MACHINE TRANSLATION SYSTEMS

Report by 

Joana Salazar and Tamara Diez

 

Abstract

Introduction

Body

        - Common expressions  

        - Interrogative and exclamative sentences

        - Spanish proverbs

        - Phrasal verbs

Conclusion

Sources

 

ABSTRACT

Our report is focused on "Machine Translation" and in showing the difficulties that we face when we use online translators. Nowadays, it is a great advance to be able to use systems that allow us to translate documents very fast. Nevertheless, there are some difficulties in the linguistic aspect of a language that we will work on.

The examples we are going to use to show the problems that we can find in this translation system are the following ones: common expressions, proverbs, phrasal verbs, and interrogation and exclamative sentences. We are going to study the problems in the translation from Spanish to English, and in some cases also from English to Spanish. For this study, we have use the following online translators: Systran , Reverso and PROMPT-Online.

 

INTRODUCTION

We realize the term Machine Translation as the fully automatic translation. However, we also have to consider the whole range of tools that may support translation and document production in general, which is especially important when considering the integration of other language processing techniques and resources with MT. Thus, we define Machine Translation to include any computer-based process that transforms (or helps a user to transform) written text from one human language into another. 

Moreover, related to Machine Translation we have similar terms that also involve translating documents but with the help of other sources. Thus, we can mention the term Human-Assisted Machine Translation, which is the style of translation in which a computer system does most of the translation, appealing in case of difficulty to a human for help. Another one is the Machine-Aided Translation, which is the style of translation in which a human does most of the work but uses one of more computer systems, mainly as resources such as dictionaries and spelling checkers, as assistants.

To sum up, it is clear that human's help is necessary when we want to translate a document. We clearly see that both in Human-Assisted MT and in Machine-Aided T, in which more than a machine is needed to complete a work. There are some problems related to linguistic features that machines cannot completely solve and that we will cover in the following paper.  

 

BODY

COMMON EXPRESSIONS

The following table shows the grammatical mistakes that we find when we use the 3 machine translators mentioned before.

SOURCE (Spanish)

SYSTRAN

REVERSO (El Mundo)

PROMPT ONLINE

CORRECT TRANSLATION

Me voy de compras I go away of purchases I me go shopping I me go shopping I am going shopping
Mi casa está en frente del parque My house is opposite of the park My house is opposite the park My house is opposite the park My house is in front of the park
Me gusta tu vestido I like your dress I like your garment I like your garment I like your dress
Mi hermano se llama Txomin My brother is called Txomin My brother is called Txomin The name of my brother is Txomin My brother is called Txomin
Me estás molestando You are bothering to me  I am not able stand you You are bothering me You are bothering me 
Te has equivocado You have been mistaken You have been wrong You have been wrong You are wrong
No merece la pena ir a clase  It is not worth the trouble to go to class It(He,She) is not worth going to class  It is not worth going to class  It is not worth going to class
La clase está llena de estudiantes The class is full of students The class is full of pupils The class is full of pupils The class is full of students

This table shows mainly grammatical mistakes. However, as we have not analize a great amount of sentences, we cannot determined the level of the errors that these translators make. What we see is that some systems are better than the others; for example, Reverso and Prompt use old-fashioned vocabulary ("garment", "pupils"). In other cases, there are random mistakes in all of them, such as the wrong use of prepositions (Systran: "You are bothering to me"; "My house is opposite of the park"). 

 

INTERROGATIVE AND EXCLAMATIVE SENTENCES

SOURCE (Spanish)

SYSTRAN

REVERSO (El Mundo)

PROMPT ONLINE

CORRECT TRANSLATION

¿A dónde vas? To where you go? To where do you go? To where do you go?  Where do you go?
¿Qué dan en la tele? What give in tele? What do they give in the TV? What do they give in the TV? What is on TV?
¿Cuándo termina la clase? When finishes the class? When does it(he,she) finish the class? When does it finish the class? When does the class finish?
¿Cuántos años tienes? How many years you have? How many years you have? How old are you? How old are you?
¡Qué tarde es! What late is! What evening it(he,she) is! What evening it is! What late is!
¡Qué bonito es tu vestido azúl! How pretty is your blue dress! Nice what is your blue garment! Nice what is your blue garment! How pretty is your blue dress!
¡Qué mayor eres! How greater you are! Major what you are! Major what you are! How old you are!
¡Qué miedo me das! That fear gives me! What fear you give me! What fear you give me! What fear you give me!

In this table, a lot of mistakes are shown. In the previous table the errors were not as notable as in this one. It seems that in these kinds of sentences machines are likely to get confused. First of all, we noticed that Systran  does not used auxiliares when translating questions ("What give in tele?"; "When finishes the class?"). We also see that this translator does not recognize the Spanish written accent ("That fear gives me!". It should be "What").

In the other systems, we also see that a literal translation is done without taking into account the expression itself. Finally, what we realize is that both Reverso and Prompt make vocabulary mistakes ("What evening it(he,she) is!" instead of using the word "late"), use the double subject in their translations ("When does it finish the class?"), old-fashioned vocabulary ("Nice what is your blue garment!"), and in Wh- questions, instead of putting wh- words at the beginning of the question, they make a literal translation and put a preposition there ("To where do you go?"). 

