☔️ 👄 🔎 Developer's dream: manufacturing runs a business 🍚 🌃 🍲

For 15 years, we, the production of the Anti-Plagiarism company, have lived in simple and harmonious relations with business and PR. First, we rolled out new functionality to the prod. Then the PR began to PR: write news, press releases, give interviews, etc. Nothing, in fact, prevented such relations from developing further, until ... Although no, first things first.

At the end of 2019, we held a strategic session: voiced and recorded a large number of ideas that lay in the heads of developers, testers, analysts, devops, in general, of the very production. Having looked at our ideas, the management (and with it the PR) decided on a desperate step - to let the production itself tell us about our ideas, collect feedback, promote it to the market. This article is the first test step of simple production guys on an uncharted planet called the “market”.

And we also won a RVC grant for the development of artificial intelligence in the field of natural language text processing. Therefore, they felt obligated to outline the immediate plans.

The second quarter for us, on the one hand, is the hottest, since the load on the Anti-Plagiarism system is at its maximum during the session, and on the other, this is the beginning of a new production cycle. The main functionality is shipped, only bugs are corrected in it, and only new in our backlogs. It is possible to make profound changes to the kernel for the future and radically change the user experience. In addition, until the next session, we will conduct a lot of training webinars, and users will quickly get used to the new products ... In general, now you can create without looking back!

I present to you the services that we plan to bring to the market in 2020.

Anti-plagiarism. Screening

^{Image from pixabay.com/ru}

We held a strategic session in December 2019. Then no one could even think how much the topic of screenings, tests and other medical services would be. However, we also did not expect a pandemic, but thought that our prototype system based on text analysis would work primarily for our potential audience: for students who have deadlines for diplomas. And now it turns out that not only students are now in need of “text-based” checks ...

It is no secret that physical, emotional and mental conditions affect a person’s manner of communication. Written work is also a communication channel, therefore there is a very serious hypothesis, that according to the text written by a person, it is possible to determine what is happening with his health at the moment. A lot of documents pass through the Anti-Plagiarism system . We already know how to separate the text written by the author on his own from a pile of quotes or even plagiarism. The solution suggests itself: we need an algorithm that, according to the original text of a person, determines his emotional and physical state!

However, where to get the data for training? We decided to turn to specialists. Fortunately, among our clients there are medical universities and even special medical institutions. Using the connections, we went to several practicing departments, which happily helped us with the organization of data collection. Among the participants were patients in the therapeutic, surgical and psychiatric wards. We asked everyone to write a small text of 2000 characters, which included the following points: a description of their complaints and well-being before the doctor’s appointment, a description of the process of traveling from home to the queue to the office and a general description of the weather today. We tried several different models, the neural network algorithm showed the maximum result: accuracy 0.7 with a completeness of 0.8. Not bad, but it was felt that we did not all squeeze out of the data.They built an ensemble with another model that showed maximum on neural network errors. After the manipulations, the accuracy increased to 0.78, and the completeness to 0.82.

Let's get back to our prototype. Analysis of the original text of student work from last year’s session showed that only a small part of them show signs of panic, agitation, or stress. Unfortunately, most of the texts are given by business attitude and boredom - a clear sign of works written to order. In addition to the psychological state, suspicions of gastritis, several fractures and two suicidal moods were identified. We plan to include the phones of the nearest specialized medical care centers in the report on the results of the audit, so that the author thinks about his health and does not do stupid things.

This year we plan to develop this prototype to production. Due to the current pandemic situation, we are looking for contacts of people with COVID-19 to write test essays. After that, we will train the model on this data and hope that we will be able to get a very cheap, simple and mass test for identifying a variety of diseases from the authors ... There is a chance of getting a diagnosis of this terrible disease that the whole world is nightmare, at the very beginning of its incubation period. We are also open to work with mobile operators and special services to adapt this model to search for sick people by SMS and messages in instant messengers.

Antiyurist

^{Image from pixabay.com/ru}

The next service that we plan to release for production will allow us to look for weak spots in the texts of contracts without a lawyer! Imagine the near future, when you can download the contract into the automatic system and it will tell you all the weaknesses: the risks for each of the parties, the possibility of fraud without violating the contract and provide links to legal practice. Below is a screenshot of our prototype of such a system.

How does it work? The system is based on the technology by which IBM Watson is made. We have studied relevant articles on this subject and, using our ready-made developments, have created a prototype system that analyzes legal documents. To train this system, we used indexes of regulatory documentation databases, for example, ATP GARANT , and also indexed court decisions made publicly available on the websites of Russian courts. We added various contract templates from the Internet, and our legal partners made markup for training.

Technically, for a quick search, this solution uses not an ordinary single-index, but an index according to the graph of relationships between the parties to the agreement. The rights and obligations of the parties, fines, the procedure for resolving disputes - all these data turn into a graph of relationships ( one , two) Further, the graph is divided into subgraphs with partial intersection (the shingle mechanism also works approximately ), and already these same subgraphs fit into the index. This approach allows us to quickly find similar cases (the so-called candidates) and compare the document being checked more closely with a digital cast of what was in the index .

As with a regular report, the Anti-Plagiarism system only draws the user's attention to slippery moments in the contract. Now we are bringing the prototype to commercial operation and are developing an interactive report for convenient work with this type of document.

