Time Travel: 2014

However, as a specific executive, regarding the tasks proposed by Google:

——Evaluate the feasibility of implementing the algorithm proposed by LIN HUI, and consider whether it can be reproduced in a short time based on the actual situation

Eve Carly was speechless.

Perhaps in the minds of those whose butt determines their head.

Now that the technical route is clear, whether the technology can be reproduced is only a matter of time.

But the truth is not that simple.

Anyway, Eve Carly, who tried it all night, found it difficult to reproduce.

Let’s not talk about the algorithm technology itself proposed by LINHUI.

It’s the “LH text summary accuracy measurement model” that LIN HUI successfully grasped in the generative summary algorithm patent.

It would be extremely difficult for other teams to build the same model from scratch.

Speaking of which, the construction process of the LH text summary accuracy measurement model is very clear:

First, use language models to evaluate the fluency of algorithm-generated language;

Second, a similarity model is used to evaluate the semantic relatedness between the text and the abstract;

Third, in order to effectively evaluate the degree of recurrence of entities and proprietary words, the original text information model is introduced for evaluation.

…

However, it is just a simple thing to say.

When it comes to putting an elephant in the refrigerator, it’s also very simple and requires the same three steps:

——Open the refrigerator door, put the elephant in, and close the refrigerator door.

It’s useless to know how to do it, the key is to execute it.

It doesn't matter how clear the steps are if you can't follow them.

The construction process of the LH text summary accuracy measurement model has three steps.

The first step is complicated.

How to build a language model?

Follow the technical route proposed by LIN HUI.

The language model modeling process includes dictionary, corpus, model selection, etc.

The problem lies in corpora, a word that in linguistics means a large body of text.

This type of text is usually organized and formatted and marked up.

It is relatively easy to get information about the English corpus. After all, Eve's team has in-depth cooperation with the three universities of Oxford, Harvard and Yale in terms of linguistics.

But when it comes to Chinese and other text prediction information, it is completely difficult to say.

Make bricks without straw.

It’s useless to know the technical route without a corpus.

However, we can abandon the research on Chinese and other news generative summaries for the time being.

But this is almost equivalent to giving up a huge market.

And the most important thing is that the algorithm proposed by LIN HUI itself can take into account both Chinese news summaries and English news summaries.

So in the future, will LIN HUI directly develop a function to process Chinese news into English summaries?

There’s no reason why someone who can figure out a text summary processing algorithm can’t figure out a translation algorithm, right? ?

The more Eve Carly thought about it, the more likely she felt that this was possible.

Otherwise, why would a summary software have an interactive style similar to translation software?

While they were still hesitating.

The opponent has already made strides forward.

For a moment, Eve couldn't help but feel helpless after a fierce battle.

This is what is called falling behind step by step.

What else is there to evaluate in this situation?

I directly suggest that Google buy back the LIN HUI algorithm!

Although LIN HUI is Chinese.

But this technology is not an important technology related to the lifeblood of the country.

It’s not like we won’t sell it!

If you don’t want to sell it, you can also seek patent authorization!

It’s not like you have to follow others and reinvent the wheel.

Eve is not tortured or lustful.