Neurolinguistic programming (NLP) - the key to people's souls or blind manipulation?

In psychology, there are special methods that allow you to influence the psyche of an individual for personal gain. With their help, you can win over potential partners during important negotiations, as well as attract the attention of the right clients. These psychological methods of influence are called neurolinguistic programming (NLP). They can be used for promotions, solving various problems, and establishing trusting relationships with others.

What is NLP?

This is a wide range of tasks for processing texts in natural language (i.e., the language that people speak and write). There is a set of classic NLP problems, the solution of which has practical benefits.

• The first and most historically important task is machine translation. They have been doing it for a very long time, and there is enormous progress. But the problem of obtaining fully automatic translation of high quality (FAHQMT) remains unsolved. This is in a sense the engine of NLP, one of the biggest tasks you can do.
• The second task is text classification. Given a set of texts, the task is to classify these texts into categories. Which one? This is a question for the corps.
  The first and one of the most important application methods from a practical point of view is the classification of letters into spam and boorish (not spam).

  Another classic option is a multi-class classification of news into categories (categorization) - foreign policy, sports, big top, etc. Or, let’s say, you receive letters and you want to separate orders from an online store from air tickets and hotel reservations.

  The third classic application of the text classification problem is sentiment analysis. For example, classifying reviews into positive, negative and neutral.

  Read also:  Why women cheat: reasons for female infidelity and 9 signs that a woman has cheated

  Since there are so many possible categories into which texts can be divided, text classification is one of the most popular practical NLP tasks.

• The third task is named entity extraction, NER. We highlight areas in the text that correspond to a pre-selected set of entities, for example, we need to find all locations, persons and organizations in the text. In the text “Ostap Bender is the director of the office “Horns and Hooves”” you should understand that Ostap Bender is a person, and “Horns and Hooves” is an organization. Why this task is needed in practice and how to solve it, we will talk in the second part of our article.
• The third task is related to the fourth - the task of extracting facts and relationships (relation extraction). For example, there is a work relationship (Occupation). From the text “Ostap Bender is the director of the office “Horns and Hooves”” it is clear that our hero has a professional relationship with “Horns and Hooves”. The same can be said in many other ways: “The “Horns and Hooves” office is headed by Ostap Bender,” or “Ostap Bender has gone from a simple son of Lieutenant Schmidt to the head of the “Horns and Hooves” office.” These sentences differ not only in predicate, but also in structure.
  Examples of other frequently identified relationships are purchase/sale relationships (Purchase and Sale), ownership (Ownership), the fact of birth with attributes - date, place, etc. (Birth) and some others.

  The task seems to have no obvious practical application, but, nevertheless, it is used in structuring unstructured information. In addition, this is important in question-answering and dialogue systems, in search engines - always when you need to analyze a question and understand what type it is, as well as what restrictions there are on the answer.

• The next two tasks are perhaps the most hyped.
  These are question-and-answer and dialogue systems (chatbots). Amazon Alexa and Alice are classic examples of conversational systems. For them to work properly, many NLP problems must be solved. For example, text classification helps determine whether we fall into one of the goal-oriented chatbot scenarios. Let’s say “the question about exchange rates.” Relation extraction is needed to determine the placeholders for the scenario template, and the task of conducting a dialogue on general topics (“chatter”) will help us in a situation where we do not fall into any of the scenarios. Question and answer systems are also an understandable and useful thing. You ask the machine a question, the machine looks for the answer to it in a database or text corpus. Examples of such systems are IBM Watson or Wolfram Alpha.
• Another example of a classic NLP task is summarization. The formulation of the problem is simple - the system takes a large text as input, and the output is a smaller text, somehow reflecting the content of the large one. For example, the machine is required to generate a retelling of the text, its title or annotation.
• Another popular task is argumentation mining, searching for justification in the text. You are given a fact and a text, you need to find a justification for this fact in the text.

This is certainly not the entire list of NLP tasks. There are dozens of them. By and large, everything that can be done with text in natural language can be classified as NLP tasks, it’s just that the topics listed are familiar, and they have the most obvious practical applications.

What NLP methods exist in psychology?

