current position:Home>Nature: machines promote the production of human language

Nature: machines promote the production of human language

2021-08-26 09:02:40 Xinzhiyuan

New Zhiyuan Report

source :Nature

edit :Su Isabel

【 Introduction to new wisdom 】 Do robots promote human language ? In the near future Nature A magazine article tells you : Yes !

This week, ,《 natural 》 There was a scientific report in the newspaper , The researchers found that , Robots can promote the production of human language .

The significance of this report lies in , For language retardation 、 Clinical population with difficult vocabulary Retrieval 、 And the educational environment that needs to enhance language generation , Robots may play a very important role .

and , The common representation and language promotion mechanism revealed in this paper may be integrated into the cognitive architecture of social robots , To improve their communication skills in oral communication with humans .

before , Studies have shown that , At the information processing level , Robots can affect human behavior , Including sports 、 social contact 、 Higher order cognitive aspects .

human beings + robot : The naming task is completed quickly !

According to the article , The participants in this experiment are 36 Name age 19 to 35 Native German speakers between the ages of , And robots Pepper, This is Japan Softbank group and France Aldebaran Robotics A humanoid robot designed to communicate with humans , high 1.2 rice .

The robot's voice and behavior through NAOqi Frame generation , It contains stimuli in robots and visual displays Presentation Custom scripts for communication between software .

In order to highlight the natural behavior of robots , During the break , It is set to respond and pay attention to its environment . During the naming task , There will be subtle arm and head movements .

Before the experiment ,Pepper Will be introduced to participants as task partners .

Experimental materials include 320 A photograph of a natural or man-made object , It is divided into 32 There are different semantic categories ( Such as , flowers 、 building 、 birds ).

The cognitive model in the experimental device is shown in Figure A Shown , The vocabulary access cognitive model in speech generation explains how concepts activate entries in the mental vocabulary , Then it is converted into voice .

In the experiments , A participant and a robot Pepper Sit next to each other in front of the computer screen , display picture , Some of them are semantically related .

The researchers evaluated whether the robot's speech behavior was simulated , And at which processing level ( Concepts and vocabulary ).

In order to eliminate the interference of additional language input that may occur in the process of sharing naming with robots , Participants and robots wear headphones , This can mask the partner's naming voice .

The picture is named by the participants , Half is presented only visually . Under joint naming conditions , Half of the pictures in a semantic category ( Like a pet ) Named by participants , The other half is named by the robot partner . For those pictures named by the participants , Will calculate 1-5 The sequential position of .

Each experiment is based on 0.5 The second fixed cross show begins . In the experiment in which participants named objects , The picture is rendered until the named response is started or at most 2 second . Start with the picture display , The naming of participants is delayed ( reaction time ) It will be recorded .

The pictures are continuously displayed on the computer screen . Under single naming conditions , In a given semantic category , Half the picture ( for example : Fruits ) Named by participants , The other half of the pictures are presented only visually . Fruits ) Half of the participants named , The other half is presented only visually . Under joint naming conditions , In a particular semantic category , Half the picture ( Like a pet ) Named by participants , The other half is named after the robot .

Experimental results show that :

Co naming with robots can promote the generation of language .

There is no promoting effect when humans and human companions name together , And with robots , It can promote .

Robots do not collectively represent at the level of lexical choice , The language of the machine is simulated at the conceptual level , Thus, it has promoted the production of human language .

At the end of the experiment , Participants completed two questions related to robot intention : Robots act deliberately 、 Robots actively decide when to respond to stimuli , Evaluation criteria from 1 To 11.

Of course , This study is not perfect , such as , The exact nature of this common characterization is unclear . Future research should include physiological measurements , Like electroencephalogram , Or a targeted underlying mechanism .

in addition , The behavior of robots should also be systematically studied 、( synthesis ) Details of sound or visual appearance , And whether the strength of language interaction or the enhancement of robot attributes will affect the common language representation of robots .

copyright notice
author[Xinzhiyuan],Please bring the original link to reprint, thank you.

Random recommended