Ascendant AI Technologies: Unlocking New Horizons for Computational Linguists
The rising proliferation of AI technologies has created a renewed interest in computational linguistics, given the crucial role played by the latter in bestowing on AI its different abilities at parsing human language and generating text. And yet, despite having come this far, AI still has a long way to go in terms of helping us deconstruct the various nuances of language.
Computational linguistics is a branch of linguistics that operates through the application of computer science techniques for the analysis of language. Computational linguists therefore work towards ensuring that computer applications are able to execute tasks such as speech recognition, machine translation, grammar checking, and text mining, among several other tasks. For the longest time, computational linguistics has also focused on testing linguistic theories to understand how the capacity to participate in language develops within individuals, and if computers can mimic that capacity. Therefore, a couple of decades ago, natural language processing systems (NLP) had to be constructed entirely through the human hand, and based on a dataset with less than a couple hundred examples.
However, according to researchers, there has been a recent pivot in the focus areas of research within computational analysis. Owing to the popularity of large language models such as ChatGPT, which are adept at generating natural language, we now have the ability to undertake precise analysis about the workings of language, as opposed to the speculative analysis that researchers were forced to undertake earlier.
Let us then zoom into one such specific research question. Given that language works by combining individual words into phrases to generate meaning and clear ambiguities surrounding which meaning of a word is being used, it is important to undertake contextual analysis to understand this aspect of language. This is also a question that deeply troubles large language models (LLMs), as existing research shows that LLMs have difficulty parsing the metaphorical and idiomatic use of language. Therefore, in order to solve this, researchers like Professor Carole Tiberius, from the Centre for Linguistics at the University of Leiden, hope to create a dictionary containing the usage patterns of words across their different meanings in order to garner a better understanding of how language functions.
And the most interesting thing is that these new developments seem possible too. Recent models such as BERT (Bidirectional Encoder Representations from Transformers) have set higher standards for computational linguistics, which—coupled with cutting-edge hardware such as GPUs (Graphics Processing Units) and TPUs (Tensor Processing Units) to accelerate deep learning processes—have opened up the possibilities for unimaginable applications across various sectors.
Sources
https://premioinc.com/blogs/blog/what-is-the-difference-between-cpu-vs-gpu-vs-tpu-complete-overview