At first glance, NLP seems difficult to put into practice. However, as experience shows, anyone can master the methods and secret techniques of NLP. You can master the theory and practice of this discipline on your own using books and articles from the Internet or with the help of trainings. Neurolinguistic programming classes are taught by experienced specialists. For example, psychologist-hypnologist Nikita Valerievich Baturin. Also, for those who, along with NLP, are still interested in hypnosis, it is recommended to take the online course “Training in modern hypnosis.”

What NLP methods and techniques exist:

1. Change of submodalities. The technique allows you to change your attitude towards some event or phenomenon. With its help, you can take a fresh look at past events and change your feelings about them. This technique works like this: you take a situation to which you need to change your attitude, and a situation in which the individual experiences only positive emotions. Differences are found between both cases, and then the submodality of the first case is replaced by the submodality of the latter.
2. Setting software goals. This abbreviation reflects the criteria that a person’s intended goal must meet. When describing a task using this technique, the individual must clearly understand what he wants. Determine the benefits, weigh all your options, identify your motives. Calculate the time it will take to achieve a specific task.
3. Technique based on Walt Disney's experience. Any question must be looked at from three points of view: the dreamer, as well as the realist and critic. At the very beginning of any activity you need an idea. The dreamer takes on this role. A realist thinks about how to bring his idea to life and moves on to action. The critic looks for weaknesses in the project and pays attention to all possible benefits.
4. Logical levels. If a person sets out to change his life, such a task must be consistent with his beliefs and moral values. There are higher and lower levels of human perception of reality. There is a relationship between them. Having realized a problem or goal at a lower level, you need to work with it both at the same and at a higher level.
5. Manipulation techniques. If a person wants to influence other people, he can use a method such as the three “yes” technique. This technique is based on the inertia of the individual’s psyche. A person, by inertia, will answer in the affirmative if he is asked several secondary questions before the main question. Moreover, each of them must assume an affirmative answer. You can manipulate people's minds using trap words. For example, you can ask: “After this drink, do you become more cheerful and beautiful?” Any person will answer this question in the affirmative. Another technique is based on asking a person to do something not in the form of an order, but in the form of a question. People are more likely to agree to someone's request when they are asked for their opinion. For example: “Do you think the music is playing too loud? Can it be made quieter?
6. Swing. This technique helps replace destructive phenomena with positive ones. Using this method, you can get rid of bad habits and correct problematic behavior. First, the situation that requires adjustment is determined. Then the factors that cause a person to act in this way are identified. After this, they make a “swing”, that is, they change the negative image to a desirable one.
7. Generating new behavior. This technique helps an individual get rid of many problems and gain confidence in their own abilities. First, a situation is identified that does not suit the person. It is worked out in every detail. After this, a new interpretation of an already known situation is created. If a person, after examining it in detail, experiences positive emotions, then the goal has been achieved. The individual’s reaction to the manifestation of reality has changed for the better.
8. Six-step reframing. Sometimes an individual cannot get out of a problematic situation for a long time. His own thinking interferes with him. It seems to the person that things will get even worse. The essence of this method is this: with the help of meditation, an individual communicates with his own subconscious and asks it whether there are benefits in the current situation. If they are not there, then how to get rid of the problem. A person needs to carefully analyze the information received. Subsequently, he will be able to move on to real action and change his life.
9. Reimprinting. The technique allows you to find resources for changing established negative beliefs and updating incorrect behavioral patterns. A situation that causes strong feelings is reviewed, analyzed from several points in time, and benefits and positive intentions are found from reactions to it. All information received is analyzed. Based on them, the behavior model or thinking of the individual changes.
10. Reassessment of the past. The technique allows you to change your attitude towards an unpleasant event. With its help, attitudes towards people also change. For development, a period is determined that requires analysis and change. It is necessary to create a kinesthetic anchor, that is, remember some pleasant incident from life and fix this moment with a movement (snap your fingers). During the selected period, you need to remember positive and negative situations. Re-experience the positive moments of life, just watch the negative ones from the sidelines. When considering negative situations, it is important to remember your resource anchor. It will help you see the positive aspects even in negative moments.

Why is solving NLP problems difficult?

