How Natural Language Processing is Changing Our World
From speech recognition to speech synthesis, and from machine translation to data mining, natural language processing is changing our world.
In language-related applications, computers are gaining intelligence at an amazing speed. Some computers can now not only recognize basic spoken words and sentences, they can also resolve lexical and sentence ambiguity based on the context; plus, they can recognize some idioms and metaphors. To top it off, computers are learning to detect emotion and respond appropriately. By extension, automatic translation is advancing daily, which may diminish the need to learn foreign languages for future generations. Speech synthesis has advanced in such a way that systems will soon be able to translate your speech using your own voice. Theoretically speaking, you will be able to hear yourself (or your voice, more correctly) speaking Hindi, Mandarin Chinese or even Mongolian in the not too distant future, without your necessarily having learnt any of those languages. With sufficient samples of your speech, such systems will be capable of putting together new sentences for you, in the new language. The systems just need to know your voice, and they will do the rest of the work.
Of course, automatic translation is a complicated task. Poetic language and uncommon metaphors and puns pose special challenges, as do certain expressions that may be considered “untranslatable”, requiring borrowing from the source language, adaptation, substantial paraphrasing or annotation. Still, as emphasized in tcworld, an international information management magazine, machine translation is becoming inevitable: “Over the next few years, every organization’s content strategy will rely on some type of machine translation” (“As Content Volume Explodes, Machine Translation Becomes Inevitable”).
As for data mining, while we all know how search engines are speeding up our research, more advanced searches can produce even better, more focused results, further eliminating the unwanted, irrelevant types of “hits” one normally obtains with ordinary search engines. Just watch this video to see how future search results can be refined with more intelligent searches: “How Natural Language Processing Is Changing Research”.
In this impressive video, Aditi Muralidaharan, a Berkeley graduate student explains her work on a new system called Word Seer. The system can save reading time for researchers by analysing digitized literary texts quickly, using parsing that targets useful parts of sentences, such as adjectives and verbs. Instead of performing a simple keyword search, the system extracts very specific data. The student gives the example of slave narratives being analysed for their references to God. Rather than simply typing in “God”, one asks specific questions about God: “What did God do?” elicits verbs, such as “gave”, “knew” and “blessed”, while “How is God described?” extracts adjectives, such as “good”, “holy”, “just” and “great”. The conclusion could be that slaves generally had a positive relationship with God despite their misery. For those working on this project, the hope is that researchers in the humanities will be convinced to use the technology based on the efficiency of the results. Rather than having a graduate student read through selected texts (with the word God in them) in five days, one can extract relevant information using the parser in five minutes.
Such advances in natural language processing herald a bright future for the currently not so bright technologies we use.
Posted by May Mikati on 19 May 2012, 9:17 AM
From speech recognition to speech synthesis, and from machine translation to data mining, natural language processing is changing our world.
In language-related applications, computers are gaining intelligence at an amazing speed. Some computers can now not only recognize basic spoken words and sentences, they can also resolve lexical and sentence ambiguity based on the context; plus, they can recognize some idioms and metaphors. To top it off, computers are learning to detect emotion and respond appropriately. By extension, automatic translation is advancing daily, which may diminish the need to learn foreign languages for future generations. Speech synthesis has advanced in such a way that systems will soon be able to translate your speech using your own voice. Theoretically speaking, you will be able to hear yourself (or your voice, more correctly) speaking Hindi, Mandarin Chinese or even Mongolian in the not too distant future, without your necessarily having learnt any of those languages. With sufficient samples of your speech, such systems will be capable of putting together new sentences for you, in the new language. The systems just need to know your voice, and they will do the rest of the work.
Of course, automatic translation is a complicated task. Poetic language and uncommon metaphors and puns pose special challenges, as do certain expressions that may be considered “untranslatable”, requiring borrowing from the source language, adaptation, substantial paraphrasing or annotation. Still, as emphasized in tcworld, an international information management magazine, machine translation is becoming inevitable: “Over the next few years, every organization’s content strategy will rely on some type of machine translation” (“As Content Volume Explodes, Machine Translation Becomes Inevitable”).
As for data mining, while we all know how search engines are speeding up our research, more advanced searches can produce even better, more focused results, further eliminating the unwanted, irrelevant types of “hits” one normally obtains with ordinary search engines. Just watch this video to see how future search results can be refined with more intelligent searches: “How Natural Language Processing Is Changing Research”.
In this impressive video, Aditi Muralidaharan, a Berkeley graduate student explains her work on a new system called Word Seer. The system can save reading time for researchers by analysing digitized literary texts quickly, using parsing that targets useful parts of sentences, such as adjectives and verbs. Instead of performing a simple keyword search, the system extracts very specific data. The student gives the example of slave narratives being analysed for their references to God. Rather than simply typing in “God”, one asks specific questions about God: “What did God do?” elicits verbs, such as “gave”, “knew” and “blessed”, while “How is God described?” extracts adjectives, such as “good”, “holy”, “just” and “great”. The conclusion could be that slaves generally had a positive relationship with God despite their misery. For those working on this project, the hope is that researchers in the humanities will be convinced to use the technology based on the efficiency of the results. Rather than having a graduate student read through selected texts (with the word God in them) in five days, one can extract relevant information using the parser in five minutes.
Such advances in natural language processing herald a bright future for the currently not so bright technologies we use.
Posted by May Mikati on 19 May 2012, 9:17 AM
No comments:
Post a Comment