The definition of semantic relations between named entities or other objects of
the text is the task of extracting relationships. Examples of relationships are author and
book or organization and chief's office.
Using N-gram. N-gram is a sequence of n elements. In NLP N-grams are used to
construct probabilistic models, text similarity tasks, text and language categorization.
By constructing an N-gram model you can determine the probability of using a given
phrase in the text. The N-gram model calculates the probability of the last word of the
N-gram, if all the previous ones are known, while it is assumed that the probability of
occurrence of each word depends only on the preceding words.
N-grams are used in the task of detecting plagiarism. The text is split into several
fragments, represented by N-grams. A comparison of N-grams with each other allows
you to determine the degree of similarity of documents. In a similar way, the problem
of correcting spelling mistakes can be solved by selecting candidates for replacement.
Neural networks are also used for NLP. Artificial neural networks are a system of
connected and interacting simple processors - artificial neurons. The algorithm of such
processors is often very simple. For example, the processor can simply transform the
signal received at the input, using a certain mathematical function, in the output. And,
nevertheless being connected to a fairly large network with a managed one interactions,
such individually simple processors together are capable to perform rather complicated
tasks. Recurrent neural networks [9] differ from other types of networks by the fact
that except for links that move from one neuron to another directly, as in networks
direct distribution, as well as communications that take place in time. That is, the signal
from one neuron in step t will go to another (or the same) neuron at step t + 1. Thus
recurrence neural networks can be saved information in time, thereby "memorizing"
some data. These are theirs peculiarity just very much helps in the translation,
classification and processing natural text as a whole, since our language is arranged in
this way, that some data at the beginning of the block of text, may affect the
understanding and translation at its end.
SNA - convolutional neural networks have best shown themselves the recognition
of objects and images in pictures, the classification of images, highlighting features
and compressing data. However, they found application in text processing.
As mentioned before, system has adaptive evaluation method as well as some sort
of open question evaluation system that based on NLP (Natural Language Processing).
Let’s consider the different techniques used by computer-based assessment
systems, as well as existing CAA approaches. Methods of automatic evaluation of free
text are divided into three main types: statistical, extraction of information and
complete processing of natural language [10].
The statistical technique is based only on the matching of keywords, so it is
considered a bad method. It cannot solve such problems as synonyms in student
- 508 -