The formulations of the problems are not very complex, but the problems themselves are not simple at all, because we are working with natural language. The phenomena of polysemy (polysemous words have a common original meaning) and homonymy (words with different meanings are pronounced and written the same) are characteristic of any natural language. And if a Russian speaker understands well that a warm welcome
has little in common with
a combat technique
, on the one hand, and
warm beer
, on the other, the automatic system has to learn this for a long time.
Why is it better to translate “ Press space bar to continue
” into the boring “
To continue, press the
” than “
The space press bar will continue to work
.”

• Polysemy: stop (process or building), table (organization or object), woodpecker (bird or person).
• Homonymy: key, bow, lock, stove.
• Another classic example of the complexity of language is pronominal anaphora. For example, let us be given the text “ The janitor shoveled snow for two hours, he was unhappy
  .” The pronoun "he" can refer to either the janitor or the snow. From the context, we easily understand that he is a janitor, not snow. But getting a computer to easily understand this is not easy. The problem of pronominal anaphora has not yet been solved very well; active attempts to improve the quality of solutions continue.
• Another additional complication is ellipsis. For example, “ Petya ate a green apple, and Masha ate a red one
  .” We understand that Masha ate a red apple. However, getting a machine to understand this too is not easy. Now the problem of ellipsis restoration is solved on tiny corpora (several hundred sentences), and on them the quality of complete restoration is frankly weak (about 0.5). It is clear that such quality is no good for practical applications.

By the way, this year at the “Dialogue” conference there will be tracks on both anaphora and gapping (a type of ellipsis) for the Russian language. For both tasks, corpora with a volume several times larger than the volume of currently existing corpora were collected (moreover, for gapping, the volume of the corpus is an order of magnitude greater than the volume of corpora not only for Russian, but for all languages ​​in general). If you want to take part in competitions on these buildings, click here (with registration, but without SMS).

Neurolinguistic programming - how to learn?

Neurolinguistic programming is a training in demand today, so finding a course that suits you is not difficult; most often such courses are organized in universities with a specialization in Psychology. Some psychological centers teach NLP remotely. To learn the method, it is important to practice constantly. There are a lot of open access webinars where NLP techniques are discussed, which you can learn on your own by practicing at home with family and friends. Reading professional literature and mastering exercises contribute to the skill of an NLPer.

How NLP problems are solved

Unlike image processing, you can still find articles on NLP that describe solutions that use not neural networks, but classical algorithms like SVM or Xgboost, and show results that are not too much inferior to state-of-the-art solutions.
However, several years ago neural networks began to defeat classical models. It is important to note that for most problems, solutions based on classical methods were unique, usually not similar to solutions to other problems, both in architecture and in the way features are collected and processed.

However, neural network architectures are much more general. The architecture of the network itself is most likely also different, but much less; the trend is towards complete universalization. However, what features we work with and exactly how we work are already almost the same for most NLP tasks. Only the last layers of neural networks differ. Thus, we can consider that a single NLP pipeline has been formed. We will now tell you in more detail about how it works.

How it works

A person cannot perceive the world around him objectively, since he passes it through his own sensations (visual, auditory, olfactory), through the prism of acquired experience, personal beliefs and principles. Some people cannot live without church, others openly hate religion. Some are looking for a job that pays more, while for others it is more important that they like it. Some people like external beauty in people, others are delighted with the mental abilities of the interlocutor.

According to NLP, if you take into account all the above points and understand how a particular person passes information through himself and what is of particular importance to him, using certain techniques, you can achieve anything from him. But first, painstaking work must be done with beliefs and physiological characteristics.

If a manager persistently persuades a church-going grandmother to buy goods with demonic symbols, he will fail. It is necessary to offer an alternative option in time so that the purchase turns out to be pleasant for her and brings benefits to the company. This is where NLP techniques such as emotionally meaningful words and creating positive expectations come into play.

If an employee sits in a closed position (arms and legs crossed), it is difficult to demand increased productivity. But, having mastered the mirroring method, you can open the pose and talk more naturally.

