AI can help with language study

Asunción Gómez-Pérez, an expert in artificial intelligence and language appointed for the “q” position at the Royal Spanish Academy.

Expert in Artificial Intelligence (AI) and Language Asunción Gómez Perez, appointed to fill the chair “q” From the Royal Spanish Academy (RAE), he stated that “Linguistic resources can provide richness for applications of artificial intelligence”, while these can help make language study “more effective and efficient”.

PhD in Computer Science and Artificial Intelligence He is the first person with a career path related to IT and computing to hold this position. Inside SAR.

The scientist born in Extremadura, Spain, serves as Vice-Chancellor for Research and Innovation and PhD at the Polytechnic University of Madrid, where she lives, and manages the Digital Innovation Node in Artificial Intelligence and Robotics for the Sustainable Development Goals set by the commission. European.

Your name appears in the list 2% of the world’s most cited scientists and scientistspublished by Stanford University in October 2021, and directed 106 research projects, with the money of which he managed to maintain the Ontological Engineering Group he founded.

“Linguistic resources can provide a richness for AI applications, but at the same time AI applications can help linguists and lexicographers study language to make their work more effective and efficient.”Asuncion Gomez Perez

Gomez Gomez told Télam. Perez: “Linguistic resources can provide a richness for AI applications, but at the same time, AI applications can help linguists and lexicographers study language to make their work more effective and efficient.”

The RAE plenary consists of 46 members each occupying a “chair” represented by the letters of the alphabet, in uppercase and lowercase, except for the letters v, w, x, y, z, Ñ, W, p.

On Thursday, April 7, Professor Gomez Perez was appointed by the RAE plenary to occupy the “q” chair, becoming The fourteenth woman to serve as an academic in the institution and one of the seven in their position.

The ‘q’ chair has been vacant since Gregorio Salvador’s death on December 26, 2020, and the Asuncion nomination was submitted by academics Luis Mateo Diez, Pedro García Barrino and Salvador Gutierrez Ordonez.

The RAE plenary consists of 46 members, each of whom holds a chair represented by the letters of the alphabet
The RAE General Assembly is made up of 46 members and each holds a “chair” represented by the letters of the alphabet.

In an interview with Tellam, Gomez-Perez recounted the connection between what he had been researching for years regarding computing and language.

Tellam: Why is RAE interested in integrating an academic from the ICT field?
Asuncion Gómez Perez: There is now a very important project in Spain linked to the Recovery, Transformation and Resilience Plan that has been awarded 1,100 million euros to modernize the Spanish economy. Within this money is a project called Spanish Language and Artificial Intelligence (LEIA) that will be implemented by the Royal Spanish Academy.

The goal of this project is that devices, machines, that speak or write in Spanish, use the correct Spanish when creating texts, taking into account all the characteristics since there are phrases that can be used in Spain that are not used in Latin America or vice versa. The goal is for all of these devices to speak or write correctly.

T: What is the language of machines today?
A: When we interact with applications, for example, the autocomplete function appears. What is happening? that we are writing and that autocomplete refers to a series of words. Oftentimes it is very easy to succumb to the suggestion that is being proposed to us. This may prevent us from being able to choose another word, that is, it will limit communication between people.

T: Why does that happen?
A: These apps use certain types of algorithms that are able to analyze the words we usually use when we type. So if a person specializes in a field, the words he will suggest are always the same. In the end, what it does is impoverish the vocabulary of a human being when he communicates through a machine, a digital application.

T: How can this situation be changed?
A: The device, for example, may come with a synonym, or when someone uses an acronym, they may know that that acronym equals an entire word. Suggest more accurate translations between languages. This would be the way to start offering certain types of resources that Dictionary Academy should enrich this type of application.

T: How was your career path in this direction?
I have been working with words since the beginning of my career, but I express the words in a language of communication that a computer understands. If a computer understands the meaning of a word, it will later be able to think of those and other words. In this sense, I began investigating the subject of ontology in the 1990s. When we do ontology we express the meaning of a word in logical languages. While the dictionary is intended for reading, the ontologies that contain words are intended for use by computers.

T: How does artificial intelligence come to modify the traditional concept of language?
A: The relationship between language and artificial intelligence is like our two-way street. On the other hand, language resources can provide richness for AI applications, but at the same time, AI applications can help linguists and lexicographers study language, to make their work more effective and efficient.

Stereotypes reinforced by language automation

And Argentine researchers specializing in linguistics and computing warned against this Technologies “tend to standardize everything in a standard language” Stereotypes or prejudices related to gender often reinforce the capacity for phobias and abnormalities.

“Technologies and automation in general, what they do is massively amplify stereotypes in very subtle ways. If we don’t have a trained eye, we won’t see them because they reproduce something that is natural to us,” Laura Alonso Alemani, PhD in Linguistics from the University of Barcelona and Professor of Computing at the University of Barcelona Cordoba National (UNC).

In this regard, he stressed that In language automation, capacity-related stereotypes become apparent (a form of discrimination against persons with disabilities) and aporophobia (aversion to people living in poverty) that “preps us when we communicate”.

Gender stereotypes are also reinforced.

Alemani added that when translated from English, a genderless language, a word like “nurse” is “systematically translated as nurse” and “doctor,” “doctor in the masculine,” which “obviously makes us think it is normal.” . To be women nurses and men men.”

Another challenge presented by language automation is “Impairment of power between languages”Tellam Luciana Pinotti, MD, PhD in computer science specializing in computational linguistics from the University of Nancy warned.

“The languages ​​that have currently experienced the greatest technological development are English and, in the last decade, Chinese. There are other languages, such as Spanish, that are much less developed, and they are not proportional to the number of speakers and also not proportional to the number of speakers. They consider regional differences.”

He added that these techniques “tend to standardize everything into a standard language and to consider other differences incorrect,” which correlates with the disappearance of minority languages.

In addition, they need a lot of numeric language data in these different languages ​​in order to function. “So languages ​​that have little record, such as the indigenous languages ​​of the American continent – the Mapuche and Guarani – do not have the support of this kind of technology,” Pinotti explained.

In connection with the panorama in the country of studies on natural language processing, known as computational linguistics, the researchers highlighted the Need more funding, investment and training.

“There are no incentives in the country to develop this field,” Pinotti said. “The PhD scholarships that Conicet pays for people trained in computing are impossible to compete with the opportunities these students have abroad or in industry.”

Another problem is the low participation of women in research in this field and the lower presence of Latin America.

Leave a Comment