In the near future, add to the Anti-Lawyer a check of counterparties for participation in the courts, being in the bankruptcy register, in the lists of invalid passports (for individuals), restrictive legal proceedings, etc. Thus, the service will allow in one window to carry out a large number of checks and not fall into vain just one text document - the contract. And in a slightly more distant future, Antiyurist should get the functions of a financial risk assessment (yes, when our users want a figure, then we can’t refuse them :)) and the ability to automatically generate claims to the other side, and claims to courts of various levels. To support the latter function, it is planned to develop partnership programs with law firms.

PlagCoin and Anti-Plagiarism Depot

^{Image from pixabay.com/ru}

Well, what about without blockchain? We also have ideas on this. In the near future we plan to “mint” our PlagCoin crypto-coin. Checksums of deposited (i.e., publicly available) works can be woven into the blockchain . Such a service will be useful to those who want to protect the priority of intellectual property. So you can upload any document to our system, and its hash amountwill be woven into a chain of blocks - blockchain. In the event of a dispute, you can always show this document, and Anti-Plagiarism will confirm that a document with such a hash has been uploaded to our system at a specific time. The service also offers to fix the share of authorship and other categories of participation (financing, editing, data collection, etc.) in writing a particular work. Similar systems already exist and are developing. By the way, several of these companies are already clients of Anti-Plagiarism.

And with the help of PlagCoins, which ordinary users will earn, it will be possible to pay for the services of checking texts for borrowing, Screening, Antiyurist and others. We plan to build the mining process in such a way as to partially transfer the burden of performing search queries to the computers of the service users themselves from Anti-Plagiarism. Thus, the load on our equipment will be reduced and, therefore, more money and resources will be left for new research and various improvements in our main system!

PlagCoin will become one of the main services of our Total Honesty Ecosystem (EtoChe).

Plagiarism Sensitivity Regulator

Naturally, over the years of development, the system has a bunch of settings that can be twisted in one way or another. Examples of such pens are given here , here and here . It is clear that all these pens can be long and hard to improve, achieving the optimal result.

Figuratively, the whole process looks something like this, although the entire filling is, of course, hidden from the user:

^{We can’t provide a screenshot of the form with real parameters for obvious reasons, therefore only the layout from the original TK to assess the complexity and number of settings. Now the number of settings is much larger :(}

By the way, the optimization criterion is rather convoluted: it is to maintain a balance of technical support calls from students (“Bring everything back, we can’t pass our diplomas!”) And teachers (“I can smell plagiarism in your mouth and your system doesn’t see it”). You can use machine learning, deep learning, artificial intelligence and all that, but business cannot wait. Therefore, we decided to go the other way and reduce all the parameters into one:

Yes, one big button for basic life scenarios. As you can see, the service is intuitive, and everyone will now be able to get the scan report in the Anti-Plagiarism system that he needs in the current life situation. We have no doubt about the commercial success of the service: users have been asking for this modification since the creation of the system. As a bonus, such a service will allow us to hope for a decrease in technical support calls.

Antip Text Assistant

Voice assistants have already been released by everyone who could. Alice , Siri, Alexa, Maroussia and even Oleghelp make life easier for people around the world. We decided to keep up with the market, but we have a more interesting offer. It is regrettable, but a fact: modern youth are few, no, VERY LITTLE reads. Add here clip thinking, twitter, SMS and instant messengers. What do we get as a result? That it is very difficult for a modern person to independently formulate a good text for scientific work. Hence the fear of sitting down for writing a diploma or article, procrastination, putting it off at the last moment, which, as you know, leads to sad consequences. Okay, drunken parties, there are worse things: using someone else’s text, cheating, ordering work from professional writers. We finally understood what the main problem of modern students is, and we offer a timely solution!

Let me introduce: Antip text assistant from Antiplagiarism company. This text assistant decrypts your thoughts and turns your 150 characters into 2 normal pages of scientific text. The Anti-Plagiarism company remains true to its credo: as before, you need to create with your own mind. We only significantly facilitate the creative process. As well as lamplighters and telegraphists, soon writers of works to order will remain only in urban legends. Antip is cheaper, faster, better, because the test will be guaranteed unique. At the same time, he, as a personal butler with whom you have been together for a long time, will solve your scientific thought in a couple of sentences and replace your pale speech attempts with elegant and complex grammatically correct constructions.
The upcoming session will be devoted to beta testing of our assistant. Follow the news.

Instead of conclusions

For two weeks now all the employees of Antiplagiarism have been working remotely. The time that we used to spend on the road was freed up. As our time tracking system shows, part of this time suddenly went into work. The time for rallies was reduced, everyone writes the current status to the group’s channel in slack. Because of the current situation, there was even a certain sense of unity and unity. Judging by the reports from the diet, our productivity has improved significantly. The trust of company management is motivating and inspiring. I sincerely hope that we will leave the remote with new services that you just got acquainted with.

Everyone knows about what masterpieces appeared at Newton and Pushkin.after forced self-isolation, however, what should everyone else do? As screening of the Internet and a survey of familiar students showed, many seriously sat down to write their term papers, diplomas, dissertations, articles and novels laid on the table many years ago. The best times for creativity have come! Take the opportunity to explore the world for free (while many services provide free access to your resources) and create with your own mind!