Pipeline NLP

This is a way of working with features, which is more or less the same for all tasks.
When it comes to language, the basic unit we work with is the word. Or more formally "token". We use this term because it is not very clear what 2128506 is - is it a word or not? The answer is not obvious. A token is usually separated from other tokens by spaces or punctuation. And as you can understand from the complexities we described above, the context of each token is very important. There are different approaches, but in 95% of cases the context that is considered when the model is running is a sentence that includes the original token.

Many problems are generally solved at the proposal level. For example, machine translation. Most often, we simply translate one sentence and do not use the broader context at all. There are tasks where this is not the case, for example, dialog systems. Here it is important to remember what the system was asked about before so that it can answer the questions. However, the proposal is also the basic unit with which we work.

Therefore, the first two steps of the pipeline, which are performed to solve almost any problem, are segmentation (dividing text into sentences) and tokenization (dividing sentences into tokens, that is, individual words). This is done using simple algorithms.

Next you need to calculate the attributes of each token. As a rule, this happens in two stages. The first is to calculate the context-independent features of the token. This is a set of features that do not depend in any way on other words surrounding our token. Common context-independent features are:

• embeddings
• symbolic signs
• additional features specific to a specific task or language

We will talk about embeddings and symbolic features in detail later (about symbolic features - not today, but in the second part of our article), but for now let’s give possible examples of additional features.
One of the most frequently used features is part of speech or POS tag (part of speech). Such features can be important for solving many problems, for example, syntactic parsing problems. For languages ​​with complex morphology, such as the Russian language, morphological features are also important: for example, what case is the noun in, what gender is the adjective. From this we can draw different conclusions about sentence structure. Also, morphology is needed for lemmatization (reducing words to initial forms), with the help of which we can reduce the dimension of the feature space, and therefore morphological analysis is actively used for most NLP tasks.

When we solve a problem where the interaction between different objects is important (for example, in a relation extraction task or when creating a question-answer system), we need to know a lot about the structure of the sentence. This requires parsing. At school, everyone did parsing of a sentence into subject, predicate, object, etc. Syntactic parsing is something like that, but more complicated.

Another example of an additional feature is the position of a token in the text. We can know a priori that some entity occurs more often at the beginning of the text or vice versa at the end.

All together – embeddings, symbolic and additional features – form a vector of token features that does not depend on the context.

In what cases is NLP used in psychology?

NLP is used in various areas of life. For example, in psychotherapy, interpersonal communication, the art of sales. It is advisable to use it in personnel management, time management, journalism, acting, and jurisprudence. The techniques of this discipline allow you to behave correctly in any situation, find a common language with strangers and influence their way of thinking. Psychological methods help to get rid of various phobias, normalize mental state, and maintain mental balance even in crisis situations.

In what areas is NLP used?

1. In negotiations. Knowledge of this technique allows you to better understand the client, correctly build a line of conversation with him, manipulate his consciousness, insist on your own, and achieve success in any discussion.
2. In sales. All training aimed at working with clients in the field of sales uses NLP techniques. With their help, you can “hook” the buyer and sell him anything.
3. In psychotherapy. With the help of such therapeutic techniques as “Allergy Treatment”, “Collapse of Anchors”, “Getting Rid of Phobias”, “Flash” and others, it is possible to influence the consciousness of the individual and transform his internal state.
4. In setting goals. Thanks to various NLP techniques (SCORE, XCP, Mission, Time Line), it is possible to correctly identify the goal and find all the necessary ways to achieve it.
5. In modeling. To model the behavioral strategy of successful and brilliant people.
6. In acting. Special techniques allow you to change a person’s way of thinking, emotional state, and set the desired behavior. Body position can also affect the way you think, and certain emotions can cause changes in body position.
7. In public speaking. There are various techniques (Chamomile, Working with voice, Spatial anchoring, Working with your condition) that help you gain confidence and get rid of stiffness in public.
8. In education. Behavioral patterns, successful strategies, and various approaches to achieving goals are often used in the learning process.
9. In coaching. Various NLP techniques (Tuning, Leading, Goal Setting, Anchoring) are used in trainings aimed at helping to achieve clearly defined goals.
10. In interpersonal relationships. There are basic presuppositions that help to better understand other people and find an approach to each individual.
11. In self-development. You can change your way of thinking, tune in to the positive, and achieve success in the right areas of activity thanks to NLP techniques.

