmentation Ctrl N Ctrl D Ctrl J Ctrl Return Ctrl 1 Ctrl S Ctrl E Ctrl S Ctrl D Ctrl S Ctrl 1 Ctrl U Ctrl B Ctrl 1 Ctrl U Ctrl l Ctrl 1 Ctrl U Ctrl Q Ctrl 1 Ctrl U Ctrl S Ctrl D Ctrl U Ctrl Shift G Ctrl Shift B Ctrl Shift H Ctrl Shift F Ctrl Shift N Ctrl Shift L Ctrl Shift P Ctrl Shift S Insert a segment for the selected region Ctrl Insert on nix Delete current segment Ctrl Delete on nix Join the current and following segments Split current segment into two at cursor point Insert section boundary Edit section boundary type Delete section boundary on the left of
19. nix as well as a development platform to make input method development easier SCIM runs as a separate program and registers itself as an input method of X Windows system Then users change their environment setting to select their input method e g in bash export XMODIFIERS im SCIM Any tool that understands X Input Method XIM Protocol will get keyboard input from the designated input method For more information on installing and using SCIM please visit the SCIM webpage http Awww scim im org Other Methods For future versions of XTrans we are researching input methods that can be directly incorporated into the tool for instance UIM 9 2 Transcript Markup Shortcuts The following shortcuts can help the transcriber add required markup to transcripts Ctrl Shift G Insert lt background gt lt background gt in text CtritShifttB Insert lt breath gt in text Ctrl Shift H Insert lt cough gt in text Ctrl Shiftt F Insert lt foreign language gt lt foreign gt in text Ctrl Shift N Insert lt lname gt in text Ctrl Shift L Insert lt laugh gt lt laugh gt in text Ctrl Shift P Insert lt lipsmack gt in text Ctrl Shift s Insert lt sneeze gt in text Ctr1 zZ undo the last change Ctr1 yY redo the last undo 9 3 Spell Check Aspell based spell checker is implemented but only enabled in the Linux package Users must install aspell and the dictionaries themselve
20. one per speaker 4 26 0485 4 151050 4 29 7408 14 6357 gt R xA A 9 o A 420 250 an 5 an o 4a 4550 sp o 405 4100 4150 Viv Cnn eas AE RE EES EE ER EE EE a 4 J5 normal Clicking on any of these horizontal bars will select that segment in the waveform and in the transcript Overlap is not explicitly indicated in the transcript display but can easily be extrapolated from the transcript file itself by examining segment start and end times 7 Assigning Sections Story Boundary Labeling To create a section story conversation or commercial boundary click on the horizontal bar of the speaker segment in the waveform that corresponds to the beginning of the new section The reverse action clicking inside the transcript to select a segment also works to select the segment Then type ctri I ctri s to insert the new section boundary You will be prompted to select the appropriate section type report conversational or non news EE Xtrans 19981217_0130_0200_CNN_HDL tdf E 3 File Edit View Tools E Select Section Type 2 x Please choose a section type for the new section and press the OK button report nonnews 3 11 3172 3 11 3831 3 15 7358 2250 230 2 2250 2310 2350 2400 2450 2500 2550 ril ma After you ve assigned the section type the transcript itself will not change but you will notice that the box in the section column that corr
21. r unique identifier for speaker Speaker Type male female child or unknown Speaker Dialect native other for non native or foreign If you know the name of the speaker type it into the SpeakerlD box Otherwise use an anonymous ID To assign a new anonymous SpeakerlD click the New button in the SpeakerlD dialog box XTrans will automatically assign the next unused ID in the form speaker1 speaker2 etc As soon as you have completed the speaker information XTrans will record this information and update the transcript display with three new pieces of information e abox in the speaker turn column that corresponds to this segment color coded and displaying the speaker name e acolor coded horizontal bar in the waveform display that indicates the duration of the segment e acolor coded speaker entry in the speaker panel slelsd vos vas ser on NS Es Meel vecd LRS Las LAG You will notice that if you select a segment by clicking on the horizontal bar for that segment the segment will be highlighted with a blue box and XTrans will jump to the corresponding region in the transcript window Any text in the corresponding transcript segment will also be highlighted in green 5 1 Creating Additional Segments for Existing Speaker When you need to create another segment for an existing speaker you can click that speaker s name in speakerlD panel on
22. rizontal slider bar that controls playback speed The default playback speed is realtime playback Sliding the bar to the right speeds up playback sliding it to the left slows down playback Beside the slider bar is a button that displays the playback speed Clicking this button restores the playback speed to realtime To the right of the waveform display are two vertical slider bars The first of these controls waveform amplitude display The second bar controls relative audio playback volume It is relative because it doesn t change the overall volume but changes the distribution of loudness over the input audio channels Sliding this bar to the very bottom will mute the channel Above this second vertical bar are two checkboxes that control mono vs stereo playback The default setting for broadcast single channel audio is mono one box checked but transcribers can check the second box to enable stereo playback helpful if you are using headphones In the upper lefthand corner of the waveform display you will notice a series of four numbers The first number shows your current cursor placement The second and third numbers indicate the start and end time of your current selection and the fourth number indicates the duration of your current selection The long horizontal slider bar running along the bottom of the waveform display adjusts your position in the audio file Numbers displayed at the top and bottom of the waveform in
