Utter Command HumanMachine Linguistics HumanMachine Grammar and a

Utter Command: Human-Machine Linguistics, Human-Machine Grammar, and a New Speech Interface Kim Patch July 12, 2005

Introduction • Human-machine linguistics • Human-Machine Grammar • Redstart Systems and Utter Command • Demo of Utter Command speech interface software 1

Human-Machine Linguistics • Leverages the way the human brain processes language to make spoken communications with machines fast and efficient 2

Keys to a successful speech interface • Wording that dovetails with instinct (Bicycles) • A concise, easy to remember vocabulary (Talking) • Overall framework – rules (Apple peeler) 3

Human-machine grammar • System of words and rules • Designed to limit cognitive effort • Easy to learn • Doesn’t require computers to decode natural language or handle a large number of synonyms 4

Books that have informed Human -Machine Grammer • Words and Rules by Stephen Pinker • Linked by Albert-László Barabási • The Psychology of Everyday Things by Donald Norman • The Humane Interface by Jef Raskin 5

Books – Relevant points • Words and Rules - speech involves rules that are built on the way the brain works 6

Books – Relevant points • Words and Rules - speech involves rules that are built on the way the brain works • Linked - the smaller and more connected the vocabulary, the easier it is to remember 7

Books – Relevant points • Words and Rules - speech involves rules that are built on the way the brain works • Linked - the smaller and more connected the vocabulary, the easier it is to remember • The Psychology of Everyday Things difficult to use interfaces usually have design flaws 8

Books – Relevant points • Words and Rules - speech involves rules that are built on the way the brain works • Linked - the smaller and more connected the vocabulary, the easier it is to remember • The Psychology of Everyday Things – difficult-to-use interfaces usually have design flaws • The Humane Interface - good interfaces follow habit and instinct 9

Command-Building Guidelines • Match commands to meaning • Use words the user sees on the screen when possible • Balance ease of saying and ease of remembering • Keep commands succinct • Don’t use synonyms • Conserve words by using multiple meanings 10

No synonyms in the Human-machine dictionary • "This" is the only word that refers to something that is highlighted or on the clipboard. − e. g. "This Cut", or "This Bold" • "Back” and “Forward” are the only words that refer to moving a selection − e. g. "Word Back 3" moves the word nearest the cursor 3 words to the left. 11

Conserve words by using more than one meaning: “Top” • The word “Top” has two meanings in the Renji dictionary − The beginning of a document e. g. “Go Top” − The top half of a paragraph or document e. g. “Graph Top” 12

Conserve words by using more than one meaning: Numbers • Number of times to hit a key − "Backspace 3" • Selecting an object like a line or paragraph − "3 Lines" • Absolute measure − “Volume 50" 13

Word Pairs make vocabulary succinct and easy to remember • • • Back, Forward On, Off Before, After Up, Down Long, Short Open, Close 14

What do you call the word before the cursor? • "3 Befores" selects the three words before the cursor: "3 Afters" selects the three words before the cursor: 15

Combined Commands make vocabulary succinct and easy to remember • 3 Lines Bold - selects and bolds the three lines below the cursor • 3 Graphs Cut – selects and cuts the three paragraphs below the cursor 16

• Dedicated to promoting Human. Machine Grammar • We all use speech • First product: Utter Command 17

Utter Command • Will launch in Fall • DNS Pro required • Free sample macros available soon at www. redstartsystems. com • Sign up for email updates at www. redstartsystems. com 18

Conclusion • Human-Machine Linguistics • Human-Machine Grammar • Utter Command 19