NLP techniques do not cause any harm to the body. They are focused exclusively on the thought process. To a greater extent, this discipline pays attention to identifying its own hidden reserves. The main task of this direction in psychology is to discover the potential of the individual, the talents of individuals, and teach the skills of quickly assimilating knowledge. NLP is unthinkable without rules that help improve well-being. The main thing is to pay attention to gifted people every day. This way you can develop your own talents.

Context-sensitive features

Context-sensitive features of a token are a set of features that contains information not only about the token itself, but also about its neighbors.
There are different ways to calculate these features. In classical algorithms, people often simply walked in a “window”: they took several (for example, three) tokens before the original one and several tokens after it, and then calculated all the features in such a window. This approach is unreliable, since important information for analysis may be located at a distance greater than the window, so we may miss something. Therefore, now all context-sensitive features are calculated at the sentence level in a standard way: using two-way recurrent neural networks LSTM or GRU. To obtain context-sensitive token features from context-independent ones, the context-independent features of all sentence tokens are fed into a Bidirectional RNN (single- or multi-layer). The output of the Bidirectional RNN at the i-th moment in time is a context-dependent sign of the i-th token, which contains information both about previous tokens (since this information is contained in the i-th value of the direct RNN) and about subsequent ones (i.e. .as this information is contained in the corresponding inverse RNN value).

Then for each individual task we do something different, but the first few layers - up to Bidirectional RNN - can be used for almost any task.

This method of obtaining features is called the NLP pipeline.

It is worth noting that in the last 2 years, researchers have been actively trying to improve the NLP pipeline - both in terms of performance (for example, transformer - an architecture based on self-attention, does not contain an RNN and is therefore able to learn and apply faster), and with from the point of view of the features used (nowadays they actively use features based on pretrained language models, for example ELMo, or use the first layers of a pretrained language model and additionally train them on the corpus available for the task - ULMFit, BERT).

The best books on NLP

1. Balyko D. Zapretov.net. 40 NLP rules for living a high life.
2. Bandler R., Grinder D. From frogs to princes. Neuro-linguistic programming.
3. Bandler R., Grinder D. Reframing: personality orientation using speech strategies.
4. Berger E. NLP for happy love. 11 techniques that will help you fall in love, seduce, marry anyone.
5. Beaver D. How to quickly absorb a large amount of knowledge. Let's learn without pain. The latest NLP methods.
6. Bodenhamer B., Hall M. NLP practitioner. Full certified course.
7. Davydova I. NLP for business and life. The art of hypnotic persuasion.
8. J. O'Connor. NLP. A practical guide to achieving the results you want.
9. Dilts R. Modeling using NLP.
10. Pligin A. How to turn the ghost of failure into the aroma of success in life.

Before using NLP to influence other people, check how the technique works for yourself. You should start by changing your own life for the better, creating an ideal “I-image”, and then experiment in society.

You may also be interested in:

Word embeddings

Let's take a closer look at what embedding is.
Roughly speaking, embedding is a concise representation of the context of a word. Why is it important to know the context of a word? Because we believe in the distributional hypothesis - that words with similar meanings are used in similar contexts. Let's now try to give a strict definition of embedding. Embedding is a mapping from a discrete vector of categorical features into a continuous vector with a predetermined dimension.

The canonical example of embedding is word embedding (word embedding).

What usually acts as a discrete feature vector? A Boolean vector corresponding to all possible values ​​of some category (for example, all possible parts of speech or all possible words from some limited vocabulary).

For word-form embeddings, this category is usually the index of the word in the dictionary. Let's say there is a dictionary with a dimension of 100 thousand. Accordingly, each word has a discrete vector of features - a Boolean vector of dimension 100 thousand, where in one place (the index of a given word in our dictionary) there is a one, and in the rest there are zeros.

Why do we want to map our discrete feature vectors to continuous ones of a given dimension? Because vectors with a dimension of 100 thousand are not very convenient to use for calculations, but vectors of integers with dimensions of 100, 200 or, for example, 300 are much more convenient.

