Voice assistant - future trends

         In this chapter, we will discuss the future development trend and planning of voice assistant , What should the future voice assistant look like ？ Here are some of my personal thoughts .

1、 The trend of the scene ：

         Voice assistant because of its use characteristics , Need to talk face to face with the machine , This is actually difficult for some people to accept , Especially in crowded places , People will feel embarrassed , So we seldom see people using these voice assistants in some public areas , Such as the bank lobby 、 Shopping malls, etc , Of course , This is also due to the current technical limitations of voice assistant , The effect has not reached the same effect as the real person communication . Relative , In some scenes with strong privacy , The advantages of voice assistant will become obvious , Such as the home scene 、 Car scene 、 Telephone assistant 、 Office scenes, etc , In these scenarios , Because of its convenience and quickness , People are more willing to communicate with machines by voice , For example, when cooking at home, you can use the voice assistant to check the menu , You can also use a voice assistant to play with children , Play some educational games , Or play music in the car scene , Navigation etc. . Predictably, , Mining user usage scenarios in these scenarios , And make targeted technological breakthroughs and innovations , It will be a trend in the future .

2、 User trends ：

         Different voice assistants vary according to their usage scenarios , The target user groups are also different , That is to say, the product positioning is different , For example, car assistants are mainly aimed at car owners , For the voice assistant on the mobile phone , Its main user groups are K12 user , That is from 1 Students from grade to high school , It is mainly because these students have learning needs （ Check information ）、 The need for entertainment （ Chat ）, And the desire to explore 、 feeling of freshness , They will also use more skills . And for adults , The purpose of its use will be more clear , For example, check the data 、 Operating equipment, etc , therefore , To mine more users , You need to start from the specific application scenarios , Consider what users really need in a certain scenario , To get more core users and conversion rates . With the mining and filling of application scenarios , The user community will also cover more .

3、 Technology trends ：

         The trend of technology is mainly divided into several points ： Topic conversation 、 Self learning 、 End model .

         Topic conversation ： In the current voice assistant , The main form of dialogue is still single round dialogue , That is, the form of one question and one answer , Ask and answer , Go as soon as you use it , It is difficult for such an assistant to create a sense of intimacy , Users don't feel like communicating with others , So the voice assistant in the future should have better ability of multi round dialogue , You can chat around a topic , Or take the initiative to start the topic , From task assistant to topic assistant . meanwhile , Talking too much doesn't mean being inefficient , Effective communication is better than redundant operation , By communicating with users , In depth understanding of user needs , Make reasonable guidance and recommendation , It can help users more efficiently .

         Self learning ： Different from ordinary machines , The voice assistant should have the ability of self-learning , That is, dialogue and operation according to history , And the surrounding environment , Learn to improve your ability , It seems inconceivable , But it can also be realized after scene disassembly , such as ： Through active learning 、 Pseudo label 、 Learning with noise 、 Reinforcement learning 、 Distillation, etc , Can realize the update iteration of the model , In fact, this has realized the purpose of self-learning to a certain extent .

         End model ： For privacy and efficiency , Now the models are more and more popular , When the user's network signal is bad , Or when some private information is not easy to upload to the cloud , The end-to-end model can give full play to , So offline learning and reasoning , It will also be a major trend in the future .