The robot detects human gaze and responds with an attached behavioral pattern: you can easily design your own reaction!
Different semantic structures trigger different reaction scripts. Triggered script and mood settings determine the robot’s final response.
When F2 hears that the robot hits the professor, identifying itself as the aggressor, it can express its own superiority or apologize. 5/6
1
2
Other words are the tree’s nodes and leaves, their roles are taken from the Fillmore’s list (agent is for main actor; patient is for one who undergoes the action; location; instrument; etc.).
Next, all words are replaced by their meaning – a set of semantic features. 4/6
2
1
Syntactic trees are then built from an utterance using the words’ morphological features and 600 syntactic rules.
Each word takes its own place inside the tree and gets a role. A verb is always the top of the tree and its role is predicate (p). 3/6
2
1
Words from an utterance are lemmatized (dogs→dog) and marked up with their morphological features (time, gender..) using a database and guesser for unknown words.
The example below is in English, but all features are relevant for Russian. All significant features are marked. 2/6
1
2
#linguistweets #lt0545
We are creating an emotional robot companion F2, which communicates with the interlocutor using speech (in Russian), facial expressions and gestures.
How does it understand the other person?
And how does it choose a response?
see below 1/6
Забрал из проявки фотки из Светлогорска. Когнитивная конференция озадачена: как же исследовать мышление роботов? #cogconf
1