In principle, we need not attempt to impose any further restrictions on such a mapping. But since we are building such a mapping, let’s try to ensure that the vectors of words with similar meanings are also close in some sense. This is done using a simple feed-forward neural network.

Top 5 NLP techniques that will be useful for every day

If you know some NLP techniques, you can manage people and achieve the desired result from them, for example, consent, some kind of benefit. Manipulation methods help to avoid unpleasant incidents and not fall into the trap of scammers.

What NLP techniques can be used daily:

1. Joining. If a stranger approaches a person, he perceives him as a danger. It is difficult to strike up a conversation with anyone on the street. It’s even harder to inspire trust in a person and inspire something in him. You must first take a close look at the individual and copy his behavior, gestures, and manner of speaking. By adjusting to the rhythm of a stranger, you can easily lure him with your offer.
2. Rapport. Building trusting relationships with the individual. It is necessary to find common character traits in the person you managed to “join”. It is important to break down the psychological barrier, lull caution, and inspire sympathy and trust in yourself.
3. Three positive answers. The individual is put into a light trance with the help of three questions to which the answer must be “yes”. By inertia, a person will respond affirmatively to an unfavorable offer.
4. Switching attention. If they want to distract a person from an important issue, their attention is transferred to another object. An individual's brain, like his vision, is capable of focusing on only one object. By switching his attention to another situation or thing, the individual escapes previous information that may be important to him.
5. Template break. With the help of a non-standard action, you can unsettle a person and force him to do what the manipulator wants. It is important to carefully monitor the individual’s reaction and respond in a timely manner to his numbness.

Embedding training

How are embeddings trained? We are trying to solve the problem of recovering a word from the context (or vice versa, recovering the context from the word). In the simplest case, we receive as input the index in the dictionary of the previous word (a Boolean vector of the dimension of the dictionary) and try to determine the index in the dictionary of our word. This is done using a grid with an extremely simple architecture: two fully connected layers. First, there is a fully connected layer from a Boolean vector of the dictionary dimension to a hidden layer of the embedding dimension (i.e., simply multiplying the Boolean vector by a matrix of the required dimension). And then, on the contrary, a fully connected layer with softmax from a hidden layer of the embedding dimension into a vector of the dictionary dimension. Thanks to the softmax activation function, we get the probability distribution of our word and can choose the most likely option.

The embedding of the i-th word will simply be the i-th row in the transition matrix W.
In the models used in practice, the architecture is more complicated, but not much. The main difference is that we do not use one vector from the context to define our word, but several (for example, all in a window of size 3). A somewhat more popular option is the situation when we are trying to predict not a word from the context, but rather the context from the word. This approach is called Skip-gram.

Let's give an example of using a problem that is solved during embedding training (in the CBOW version - word prediction by context). For example, let the token context consist of 2 previous words. If we were trained on a corpus of texts about modern Russian literature and the context consists of the words “poet Marina,” then, most likely, the most likely next word will be the word “Tsvetaeva.”

Let us emphasize once again that embeddings are only trained on the task of predicting a word from the context (or vice versa, the context from the word), and they can be used in any situations when we need to calculate the attribute of a token.

Whatever option we choose, the architecture of embeddings is very simple, and their big advantage is that they can be trained on unlabeled data (indeed, we only use information about the neighbors of our token, and to determine them we only need the text itself). The resulting embeddings are the average context for just such a corpus.

Embeddings of word forms, as a rule, are trained on the largest and most accessible corpus for training. Usually this is the entire Wikipedia in the language, because it can all be downloaded, and any other corpuses that can be obtained.

Similar considerations are used in pre-training for the modern architectures mentioned above - ELMo, ULMFit, BERT. They also use unlabeled data when training, and therefore are trained on the largest available corpus (although the architectures themselves, of course, are more complex than those of classic embeddings).

Basics

Where to start learning neurolinguistic programming? There are basic, basic techniques that were developed half a century ago by J. Grinder and R. Bandler and continue to be improved to this day. First you need to master them theoretically, so that you can then learn to apply them in practice.

