Although Zhao Shiyan didn't feel that she was as exaggerated as Ling Zhi said, shining on the people around her like the sun. But she also didn't feel that Ling Zhi was lying or dealing with her. She could feel the sincerity in the boy's heart, so she was still very happy.
Who doesn't like others to praise themselves?
Lying on the table, she thought about it now. She described Ling Zhi as a "clean" boy. When she just said it, she blurted it out based on her feelings. But now she carefully recalled the first two encounters, she became more and more sure of herself.
The description is very accurate.
If we just talk about appearance, the boy is indeed not very handsome. He has single eyelids, small eyes, glasses, and even a little nose hair and an unshaved beard. At first glance, he looks like an ordinary student.
However, the serious look in his eyes when he played and talked about billiards, and the unhurried and unhurried movements when he took the shot made her feel that playing ball should be like this, rather than mixed with other things.
She only found it very funny when she occasionally praised her good shots and commented on her bad luck.
"What a fool. Is every missed shot a bad luck?"
Zhao Shiyan thought angrily.
Therefore, she is willing to be a golf partner with such a boy. After all, it is almost impossible to find a girl of the same level as herself in the golf room. But now, she feels that she is not only willing to be a golf partner with a boy, but even...
As the thought came together, she quickly and unconsciously shook her head. She also had her own troubles, but they were well concealed by her inherent aura.
After a while, she raised her head and patted her face.
"Well, we can't do this anymore, it's not like me at all."
The dormitory where Zhao Shiyan was staying was a single room. The senior went out for an internship during the summer vacation, and she should have moved to the new dormitory before the senior came back from the internship. She planned to use the campus card borrowed from the senior to take a shower, hoping
Can forget worries smoothly.
…
The next day, Ling Zhi came to the laboratory early. Today, Saturday, students from the laboratory will also come, but they will come later than usual. Ling Zhi did not open the paper as usual, but opened an order-taking software.
Since he learned to collect data for himself, Ling Zhi felt that this skill of his could not be buried, so he found a job for himself, sorting out some public data sets for some customers every weekend, which would not take up too much of his time.
.
Lexus hung online for a while, and today's first customer appeared.
"Hello, are you there?"
"My dear, what data do you want to collect?"
"Hello, can you help me analyze the data?"
Analyze data. Lexus does not specialize in data analysis, but whether it can be done depends on the situation. If you just do some simple statistics...
"You can send your request and I'll see if I can do it."
"Okay, I want to calculate the similarity between every two companies in the industry..."
Lexus listened to the customer's needs and asked some questions from time to time. After answering, it gave a quotation and expected delivery time based on the amount of data. The customer didn't bargain too much and the deal was finalized.
In fact, there are many ways to calculate text similarity, such as counting whether words appear at the same time in two texts, whether the frequency of occurrence is similar, whether the frequency of punctuation marks is consistent, etc.
However, these statistics are superficial statistics, and the deep semantic similarity of sentences cannot be calculated through this method.
For example, "happy" and "happy" have the same semantic meaning, with a similarity of 0.97, while "happy" and "sad" have opposite semantics, with a similarity of 0.02. It is impossible to draw the above conclusion using the most superficial statistical methods
.
Therefore, at present, deep learning methods have become the main method for learning the deep semantics of sentences.
Lexus doesn't want to fool customers, he wants to do his best since he wants to do it.
Although the customer is not professional in this area and does not know what methods are available to calculate similarity, nor has he asked Lexus how to calculate it, Lexus still wants to use deep learning methods to help customers accurately calculate the similarity of each company.
.
Lingzhi looked at the representative products of each company sent by customers, and found a word vector table that he had trained before. He first segmented the product names using a word segmentation tool, and then mapped them to the word vector table. Finally, he divided the product names between the two companies.
The vectors corresponding to the products between the two are calculated by pairwise Euclidean distance, and finally the similarity between each pair of companies is obtained by averaging...
It didn't feel complicated, but the day passed before I knew it. Lexus sorted out the data and sent it to the customer.
"Hello, the similarity calculation has been completed. Are you satisfied?"
After a while, the customer replied:
"[Hand over] The arrangement is too detailed, thank you. But how do you calculate the similarity? If it is convenient, can you tell me, I want to write it in my introductory tweet."
"Oh, it's actually not complicated. I use word vectors to calculate them, and word vectors are trained using deep learning methods..."
Ling Zhi carefully typed a large paragraph and briefly introduced the principle without going into too much detail.
In fact, the training process of word vectors is inferred by the deep neural network based on the context of the word. For example, the words that often appear around the word "happy" are completely different from the words "sad".
