6. 海外の動向
Winograd (1970s) SHRDLU: シミュレーションでの物体操作
Kollar+ 2010
HRI 2010 Best Paper
• 入力:移動表現、LRF、オドメトリ、画像
• 例:”Go down the hallway”
Yu+ 2013
ACL 2013 Best Paper
• 入力:ビデオおよび内容を表す文
• 例:”The person to the left of the backpack carried
the trash-can towards the chair”
DARPA BOLTプロジェ
クト
• 約44億円/年を投資(2011~15年)
• 翻訳と並びGrounded Language Learningが1つの柱
13. (付録)座標系の推定結果
Place-on Move-closer Raise Rotate
Jump-over Move-away Move-down
Loglikelihood
Position
Velocity
Training-set likelihoodMotion “place A on B”
No verb is estimated to have WCS
-> Reference-point-dependent verb
40. rospeexが提供する機能
rospeex core
Dialogue
management
(written by user)
Speech
synthesis
Speech
Output
Speech
recognition
Rospeex cloud TTS
Noise
reduction
Voice activity
detection
Third party’s ASR API
Browser UI
OR
Smarphones On-board mic
Task management
Rospeex cloud ASR
What time
is it? It’s 6 pm.
Third party’s TTS API
Speech synthesis
designed for robots
WER = 7.9% for IWSLT tst2011
(1st Place Winner: IWSLT12, 13, 14)