From the product manager's point of view lucky Fast Three source code development, look at the phone side voice assistant

This article from the PM point of view of the mobile phone-side speech assistant, including the current market situation, PM in the design of the product ideas and so on.

From the product manager's perspective, look at the phone-side voice assistant

First, the status of mobile phone-side voice assistant

The advent of Apple Siri has led to the development of smartphone-side smart assistants, now, Apple Siri, Amazon Alexa, Google assitant, Microsoft Cortana, Samsung Bixby have been in the mobile phone-side layout of the voice assistant, Samsung Bixby also set up on the phone side of the Entity button Outbound helper, Mobile assistants are gradually moving to mobile development.

In the future, in the smart hardware "interconnection" trend, smart speakers, smart watches, smart headphones and other products will cover most of the needs of people's lives, mobile phone in the intelligent hardware and the impact of the app, should look for features of the landing scene, combined with mobile, screen, information aggregation characteristics, to visual, tactile and other multi-modal development, Avoid homogenization.

Second, PM need to pay attention to which issues

In this context, PM needs to think about several issues: Scene selection, user experience, traffic entry, user stickiness.

From the product manager's perspective, look at the phone-side voice assistant

1. Scene design

Scene design is an important part of product design, user demand, product realization, bot creation, semantic comprehension difficulty are inseparable with the choice of scene. The following three points need to be focused on digging a scene:

(1) Clear objectives

(2) Limited input

Intelligent assistant mainly through dialogue to complete the interactive lucky Fast Three source code development QQ2952777280 "Words Fairy Source Forum" hxforum.com "Papaya Source Forum" Papayabbs.com, the user each sentence contains the amount of information to be maintained within a certain range otherwise the user does not know what to express, the machine can not be well understood. For example:

Bad case:

Q: I'd like to order a meal

A: OK, what would you like to eat?

Q:emmm ... I want to eat egg tomato fried rice, put a little sugar I'm afraid salty, thank you

A: Sorry, did not understand you said, can you tell me again?

Q: I would like eggs fried tomato rice bowl, less sugar

......

The above case, the user input domain is too open, the user has been doing "simple problem" rather than "choice question", increase user costs, but also not conducive to natural language understanding.

The same scenario can be optimized to:

Q: I'd like to order a meal

A: OK, would you like to have rice bowls, noodles or hamburgers today? (Graphical interface available here)

Q: Rice Bowls

A: Well, we recommend some of the highest-selling rice bowls: egg and tomato bowls, braised beef risotto, more

Q: Egg and tomato rice bowl

A: OK, this is our sign, if you have the following special needs, please choose, less sugar, less salt, if not do not have to reply

Q: Less sugar

.......

(3) Fast convergence of conversations into task instructions

When the bot recognizes the user intent A, it will not complete the corresponding task instruction. Therefore, it is necessary to quickly refine the intention and slot in the scene.

From the product manager's perspective, look at the phone-side voice assistant

We re-quoted the case of the order.

Create a booking bot according to case:

Intention Intent: Order meal

Training dialog Sample hit intent: query= I'd like to order a meal.

Slot 1: Dish name dictionary contains: Egg tomato bowl, braised beef rice, etc. (corresponding to the database)

Slot 2: Special Needs

From the product manager's perspective, look at the phone-side voice assistant

In order to set meal scene, through the parameter extraction in the conversation, quickly converge to the order meal intention, finally satisfies the user demand. Voice assistant before Nlu, there will also be a process of speech recognition, you can refer to the two previously shared articles

2. User Experience

User experience is all PM need to pay attention to the issue, we often say that listening, understanding, satisfaction is to enhance the user experience of the point. So in the user experience, you should focus on the following three points:

Can you solve user needs?
What are the minimum standards accepted by the user?
What is the standard that exceeds the user's expectations?
We use the case description of the weather.

Address user needs: Phone Assistant Check the weather
User accepted minimum criteria: Query a time, location for the weather
Exceed user expectations: Severe weather alerts, rain and snow weather in advance to ask the user whether the car, etc.
The principle is similar to the Kano model in demand analysis, that is, the basic demand, the expected demand, the excitement demand, can refer

From the product manager's perspective, look at the phone-side voice assistant

3. Flow entrance

Individuals think that mobile phone voice assistants and smart hardware traffic entry, smart hardware must find enough just needed for the scene to be sustainable development. Mobile phone has a congenital flow advantage, to let the voice really landed, become the user accepted the most commonly used interactive way, the final thing to solve is the user head needs.

So what are the head needs that voice assistants need to focus on? I think, check the weather, check the time is not the head needs, these are the test/get Started voice assistant the simplest function. The basic needs of the user are the daily necessities and communication. For example, smart home, control home equipment is the head of the demand, in the car environment, telephone, texting is the head demand.

In the food and shelter scene, respectively, corresponding to different needs: voice shopping, ordering meals, restaurant reservations, hotel reservations, booking travel tickets and attractions tickets, are the public demand for life services, which makes it easier for users to know what the Voice assistant can do for them.

4. User stickiness

This is a data report on the AI assistant, which shows that Apple Siri's user volume is the first, but the monthly activity has been declining. This leads to: in addition to the flow entrance, PM should also focus on user stickiness.

From the product manager's point of view lucky Fast Three source code development, look at the phone side voice assistant