If Fanyu wants to completely subvert the social field and become the terminator of Facebook and Twitter, it is not enough to rely on the perfect translation of the moment.

Need to go one step further and have more selling points!

At present, the interaction of social software is not only limited to traditional text, but also includes video, voice and so on.

More importantly, the integrated interaction with each other, such as the translation between speech and text.

Therefore, if the mortal forum or ordinary language wants to subvert the entire field, it cannot escape keeping pace with the times.

Video interaction is easy to solve, not much difficulty, but voice is different.

Although there are many intelligent voice input software at present, accuracy and fuzzy recognition have always been a big problem.

When the user speaks excitedly to the voice input method, the result is a lot of recognition errors, word drops, and even semantics are distorted.

As for those non-standard Mandarin and dialects, the results are even more miserable.

It's not an exaggeration to say it's chicken ribs.

What's more, Wang Fan is even more demanding!

He wanted to translate the phonetics of one language into the phonetics of another.

Perfect dialogue interaction between different languages, rather than simply limited to perfect interaction on text.

For example, conversational interaction between English and Chinese.

This requires the conversion of English pronunciation into English. Then it is perfectly translated into Huaxia, read it again, become Huaxia, and thus pass on to another person.

The same is true of antisense.

The whole process needs to be accurate, especially the first process, speech to text, once there is a mistake, the whole process will be problematic in the future.

And the perfect translation of the two scripts has been achieved now, and as for the final text-to-speech reading, it is not a problem.

It's just that the process of recognizing speech so accurately, this kind of approximate "intelligence", of course, Wang Fan can't do it with his own strength.

Even the world's leading giants are still limited to the imprecise stage of "fuzzy recognition".

But with a technology-derived system, everything will be different.

Wang Fan personally selected some speech recognition software, iFLYTEK, Qiandu, Lingxi, Cloud Voice, GoogleNow, Siri and Microsoft Cortana.

A comparison, each has its own merits.

Xunfei, Qiandu, Lingxi, etc., are more suitable for China, but they are not very effective in dealing with international aspects.

GoogleNow, Siri and Microsoft Cortana are good at international, and many Chinese aspects are not in place.

After some trade-offs, Wang Fan finally chose GoogleNow, a comprehensive and superior speech recognition software.

On this basis, Wang Fan derived a new speech recognition software - TT Zhiyu.

After some testing, Wang Fan was full of praise for the power of this speech recognition software.

It can not only intelligently recognize common languages such as Chinese, English, Russian, French, etc., but also slang.

Even if Wang Fan speaks foul language to TT Zhiyu, he can intelligently recognize or blur the translation.

The reason why the translation is a bit vague is because many dialects or foul language are too "high-end", and there are no specific Chinese characters corresponding to them at the moment, and even we can't write them.

However, Wang Fan soon discovered the problem, for Mandarin or nearly Mandarin, the accuracy of intelligent recognition is almost 100%.

However, for the recognition of dialect audio, it is somewhat unsatisfactory, and the accuracy rate is only about 90%.

Although this percentage is already high enough for users. But as a black technology produced by the system, it is still a little insufficient.

In addition, for the recognition of dialect audio, there will be two versions.

For example, when intelligently identifying "Our Tunli People", two results appeared.

One is the standard Mandarin version, and the other is a directly identified version: even Di Lao Ga, which is grouped in Ze Getun. The even line is a tunli, and the native sheep are raised......

And when connected to the Internet, the Mandarin version is in priority. In the disconnected state, the version is directly identified as the priority.

However, when Wang Fan chooses to directly identify the version, the result is that the direct identification version is in priority, regardless of whether it is connected to the Internet or not.

Later, Wang Fan experimented with some other dialect audio, and the same was true.

It's just that with his continuous attempts and choices, TT Zhiyu recognition has become more and more accurate, and the dialect accuracy rate has increased to 97%!

And the version selection is all directly identified version!

Wang Fan couldn't help but sigh that TT Zhiyu recognition not only has strong recognition ability, but also can continue to improve and progress.

By recording and analyzing the user's habits, the dialect's own dialect and speech library are passively expanded, so that the recognition results are more in line with the user's requirements, and the accuracy of dialect recognition is gradually improved.

For example, if Wang Fan selects the direct recognition version, other dialect audios will also give priority to the direct recognition version.

The recognition accuracy of dialects has also increased from 90% to 97%.

In order to verify his conjecture, Wang Fan found some dialect audio for TT Zhiyu to test.

If something is wrong or ambiguous, he will also correct it or make a choice of preference.

Over time,

97.5%，97.9%，98.2%，98.4%，98.5%……

The accuracy of dialect recognition continues to improve, but the magnitude of the improvement is obviously getting smaller and smaller.

In the end, two hours later, it still did not increase from 98.7 to 98.8.

Although the figure of 98.7% is quite satisfactory, the subsequent increase is really small and pitiful.

In desperation, Wang Fan could only give up temporarily.

After a short break, Wang Fan began to think, there is language intelligent recognition software, there is also perfect translation software, and only speech reading software remains.

This kind of software is very low-tech, and there is a large part of the application market.

Wang Fan looked for a few, but he was not satisfied.

It's either a cold machine language, or a limited number of celebrity voices, such as Teacher Guo, Sister Chiling, and robot cats......

So Wang Fan once again derived a simultaneous reading software.

This simultaneous sound means that when reading aloud, it perfectly simulates the previously recorded sound. Even the tone and speed of speech are simulated vividly.

For example, Andy Lau sang about the people in Tunli, and finally perfectly simulated Andy Lau's voice and read it out further.

In the end, Wang Fan combined TT intelligent speech recognition, perfect translation, and simultaneous reading into a new software - TT Intelligent Translation.

TT can perfectly translate the sounds of one language into the sounds of another.

In the whole process, the speaker's semantics, tone, speed of speech, and pronunciation are almost exactly the same.

What's more, the response speed is amazing, and the whole process is basically translated in seconds!

Wang Fan selected the Hebrew output, and then said to TT Zhiyi: Happy New Year, Happy Year of the Monkey!

TT Zhiyi immediately imitated Wang Fan's voice and tone, and replied in Hebrew: "The Year of the Monkey is auspicious."

This gave him a feeling of trance, as if he had become proficient in Hebrew all at once.

Wang Fan raised his head tiredly, the night was thick outside the window, and it was already late at night......

――――

PS: Happy New Year, Happy Year of the Monkey! (First Update!)

Thanks to Falling in Love with Satan 12, Big Miki, Penguin's Brother, GreenC, The Novel is Small, Account Password Table, House Prices Are Very Good, Benevolence, Jie Jie Together, Sugar Water Brother, Celestial Dragon of the North Pole, Universe F, Drifting Wind and others for their tips. As well as the recommended support of wisteria tree flowers, huixing, i love water, etc. Thank you.

During the Spring Festival, not only is it updated normally, but it is also added, and I haven't even watched the Spring Festival Gala, and I have been coding words, asking for a recommendation ticket for comfort......