Could dictionaries be written by robots?
In the last two steps we saw how technology can help us detect new words, and new meanings of existing words, as they enter the language.
As these technologies develop, they will bring significant benefits for dictionary-makers and save them a lot of laborious work, especially when the dictionary is being updated. But what about creating entries for all the words in the dictionary: how far can this process be automated so that machines, rather than humans, do most of the work?
Three central tasks in building a dictionary entry are:
Deciding how many senses each word has.
Writing definitions for every word or word sense.
Providing example sentences when needed, usually by selecting good examples of usage from your corpus data.
All three tasks are time-consuming and require a considerable skill – but can technology help us here too? The job of automatically finding word senses – word-sense disambiguation (WSD) – is a big research topic in the Natural Language Processing (NLP) community, because automatic WSD will bring all sorts of benefits: for example, improving the performance of search engines and machine translation systems. The solution is likely to involve looking at the contexts in which a word typically occurs – but research in this area continues, and the problem of automatic WSD hasn’t been cracked yet.
In the next two steps, we will look at the other two tasks – writing definitions and examples – and learn more about how much progress has been made towards automating them.
© Michael Rundell. CC BY-NC 4.0