23. s since the current XTrans package doesn t include those components Spell check will be added to the Windows version in the future 10 Other Features To be added 11 Data Format The default XTrans data format is tdf or Tab Delimited Format This is a simple file format in which data is represented as a set of records which are in turn a set of fields separated by tab characters The XTrans tdf format is a set of 13 field records plus some meta information as described below 1 file file name or id string 2 channel audio channel number 3 start start time number 4 end end time number 5 speaker speaker name or id string 6 speakerType speaker type string 7 speakerDialect speaker dialect string 8 transcript transcript string 9 section section id number 10 turn turn id number 11 segment segment id number 12 sectionType section type string 13 suType SU type string Each record in an XTrans tdf file refers to a segment In addition to the body of segments there are a few lines of meta information within the tdf file The first line of the format declares the above field specification for segments in the following form file unicode channel int start float The second and third lines specify where the real section boundaries are and what types they are For example 7 MM sectionTypes u report u nontrans None MM sectionBoundaries 0 0 425 3 9999999 0 means that the first section starts at 0 0 second an
24. t and select Merge into another speaker single The speakerlD dialog window will appear and you can select the new speaker assignment You can also change speaker information for a single segment by first placing your cursor within that segment then hitting the MRG merge current speaker into another speaker button on the widget panel To reassign speaker information for all segments from a given speaker i e merge two speakerlDs into one for the whole file you can right click on the speakerlD box for one of the segments and select Merge into another speaker global or you can click the MRGg merge global button on the widget panel 6 Overlapping Speech As noted above XTrans prevents users from creating overlapping segments for a single speaker However transcription of overlapping or simultaneous speech is very straightforward Although broadcast audio is single channel XTrans creates a separate virtual channel for each unique speaker There is then no prohibition on overlapping segments from two or more different speakers To create an overlapping speech segment transcribers simply perform segmentation as described in Section 4 mark the start and end of the segment associate it with the correct speaker and insert the segment Overlaps of two three or more speakers are possible XTrans visually displays the overlapping speech regions in the audio panel by showing overlapping color coded horizontal bars for each segment
25. thon with some components written in C The current version of XTrans runs on FreeBSD Linux and Windows platforms 2 Tool Overview This section gives a very brief introduction to the tool design and layout XTrans is divided into five main areas labeled on the image below WB Xtrans 19981217_0130_0200_CNN_HDL tdf Fi ls omprehensive operation he operation employs US Navy aircraft flying from the decks of the USS nterprise US Air Force and Royal Air Force aircraft operating from land pases in the region and Tomahawk cruise missiles launched from US Navy ships at sea and United States Air Force B fifty twos he Desert Fox manoeuver diverted America s attention away from he crossfire on Capitol Hill surroundi eachment hat has left some people sceptical about the timing of the operation olf Blitzer reports President Clinton is trying to allay those concerns President Clinton gave the final order to attack Iraq on the eve of an almost ertain vote to impeach him But he told the nation that wasn t a factor Badaam Husein and the other enemies of peace may have thought that the serious debate currently before the US House of Representatives would distract America or weaken our resolve to face him down vas SRT CLA NSI ESig MRG ME Ls LAS LAG reed N wa met aat in Amarinale vital intaracte var 1 The Segment Panel Th
26. urns sentences or other salient units Segmentation can be done with the keyboard only with the mouse only or witha combination of both After you ve become familiar with basic XTrans functionality you will find that using only the keyboard is both faster and more intuitive than using the mouse 4 1 Mark a Region and Insert a Segment To begin place your cursor at the start of the waveform and type ait m This starts playback and marks the start time of your selection When you reach the end of a segment you wish to mark for instance the end of a sentence type ait m again This will stop playback and also mark the region you have just played surrounding it with a red box in the waveform display This step will be done automatically in future versions of XTrans For Quick Transcription of broadcast data we recommend starting with sentence based segmentation 2 Selecting a region to segment can also be done with the mouse holding down the left mouse button and swiping over the selected region but it is strongly recommended that transcribers use keyboard functions for segmentation instead of the mouse After you ve marked your segment you must insert it into the transcript To do this hit ccri n or on NIX ctri iInsert To proceed with segmentation go to Section 5 4 2 Modifying Segments To delete an ex
Download Pdf Manuals
Related Search