14 Oct Artificial Intelligence in the Language Industry
In the past few years Artificial Intelligence (AI) has been a hot topic in virtually any field of knowledge. From Cybersecurity Defence to Health Care, from Market Prediction to Accounting, from Human Resource Management to Logistics. The language industry is not exempt from it either.
Here are a few examples of Artificial Intelligence applied to the language industry. We’ll briefly go through some interesting applications, including, but not limited to, those related to translation and interpreting.
Artificial Intelligence in the Language Industry
As you all know very well, Machine Translation (MT) is the translation of a source text into a target language done by a computer, without human intervention. The use of AI in Machine Translation has led to the development of Neural MT systems, which are based on Machine Learning.
Training a neural system requires feeding a large quantity of material into neural algorithms. Those of you who have seen NMT put in practice have seen how impressive the results can be. As it does mirror the way a human mind works. However, we are at a stage where the human effort is still required, as the translated text may sound very natural and fluent. But it may not match exactly what the source text itself. Post-editors need to learn how the engine works to better detect the most typical errors and make the most out of the neural systems.
Text-to-speech software can easily convert written text into audio and it is used for example to convert manuals or training content into multimedia courses. By recognizing single phonemes, the voice synthesizer (or, alternatively, human recordings of the single phonemes) generates the different sounds in a given language.
Today these systems are more accessible to everyone. And thanks to AI and machine learning they are getting more and more accurate in a way that it is sometimes hard to distinguish a real voice from a synthetic one. In addition to being accurate, these systems are now more widespread and accessible, which has led to their use in other fields, such as the creation of multimedia or audio content from written documentation.
Opposite to Text-to-speech, Speech-to-text software convert audio into written text (used, for example, for dictation or transcription). In order to be able to do so, voice recognition programs use statistical and pattern recognition techniques. The most sophisticated tools in this sense are fed with grammatical rules (so that they recognize which types of phrases make sense) and integrated with dictionaries to recognize which words tend to follow each other.
By using these systems, users get texts written more rapidly, by dictating while doing something else and the result is that they spend less time in front of the PC too. These systems are not yet 100% accurate, as the way of speaking still has an impact on the result of speech recognition.
The Automated interpreting essentially combines three forms of previously existing AI technologies: voice recognition software (Speech-to-text and Text-to-speech) and Machine Translation.
Automated interpreting can be broken down into three basic steps:
- Voice recognition software picks up a speaker’s voice and converts it into text.
- The text is run through a Machine Translation engine.
- The translated text is converted automatically into speech in the target language.
Even though this seems like a simple process, there are quite a few glitches in voice recognition software and machine translation, which make Automated interpreting still an unreliable system. At this stage it could be used in cases where you only need to grasp the intent of a spoken communication, as you would still need a human interpreter to fully convey a speaker’s message.
With the help of Artificial Intelligence, sentiment analysis runs through texts to look for emotions and opinions of employees and customers and it is less expensive than paying for a human analyst to examine every news and social media post for a specific brand or product.
This software assigns negative, positive or neutral scores to news and/or social media posts, based on the analysis of the text run by the system. The software scans the contents of news and posts that mention a brand or a product and based on the algorithm a score is assigned to analyse the trend throughout different texts.
Neural networks are currently used to teach algorithms more complex grammatical structures and the accuracy of Sentiment analysis has improved and considerably over the past few years, with an accuracy rate of more than 90%.
Intent Analysis is a system based on AI that detects the intention of a text (e.g. sale, complaint, purchase, etc.). In the same way Sentiment analysis evaluates a text, Intent Analysis, thanks to an algorithm, identifies its intent and hence facilitates response actions. It can lead to many benefits for the Company who adopts this system. It can help understand feedbacks from the customers about new products or understand the nature of a service which we have to provide or improve.
Natural language generation
Natural language generation is an automatic system used for the creation of texts from non-linguistic data, which is useful for summarizing or explaining the contents of a computer database. Automated NLG can be compared to the process humans use when they turn ideas into writing or speech.
Through the use of Artificial Intelligence in language industry, the software identifies the most salient insights by understanding the context and a natural language text is generated to convey complex concepts in an easy way. These systems are fed with grammatical rules and dictionaries and the different templates can be tailored-made for the specific end-use.
These are just a few examples of how AI can be of help when it comes to the language industry. But there are many more applications of use. As the AI techniques and software improve day by day, there are increasing ethics issues revolving around the use of AI and the human language.
We can always choose whether to embrace these new systems or not. But if we want to play a key role in what seems to be happening anyway, we need to be one step ahead and know exactly how these systems works, so that we are not left behind.