1. Subjective experience: a person’s past, which has a huge influence on him.

2. Presuppositions - semantic components of NLP:

• the map is not yet the territory;
• consciousness, like life, is a systemic process;
• positive intention is the basis of any behavior;
• rapport (building trusting relationships);
• feedback, not defeat;
• and many others.

3. Representational systems: how a person receives information mainly from the outside (visual, auditory, kinesthetic and discrete images).

4. Metamodels and metaprograms.

5. Milton models: empty information that a person fills in himself in accordance with his beliefs, desires, and principles.

Access Keys: A single thing or phrase can force a person to take an action.

These are the basic theoretical principles of NLP. Without mastering them, achieving results in practice will be extremely difficult. You can find their detailed descriptions in popular science literature (the list is given below).

Why are embeddings needed?

As already mentioned, there are 2 main reasons for using embeddings.

• First, we reduce the dimension of the feature space, because it is much more convenient to work with continuous vectors with a dimension of several hundred than with Boolean feature vectors with a dimension of 100 thousand. Reducing the dimension of the feature space is very important: it affects performance, it is more convenient for training, and therefore algorithms learn better.
• Secondly, taking into account the proximity of elements in the original space. Words resemble each other in different ways. And different embedding coordinates are able to catch this similarity. I’ll give you a simple, rude example that has set everyone’s teeth on edge. Embedding is quite capable of grasping that the king differs from the queen in much the same way as a man differs from a woman. Or vice versa, a king differs from a man as a queen differs from a woman. Similarly, the connections between different countries and their capitals are similar. A well-trained model on a sufficiently large corpus is able to understand that Moscow differs from Russia in the same way that Washington differs from the United States.

But one should not think that such vector arithmetic works reliably. In the article where the embeddings were introduced, there were examples that Angela treats Merkel in much the same way as Barack treats Obama, Nicolas treats Sarkozy, and Putin treats Medvedev. Therefore, it is not worth relying on this arithmetic, although it is still important, and it is much easier for the computer when it knows this information, even if it contains inaccuracies.

In the next part of our article we will talk about the NER problem. We will talk about what this task is, why it is needed, and what pitfalls may be hidden in solving it. We will tell you in detail how this problem was solved using classical methods, how it began to be solved using neural networks, and we will describe modern architectures created to solve it.

NLP: rules for beginners

In neurolinguistic programming, special importance is attached to the habit developed or learned by the human brain to react in a certain way to certain phenomena.

We all develop these habits from birth. We are influenced by parental attitudes, as well as acquired life experiences and beliefs.

To be able to implement effective changes, it is first necessary to establish in detail the mechanism by which the brain triggers certain reactions to a specific situation.

In other words, you need to understand what a person feels - sees, hears, smells or touches at the moment the situation of interest occurs. Just before you react to it in a certain way.

In psychology, external stimuli are associated with sensory systems of their perception - modalities.

In total, there are five main ones: visual, auditory, tactile (kinesthetic), olfactory and gustatory. According to human senses. Each modality has its own submodalities.

For example, a visual modality may have the following submodalities : bright, pale, distant, close, colorful, monochrome.

95% When working with a client, an NLP practitioner spends most of his time studying a picture or image that pops up in a person’s mind when he is mentally immersed in a “working” situation. It is believed that in order to change something, you need to know exactly how “it looks” in a person’s head.

And only 5% of the time is required to change this. After all, the brain learns quickly!

But the most important thing that practitioners of neurolinguistic programming need to learn is the need to check the desired changes for environmental friendliness (safety) for humans.

Before changing a person’s beliefs and his systems of reactions to situations, one must make sure that the changes will not harm his personality and will only bring benefits.

The benefits and harms of NLP

In pursuit of success, you can actively use popular techniques, moving towards your goals. If a partner does not know how to recognize NLP techniques in communication, he will follow the invisible path outlined for him without hesitation.

The impact on relationships within a couple can cause harm. If a guy initially did not have tender feelings for a girl, but she programmed him for a relationship, such a union may turn out to be unhappy for the “in tune” partner.

Developing your abilities and becoming deeply interested in human psychology is a useful activity. The human brain is fraught with countless abilities that you need to try to develop, which will certainly reveal your potential and allow you to enjoy life.