Therefore, deep semantics are essentially determined by its context, which is what Lexus roughly explains to customers.
As for how to derive word vectors based on context and how broad the scope of context is, Ling Zhi did not go into details. Customers did not care about it. He would only focus on the key points.
"Thank you. By the way, I actually still have a large amount of data that needs to be calculated for similarity. Can you teach me how to use the code? That way I won't have to trouble you in the future."
Teach code. Ling Zhi felt a little nervous. That’s not to say it’s not possible, but teaching people who are not in the computer industry to run code may cause various minor problems. But he didn’t refuse. This kind of demand is normal, and generally
After the customer asks for the data, if he wants code, Lexus will give it to him for free, and will also patiently teach the customer how to install various environments.
"Okay, I'll send you a document. You can install it according to the instructions in it."
Lexus simply revised the previously written document and sent it over.
after awhile.
"Sorry, I got an error when I installed the gensim package..."
"Oh, that may be a problem with the download source,..."
"Does this address need a space after the slash?..."
The customer kept asking various questions, Lexus answered patiently, and finally helped the customer solve the problem.
"Thank you, sorry for your trouble."
"You're welcome, that's right."
Ling Zhi looked at his watch. It was time to have dinner. Anyway, he had already added a customer on WeChat. Let’s talk about anything.
…
At 7 o'clock in the evening, Lingzhi picked up the "Alphabet Mystery" he had just bought and prepared to enter his own suspense world. After reading a few pages, his phone vibrated. Lingzhi turned on the phone and found that the customer had sent him another message:
"I'm sorry to disturb you again tonight. I think the table format I designed before is not very beautiful because I didn't think about it at the beginning. I have to trouble you to help me layout it again..."
Later, Lexus received the legend sent to it by the customer, indicating that it should be typed according to this format.
It wasn't too troublesome. Lexus turned on his computer and started working overtime, and he quickly redesigned it.
"Wow, you are so awesome. You can help me in time so late. Thank you so much."
"It's okay, it should be."
Ling Zhi thought for a while and added shamelessly:
"If you don't mind the trouble, you can leave me a positive review of more than 20 words."
“A must, very satisfied, must praise.”
"Yeah, thank you for your support [smiley face]"
It should be considered over, Ling Zhi smiled and shook his head. Generally, he would not take the initiative to ask for positive reviews, unless he felt that the customer was very satisfied. Think about when he developed this habit?
Ling Zhi didn't know, but he remembered something else.
When I visit other stores, sometimes I ask too many questions, and the customer service response is very slow, or simply responds with "It can't be done." You can't help but sulk in front of the screen. There are too many cases.
In the head.
Ling Zhi couldn't guarantee that his service would be 100% satisfactory, but he still wanted to do his best because he didn't want to become the kind of person he once hated.
Therefore, even if the customer doesn't even understand his own needs, as long as he can get the general direction, the rest is to help the customer patiently improve the details. As a result, he has gained some old customers who are willing to come to him for help frequently.
After a while, he received a comment from the customer: "A very good shopping experience. The code written by the staff was very professional, and there were detailed instructions on how to run the program. The whole service process was very patient, and they did their best even during non-working hours."
."
The customer feels that he has met a patient me, and I feel that he has met a polite customer. Ling Zhi thought.
…
The next day, Lexus opened the order-taking software and read the paper. Soon, business came to his door.
"Hello, I would like to ask if you have any data on high-end peripherals here?"
When Ling Zhi saw this sentence, although he didn't understand what peripherals meant, he felt slightly uncomfortable. But out of politeness, he still responded:
"Hello, we don't have this kind of data. Do you have a website with relevant data? If it is a public data set, I can help you collect it."
"Well, is it convenient to add WeChat? Some content is not convenient to say on the platform."
Ling Zhi's bad feeling was even worse, but he had no reason to refuse, so he scanned the customer's WeChat account.
"This kind of data is very scarce now. If you get it, you can sell it for a lot of money."
"Oh, what is a high-end peripheral? How can I find the data you need?"
"You can find out by searching on the Internet."
Lingzhi searched for it. Although he already had a premonition, when he saw the first result, he immediately verified his idea. He immediately replied to the customer and said:
"Hello, we are not allowed to collect this kind of data, sorry."
This customer seemed unwilling to give up, and promoted the preciousness of this kind of data endlessly and self-centeredly. In the end, it seemed that the other party was not interested, so there was no reply.
Ling Zhi breathed a long sigh of relief. The forest was really big and there were all kinds of birds. Things in the gray area must not be touched.