Top Banner
Realtime Speech to Text: A Means to an End Mark J. Golden, CAE Executive Director and Chief Executive Officer National Court Reporters Association LangTech 2008 February 28, 2008
17

Realtime Speech to Text: A Means to an End

Jan 15, 2016

Download

Documents

maisie

Realtime Speech to Text: A Means to an End. Mark J. Golden, CAE Executive Director and Chief Executive Officer National Court Reporters Association LangTech 2008 February 28, 2008. Speech Capture. Three variables The human element - PowerPoint PPT Presentation
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Realtime Speech to Text:  A Means to an End

Realtime Speech to Text: A Means to an End

Mark J. Golden, CAEExecutive Director and Chief Executive Officer

National Court Reporters Association

LangTech 2008

February 28, 2008

Page 2: Realtime Speech to Text:  A Means to an End

Speech Capture

• Three variables– The human element

• Skill and other human judgments that go in to creating the input to the text processing system

– Technology– The functional requirements of the party for

whom the text is being produced• What will the text be used for?

Page 3: Realtime Speech to Text:  A Means to an End

Functionality

• Two parameters– Speed

• How quickly is text required?

– Accuracy• How exact must the translation be?

• Interrelationship of these two parameters– E.g.; if some delay in production of the final text is

acceptable, there is an opportunity to review and revise in order to correct for any deficiencies in the initial capture and translation.

Page 4: Realtime Speech to Text:  A Means to an End

Realtime

• There are many speech capture/text production applications where you need text as soon as possible but not immediately

• There are some applications where a comprehensible text is needed immediately, but a rough translation is adequate

• There are numerous applications, however, that rely upon true realtime– Instant text production, with no opportunity to revise or

correct– High degree of accuracy

Page 5: Realtime Speech to Text:  A Means to an End

High-Tech Courtroom

• Reduce costs

• Improve efficiency

• Better customer service

• Secure data

Page 6: Realtime Speech to Text:  A Means to an End

Litigation Support• Immediate access to

the record– Annotation– Highlights– Search– Instant message

• Complete multimedia record

Page 7: Realtime Speech to Text:  A Means to an End

Evidence Presentation

• Realtime text hyperlink to multimedia record

• Immediate display of evidence and other documentation

Page 8: Realtime Speech to Text:  A Means to an End

Virtual Justice• Complete access

to multimedia record

• Remote video participation by parties or witnesses – from anywhere in the world

Page 9: Realtime Speech to Text:  A Means to an End

Communication Access

• Full and effective participation

• Complete understanding

Page 10: Realtime Speech to Text:  A Means to an End

Stenographic Realtime

• Only current method for high accuracy, immediate voice-to-text translation– Multiple speakers

– Near perfect accuracy at high speeds

• Serves as foundation to the applications previously described

Page 11: Realtime Speech to Text:  A Means to an End

The Question of Accuracy

Page 12: Realtime Speech to Text:  A Means to an End

Realtime Accuracy: 100 Percent

The hurricane, with winds exceeding 150 miles per hour, is expected to hit the coast by 10 in the morning. Local government officials have ordered all coastal residents to leave their homes and move farther inland. More to come later.

Page 13: Realtime Speech to Text:  A Means to an End

Realtime Accuracy: 90 Percent

The hurricane, with winds exceeding 150 miles per hour, is expected to hit the coast by TEPB in the morning. Local government officials have ORD/-D all coastal residents to HRAOEFB their homes and PHOFB farther inland. More to come later.

Page 14: Realtime Speech to Text:  A Means to an End

Realtime Accuracy: 80 Percent

The hurricane, with winds exceeding 150 miles per hour, is EBGS/PEBGT/-D to HEUT the coast by TEPB in the morning. Local government officials have ORD/-D all coastal RES/TKEPBTS to HRAOEFB their homes and PHOFB farther EUPB/HRAPBD. More to come later.

Page 15: Realtime Speech to Text:  A Means to an End

Mistranslates: Steno and Voice• How do you justify stopping the defendant's

vehicle?– How do you just if I stopping the defendant's vehicle?

(voice)– How do you just if he stopping the defendant's vehicle?

(steno)

• Do you recognize this document I've just marked as Exhibit 1?– Do you recognize the stockman I've just marked as it's a

bit 1? (voice)– Do you recognize this doctor I've just marked as Kent

1? (steno)

Page 16: Realtime Speech to Text:  A Means to an End

Voice Capture ≠ Instant Translation

Appropriate Technology +

Competent Reporter

High Accuracy and Immediate Translation

Full, Complete and Instant Understanding

Page 17: Realtime Speech to Text:  A Means to an End

Thank You!

Questions and Answers