Brother Long is dead, Brother White is injured, how can big data help both sides to kill

Text / Han Cheng

Brother Long is dead, Brother White is injured, everyone is talking very lively, Brother Dragon's death is very popular, Brother White is injured, don't be sentenced, it's an injustice, a contemporary hero, etc., jumped into the circle of friends. After reading the various texts, it can be roughly divided into systematic analysis, technical details analysis, social value analysis, professional case analysis, etc. In short, they are very exciting and positive. I did not want to use big data to solve this problem, so the author tried to use computer and data science to propose a solution to this problem.

Without further ado, go directly to dry goods:

The first step, murder big data collection

This step is naturally to collect as many murder data cases as possible in human history. To be sure, most of them are text descriptions. No matter how, we assume that 100,000 typical cases have been collected to form 10T text data.

Data source: case library, Internet crawler, as long as the keywords collected, various case library URLs at home and abroad, and other rules are believed to be easily completed by professional companies;

The second step, killing big data governance

The text data needs to be managed by NLP technology. The classical Chinese text should be modified into vernacular, and the English to Chinese translation should not be limited to details. There is a system to do it. Because natural language is the crystallization of human wisdom, natural language processing is also one of the most difficult problems in our understanding of cases. We have obtained 10 T text data and 100,000 cases. We need to use NLP to analyze these cases.

First of all, let's assume that we are mainly divided into 4 categories: 1) Bad guys kill good guys, direct killing succeeded 2) Bad guys kill good guys without success but were killed by good guys 3) Good guys kill bad guys, directly killed 4) Good guys killed The bad guys were killed by the bad guys. Here you need to do some searches for good and bad people, such as using public opinion to define whether the protagonist is a good or bad person. The machine first recognizes it and then divides it again according to the classification of people.

Then, for classified cases, we do two things to automatically segment words:

Good guy's name, characteristics, information marked on the network;

Bad guy characteristics, name, network label information;

Then we will form a broad list according to the classification, including the murderer, the victim, age, motivation, time, country, region, reason, occupation, social conference theory, social impact, social environment, specific description, etc .;

The third step is to kill big data algorithms and modeling

According to our previous table, we need to extract the labels of good people, for example, to be brave, to be loyal, just, to eliminate harm for the people, etc .;

Party tags: Mainly distinguish the attributes of the parties. For example, occupation information can be set such as ancient heroes, historical loyal officials, corporate executives, ordinary white-collar workers, newcomers in the workplace, full-time mothers, small businesses, and social well-known people. Character classification is difficult and needs Label the data based on the experience in the case.

Case label: mainly set the label according to the scene of the case, day, night, cause of the event, etc. This requires a lawyer to participate.

Behavior tags: killing with a knife, killing with a gun, killing with a body, pushing downstairs, using drugs, or burying alive, this should be able to be extracted from 100,000 cases according to word frequency, or it can be summarized based on the experience of lawyers;

Social background tags: social background, humanities, dynasties, cultural attributes, etc .;

Determine the model design of the application scenario. Suppose we build two models

1) Tell the murderer, the consequences of the murder?

2) Tell the slain how do you defend properly?

Algorithms that may be used here

Recommendation algorithm : Association Rule-based Recommendation (Association Rule-based Recommendation) is based on association rules, takes the previous cases as a reference, finds similar cases based on the current case label, makes recommendations for the recommended objects of this case, and mines association rules You can find the relevance of different cases in the implementation process, and provide behavior references for the parties in the case. It can be understood that there must be a connection between the behavior rules of the parties with the same attributes and the outcome of the judgment. Through this connection, the outcome of the judgment is predicted.

