What kind of product is Iflytek Input Method?

Iflytek input method is an input software that integrates various input methods such as voice, handwriting and pinyin. In view of its greatest feature is speech recognition input, so today we focus on the speech input method.

Although I am Fan Luo, frankly speaking, Iflytek has been used as an input method for a long time before the hammer conference, and speech recognition technology is nothing new, but the advertisements of Iflytek and Lao Luo have once again attracted public attention.

In fact, the Chinese speech recognition technology of Iflytek input method is indisputable and very awesome (if you don't accept it, you can download and test it yourself, and I'm tired of it, from English to Chinese to mainstream dialects). On the 23rd of this month, Iflytek held an annual press conference at the Beijing National Convention Center, and the video of the new product "Dubbing Hall" immediately became popular in Weibo. Iflytek's technology has been able to simulate human voice dubbing through technical support.

1.？ Why are fewer and fewer people using Iflytek input method?

However, many people around us use Baidu, sogou and QQ input methods, but few people use Iflytek input methods, which is why Iflytek has been tepid. There are two main reasons:

First, apart from the voice input function, other functions in Iflytek are difficult to win compared with the mainstream input method. After trying a variety of input methods, the interviewee finally chose sogou because of its powerful thesaurus and cloud association function.

Secondly, it is also an important question to be discussed, why the best voice input function in Iflytek cannot be popularized (at least temporarily).

2.？ Why is Iflytek's voice input function awesome, but it is still difficult to impress users?

(1) Pay attention to whether everyone, including ourselves, usually type details like this: thinking and modifying, which is what we often do when facing the keyboard, but once we want to modify the voice input, it will be very troublesome and we need to keep moving the cursor.

The answer is that it is common for online novelists to watch ten thousand classes every day. My husband often says that you can use voice input if you don't feel strange, but in fact, even if I am alone at home, I won't choose, because the process of codeword is actually a process of repeated thinking and revision.

(2) We usually need punctuation marks when inputting text, even if it is a bunch of meaningless spaces, commas or ellipsis. If punctuation can't be inserted accurately, the experience of converting speech into text is very poor.

(3) It is true that at the hammer conference, we saw the accuracy rate of 97% mentioned by Lao Luo, but it is actually a process of trial and error-what if you have an accent? What if your pronunciation is incorrect? Even if the surrounding environment is chaotic? At this time, the recognition degree of voice input will become very low.

(4) When it comes to voice input, siri is actually the first thing most people come into contact with. Do you remember the feeling of talking to her? Is it weird to chat with AI in sci-fi movies?

To tell the truth, when answering the voice input method, they always find a corner where no one is at home. This experience is different from the voice chat of WeChat, because at least that one is sent directly by you, which looks like communication with people, while voice-to-text is more like interaction with machines. Although we have to admit that Iflytek's recognition of dialects, English and Chinese is really high, pronunciation is only an input method, and the ultimate goal is to communicate with people.

(5) Now even QQ has the function of converting speech to text, but in public places, if we talk about things, we often want to have a private space instead of transmitting everything by voice.

In addition, do you remember when WeChat voice became the most taboo thing among colleagues? It's convenient for you to talk quickly, but someone else is in trouble, okay.

In fact, having said so much doesn't mean that Iflytek's voice input method is meaningless, but there is still a lot of room for optimization in the application of scene logic.

I hope Iflytek or other voice input products can solve this strange information input mode as soon as possible, so that voice can really play a role. After all, "a gentleman talks but doesn't do it" (laughs ~)

= = = = = = = = = dividing line = = = = = = = = = = = = = = = = = =

Dear, if you like it, or find it helpful, just like it before you go (* ~ 3) (ε ~ *).