相槌

US2015209664
"[0142] As another example, the magic mirror device 1500 could communicate with an eBook reader controlled by the user to determine which book the user is currently reading and which page of a particular book the eBook reader is currently displaying. The magic mirror device 1500 could then chime in（相槌を打つ）at exactly the right time (e.g., momentarily after（一瞬後）the user has turned the page) by displaying a head of a fictional character and outputting an audio stream with commentary to enhance the story being told on the legacy media platform (i.e., the book the user is reading)."

US2008256063
"[0044] Taking the text at the start of the conversation as an example, the customer says "Can I make a reservation? (can make a reservation?)" in the text 30-1 while the customer says "I am looking for a rental car (i am looking for car)." in the text 35-1. As can be seen, the contents are different. In addition, the customer inquires the grade and price of the car in the text 30-1 while such inquiries are few and the main subjects are the driving record and the like in the text 35-1. On the other hand, both of the texts 30-1 and 35-1 include various contents, such as an exclamation（相槌）"Ahh" and a phone number, which are hard to say that they determine the occurrence of the event." 

WO9855980
"Deciding when to intervene, if the user has not asked for help, is another difficult matter. Speech systems that use an open microphone, by necessity, interpret a period of silence at the end of an utterance as the end of a conversational turn. In the oral reading tutoring task, pauses of several seconds in the middle of an attempt at reading a sentence are not uncommon. Therefore, the standard assumptions about turn-taking behavior in spoken language systems do not apply to this task. In addition, there are times when it may be appropriate for the tutor 10 to intervene twice in a row; for example, when the student struggles with a word or is unsure of what to do next. The appropriate length of time to wait during a silence before intervening may depend on several factors, including the student, the difficulty of the text, and the last action taken by the student and the tutor 10. 

Various mechanisms have been proposed to allow a tutor to take turns. Proposed solutions include dynamic constraint satisfaction, dialogue scripts, and linearly combined feature vectors. One proposed solution was to use prosodic rules to trigger backchanneling（相槌）in a computerized "eavesdropper（立ち聞き部）" that listened to conversations and interjected（挿入）"mm" when its rule fired（発火）."

"The tutor 10 uses seven turn-taking rules:
o Interrupt the user if the user's turn is very
long;
o Self-interrupt (stop speaking) if the user has 
overlapped;
o Backchannel（相槌を打つ）if the user pauses;"

"The forgoing rules do not fire immediately upon entering the appropriate turn-taking state. Instead, each rule has a delay associated with it that indicates how long the turn-taking state must remain in the appropriate state before the rule will fire. That delay is compared with the appropriate event timer(s). For example, the rule for backchanneling compares the backchanneling delay against the dead air timer and the user action timer. Because results from the education literature indicate that delays of more than three seconds between teacher questions and teacher-supplied answers lead to increased student learning, we set most of the timing parameters to be greater than three seconds. The exception was backchanneling because the tutor's backchanneling was not intended to be perceived by the student as taking a turn. The delay for backchanneling was originally set to 1.5 seconds, but we increased it to 2 seconds because it seemed too fast for this task. At 1.5 seconds, despite being longer than normal conversational pauses, the tutor 10 seemed to interrupt students who were struggling with difficult words. Increasing the delay to 2 seconds made the tutor 10 seem more patient." 

Backchannel, Wikipedia
"In linguistics, backchannels are listener responses in a primarily one-way communication. These can be both verbal and non-verbal in nature, and are frequently phatic expressions, primarily serving a social or meta-conversational purpose, rather than involving substantial two-way communication."

"In English, sounds like "uh-huh" and "hmm" serve this role."