Machine Translation (MT) is the automated process computers use to translate text from one natural language to another (e.g. Google Translate). Human-translated text is based on linguistics and grammatical understanding of a language pair and is usually quite accurate. However, not every language has a word-for-word translation available, so correctly getting the sentiment is of more importance. MT is much quicker, however, even with the backing of immense amounts of diction and grammatical language software, the translation quality can be lower.
Machine translation technology includes rule-based machine translation, statistical machine translation, and neural machine translation. Rule-based MT involves large amounts of back-end dictionaries in both source and target languages and dialectal rules to provide high-quality translations. By utilizing the sentence structure and sentiment of the source language, rule-based MT can better transition to the target language without losing substance or context. This process can take longer than statistical MT, but the quality to close to a human translated text.
Statistical MT tends to be the lower cost, but lower quality language translation software. Statistical MT utilizes translation models based on machine learning. Therefore, they change over time when they get new examples and translations. This can lead to more fluent understanding, but more inconsistent translations. Neural machine translation is further specialized, using neural network models to focus only on the source and target languages to provide better translations than traditionally phrase-based statistical methods.
As a product of machine learning, machine translation is based on analytics and probability. This is not always an accurate representation of the original text. Applying linguistics software to a large text may produce the most likely possible translation, but not necessarily the accurate one. Machine translation software is still expanding and developing to bridge the gap between fluency and accuracy.