Monday, March 28, 2011

Awkwardly Homophonic

Homophones are usually defined as words that share the same pronunciation, regardless of how they are spelled. While the presence of homophonic words in any language must have decreased the size of a language phonetically, but to a user it might some time create awkward situations. A few days ago, I fell to the pray of such a word... "come". The homophone of this word is never expected to be uttered in any casual conversations. But, in one way or the other, my pronunciation or expression had put me in a very very gawky situation and I struggled hard to justify myself that I never meant the other word. And, I curse the linguists who must have put such a pathetic homophones to such a common word in English language. Later, the situation reminded me of the effective communication courses taught during the training programs of my previous employer. The lady instructor while teaching us about correct way expressing and pronunciating, had explicitly put forwarded the example of homophone of "come" and told us to be very careful while interacting with clients. I never found the usefulness of those lessons until recently.
             While discussing about this with my friends, we thought that it will be nice to devise an algorithm in Natural Language Processing, which can detect such anomalies. I somehow think that, one heuristic way of detecting whether a person meant the common word or its homophone will be to make use of past history of the person about usage of words in his daily life w.r.t to the context. And, as the usage gets detected, count for each such words gets incremented. Now, if a tolerance limit of say 10 is set, then when usage of the homophone exceeds the tolerance limit, the system will detect that indeed the person meant the homophone, instead of the common word. For example, as I see myself, I never use such homophones in social gatherings, so my count was initially zero. So, when my count reaches 10, the natural language processor will detect that I indeed meant the homophone of "come". Again, the algorithm can make usage of weightage for each such homophones. So, in gist, making usage of historical data we can put some anticipation about what the person actually meant.

Related Posts by Categories

Widget by Hoctro | Jack Book

No comments: