The overall objective of any system of processing 

Natural Language (PLN) is to obtain representation 

message content of the phrases. automatic processing 

of a language is a problem of great complexity in which 

involving diverse and complex sources of knowledge: 

phonetics, morphology, syntax, semantics, pragmatics, 

knowledge of the world, and so on. Although in some cases these 

information sources can be considered independent 

in general have a relationship, without which, no 

interprestación can get a proper meaning and 

the role of words in a sentence. 
Because of this complexity, to address 

understanding of a language is usually followed one of the 

following ways: 1) It resolves some more subproblems 

simple, in some cases, they should make simplifications 

to be treated in an automatic way, such 

as morphological analysis, lexical tagging of texts, analysis 

superficial syntactic sentences prepositional linkage, 

sesambiguación the meaning of words, treatment 

specific linguistic phenomena as anaphora, ellipsis, 

etc. 2) Dr dimplifica task considering the language 

restringidasm in the size of vocabulary, complexity of 

used syntactic structures or semantic domain of 

application. 
Over recent years we find a great deal 

examples taken from the commentary tracks. In 

speech recognition applications that are restricted to 

bounded vocabularies, sonsultas to specific databases, 

dialogue systems on specific tasks, etc.. In other 

areas more directly related to the PLN, 

applications are machine translation, extraction 

and retrieval of information, summaries of texts, etc. in 

which, to a greater or lesser extent, are restricted to domains 

specific sonseguir acceptable results. 
Moreover, the fact of having large corpus of 

data, written or oral, annotated with linguistic information 

different nature of morpho-syntactic information, analysis 

total or partial syntactic, semantic information - along with 

operational, has the appearance and use of 

inductive approaches or methods based on corpus 

within the field of Computational Linguistics, which 

applied to different tasks PLN obtained a high degree of 

benefits. 
Inductive approaches, with or without information 

Statistics are of great interest to achieve 

disambiguation of natural language (LN), since in addition to 

provide acceptable results, models 

relatively simple and its parameters can be estimated to 

from data. This makes them particularly attractive, since 

that the change from one task to another, or even language, it is 

substantially reduces human intervention. Nevertheless, 

some cases of ambiguity can not be solved this 

way and must turn to a human expert to enter, 

For example, certain rules or restrictions that will help your 

resolution.