Language is a central element of human interaction; it is what enables civilization, and develops in lockstep with it, encompassing new concepts or describing theoretical frameworks. With computer development reaching processing capacity and algorithms that enable them work with language, this field of technology is starting to affect language, as well. When using computers to identify, parse, or translate texts or speech, neural networks are the tool of choice for linguists and consumers alike.
Neural networks are computer systems designed to imitate human perception and thought processes. They achieve this by using a large number of processing nodes, called neurons, that activate in response to the presence or absence of particular elements in the data they receive. Specific patterns of activation correspond to real-world outcomes, such as identifying an object in a photograph, “guessing” a pronounced word, or the translation of a phrase in a foreign language. Neural networks can also be “trained” or “learn” from working with large arrays of input data by having the activation criteria of their neurons adjusted to align with expected outputs.
Neural networks have found use in consumer applications for text analysis and translation. Google Translate, for instance, allows arbitrary text to be translated between any two languages. By using a neural network for this, this service achieves relatively reliable and accurate translations. Besides its utility in allowing regular users to access content in unfamiliar languages, Google Translate has been found useful as a tool for preliminary text analysis (De Vries, Schoonvelde, & Shumacher, 2018). In this capacity, Google Translate’s output was found to only have small differences from the same text translated by professional translators (De Vries, et al., 2018). As such, this service represents an advancement in the fields of linguistics and translation, achievable by a neural network and novel computational technology.
Reference
De Vries, E., Schoonvelde, M., & Schumacher, G. (2018). No Longer Lost in Translation: Evidence that Google Translate Works for Comparative Bag-of-Words Text Applications. Political Analysis, 1-14. Web.