 

SPANISH PROVERBS

SOURCE (Spanish)

SYSTRAN

REVERSO (El Mundo) PROMPT ONLINE CORRECT TRANSLATION
A buen entendedor, breve hablador To good entendedor, brief talkative To good one who understands, brief talker To good one who understands, brief talker To a good understander a short speaker
Agua de Mayo, pan para todo el año Water of May, bread for all the year Water of May, bread for the whole year Water of May, bread for the whole year May water bread for the whole year
Al cabo del año mas come el muerto que el sano After the year it more eats the dead that the healthy one After the year more the dead man eats that the healthy one After the year more the dead man eats that the healthy one At the years end the dead eats more than the living.
Año de nieves año de bienes Year of snows year of goods Year of snow year of goods Year of snow year of goods A year of snow, a year of plenty
En boca cerrada no entran moscas Closed mouth flies do not enter Closed mouth flies do not enter If you keep your mouth shut you won't put your foot in it When the mouth is closed, flies don't enter
Hombre narigudo pocas veces cornudo Not very often cornudo narigudo man Long-nosed man rarely horned Long-nosed man rarely horned A long-nose'd man is seldon a Cuckold
Mas vale tarde que nunca But it is worth late that never Mas costs(suits) late that never Better late than never Better late than never
No hay mal que por bien no venga It does not have bad that by good does not come Every cloud has a silver lining Every cloud has a silver lining There's no ill but time may ease the smart 

We can see here that the translators we have used (and surely others, too) are not useful for these sentences. They are not as what we expected them to be. Proverbs are expressions that are typical of the language in which they are told, so in introducing Spanish proverbs in the 3 translators, we were not looking for a literal meaning. We were expecting a correct form, that is the way in which they are said in English. They have their equivalent meaning, but those translation systems are not able to recognize them. They are made to work more or less word by word, without looking to the whole and without paying attention to the changes we find, for example, in proverbs. Even if we found a lot of mistakes in this area, we noticed that Systran works worse than the other two translators.

 

PHRASAL VERBS

 SOURCE

SYSTRAN

REVERSO (El Mundo)

PROMPT ONLINE

CORRECT TRANSLATION

Shut out       Cerrado hacia fuera   Cerrado hacia fuera cerrado hacia fuera  Excluir
Carry off Llevé apagado Llevar Llevar  Salir airoso/a
Get into Consiga en  Entrar Entrar Entrar
Look out Mirada hacia fuera Mirar Mirar Ir con cuidado
Get up Levántese Levantarse Levantar Levantarse
Go away Salga Marchar Marchar Marcharse
Put off Puesto apagado Disuadir Aplazado Posponer
Bring back Traiga detrás Devolver Devolver Devolver
to look down on somebody  para mirar abajo en alguien despreciar alguien despreciar alguien Despreciar a alguien
to get over something para conseguir sobre algo terminar algo terminar algo Recuperarse de algo
To give in Para dar adentro Presentar Presentar Ceder
To give something in Para dar algo adentro Presentar algo Presentar algo Entregar algo
To bring out Ponga en evidencia Publicar(Recalcar) Sacar Publicar
To take away  something for someone Para quitar algo para alguien Llevar algo para alguien Llevar algo para alguien Quitar algo a alguien 

In the case of Phrasal Verbs, the main problems is that the literal meaning is not the correct one. That is their main characteristic. In Spanish such terms are not used to define some expressions. So machine translators should be able to recognize phrasal verbs in the correct form.  

We have used two different forms in this table to translate these terms. The first ones are written without the preposition "to" that indicates the infinitive, and the rest of the table using it. The result in this case is that Systran makes the worst translations, that is it makes a literal translation. It takes into account the two or more words that form the phrasal verb and not everything as a whole. As a result, it makes translations such as the following ones: 

- Shut out = "Cerrado hacia fuera"  (instead of "excluir")

- to look down on somebody = "para mirar abajo en alguien" (instead of "Despreciar a alguien")

 

CONCLUSION

First of all, it is obvious that the translators that we have used do not understand the context and translate what we write literally, that is word by word. In fact, all translators work in a similar way, but some of them are better than others as we see in the tables we have created. So we can say that we cannot rely on these machine translators as the only source for our translations. They are a good help that save us time in our work, but we should be careful in their use and use our knowledge in the language we are working on. 

Thus, nowadays we could say that although machine translation systems are hugely developed, we still need the human source. The mistakes are so big that they shock the person who knows what the real terms should be. This does not happen because of lack of vocabulary. It is just that they do not have the capacity to understand the meaning of some combination of words and give the proper translation as a person does. They are not able to adapt what they have in their database to the correct usage of each sentence or expression. In fact, they are machines, so we cannot expect them to work as a human mind. The ability to make an understandable translation can only be achieved by human beings. 

There is not much to say at this point. We have seen the general idea of these systems, about how they work and the problems they have. As a result, it is evident that Human sources will still be neccessary, if not always. One of the main problem that we think these sytems have is that some of them are not updated, so they work with old guides of language. Also, there are some words that have several meanings, but machines do not identify when we need one meaning or the other. It is also important to mention that we have noticed that there are a lot of syntactical mistakes, that is the correct word order is not the appropriate. Finally, big mistakes are also made in the use of grammar, as we stated before.

 

SOURCES

ONLINE SOURCES

    - REVERSO: http://www.elmundo.es/traductor/ 

    - SYSTRAN: http://www.systransoft.com

    - PROMPT ONLINE: http://translation2.paralink.com/

    - About Machine Translation Systems: http://www.languagepartners.com/reference-center/whitepapers/catinto.htm 

BOOKS:    

    - "PROVERBIOS, REFRANES Y TRADUCCIÓN " by James Howell. (Estudio, compilación y notas por F. Javier Sánchez Escribano. Published by the University of Zaragoza.

    - Harrap´s "PHRASAL VERB. DICTIONARY". Published by Harrap's Dictionaries.