Clustering: mainly to solve the dynamic classification of criminal cases, the collection is divided into multiple classes of similar composition label party labels, case labels, social labels, label the behavior of this accord, " Like attracts like, people in groups ", where poly Class analysis can be called group analysis, which is a statistical analysis method for studying classification problems. Cluster analysis originated from taxonomy, but clustering is not equal to classification. The difference between clustering and classification is that the classification required by clustering is unknown. Cluster analysis content is very rich, including systematic clustering method, ordered sample clustering method, dynamic clustering method, fuzzy clustering method , graph theory clustering method, cluster prediction method, etc. It is estimated that after clustering, there are more than 10 thousand cases. In this way, we can deal with it more conveniently. Finally, we can perceive the case graphically. Provide the basis for decision-making for the people at that time;

You can also apply regression, decision tree :, support vector machine, deep learning, neural network and other algorithms, specifically the process of continuous improvement and feedback;

The fourth step is to kill big data analysis

There is a very important data source to be input at this time, that is, the murder situation at that time. Assuming that we take the mental input, I want to kill, willing, motivate, murder assumption, and I want to defend, why do I defend, etc .;

Scenario 1: After inputting the data, we will analyze the model according to the model 1) homicide consequences and give you the results immediately. What are the consequences? Go to jail, lose relatives, children, wife, life embarrassment of the parents for the rest of their lives, etc .;

Scenario 2: After entering the data, our legitimate defense analysis model 1) How to defend properly, the timing of proper defense, whether to communicate before defense, the risk of cases that may be encountered after defense, economic losses, and how long the sentence is.

The model is hypothetical, clear logic, and trial and error (this kind of thing does not know how to trial and error, the author has not thought about it);

The fifth step is to kill big data application products

Download the "anti-homicide" APP, manually enter the motive, time, method, tools, surrounding environment and other factors of the killing, and then the system will automatically tell you that you still don't want to kill TMD.

In this way, our big data products from data collection, to data governance, to data structuring, data labeling, tag combination algorithm models, application analysis, and then recommend the results to the APP side basically complete a killing big data solution;

Of course, big data killing big data solutions summarize the talents and tools needed:

Talent team: product managers, data analysts, technical engineers, architects, lawyers.

Tools: Big data acquisition platform, natural language analysis platform, big data label management system, big data modeling, data analysis and mining, data visualization, APP rapid development of customized systems, etc., because of real-time consideration, the system must be deployed in the cloud;

Of course, it is definitely not that simple. The data needs continuous training and feedback optimization. The problem solved by our big data killing system is to punish Yang Shan, use big data technology, use machine learning artificial intelligence, and assist killing and killing people. Make rational judgments; data and technology merely build a bridge between people and the world, and dig out and apply the value hidden in the case data. This not only conforms to the current big data thinking model, but also conforms to the future development direction of case judgment.

Finally, based on this case, we can easily find the following important construction ideas for big data in legal cases:

Opinion 1: The importance of case data. Extensive collection of various related case resources around the world, data resourceization, resource assetization is the key. We should form a resource database of legal cases.

Opinion 2: Case big data and data science will be deeply integrated, use various machine learning algorithms to understand the commonality of the case, refine the personality of the case, provide effective behavior measures for both parties, and the idea of association analysis proposed by big data The law provides new governance ideas;

Opinion 3: Big data will change the paradigm of legal case research. Law is not a science in a strict sense. The content of law design is many sociology, physics, chemistry, mathematics, psychology and many other factors are very similar to the scope of big data research. Now the big data colleges established in the country also have a multi-disciplinary communication system .

Opinion 4: National Circular 39 requires deepening the innovative application of big data in various industries, exploring new formats and models for synergistic development with traditional industries, and accelerating the improvement of the big data industry chain. Accelerate key technological breakthroughs in the fields of massive data collection, storage, cleaning, analysis and excavation, visualization, security and privacy protection. Promote the development of big data software and hardware products. Improve the big data industry public service support system and ecosystem, and strengthen the standard system and quality technology infrastructure. The opening and sharing of legal big data will inevitably bring about changes in the new legal research and judgment system;

I believe that one day, data will assist in adjudication of cases, and the results of the cases will be compelling, and the use of data to govern the case industry will be expected! And big data is challenging the traditional legal case research method and the legal research method.

Brother White goes well, big data analysis is late!

Zhang Hancheng (Welcome Wechat Consulting: waynezhanghc is welcome ) Research areas include: Introduction to Big Data Basics, Big Data Application Practices in Enterprises and Governments, Business Models for Data-Driven Business Transformation, Medical Big Data Operating System, Finance and Tax Big Data, Customs Big Data, Poverty alleviation big data, operator big data construction plan, tourism big data platform construction plan, data asset management, big data industry ecological analysis, data transaction market, blockchain, artificial intelligence and other new technologies to the value and empowerment of traditional enterprises .