You are here
Summary
Mon, 03/16/2015 - 12:04 — antardhvani
Avoid artistic flourishes as these are unlikely to be parsed correctly by most Deep Parsers, and hence have a low chance of being translated correctly.
Extremely long sentences are unlikely to be parsed correctly by most Deep Parsers. In our case, the reliability of the parse-tree declines sharply as the number of words in a sentence increases beyond 25. The Deep Parser that we use is one of the few parsers that can identify high-quality parse-trees. We do not translate sentences that have parse-trees of low quality, as this output is almost sure to be defective, . If required, translations of individual words can be provided for such sentences.
A sentence that has a very complex structure will, in all likelihood, be translated poorly. The WMScore program can identify sentences in the English passage that are very complex.
Another point to observe is that splitting a sentence containing a Relative clause into an inner clause and an outer clause can make the translation easier to comprehend. It is also more likely to get Case, Number, Gender, and Tense right in each clause.
Ensure that a Relative Clause is immediately adjacent to the Noun that it modifies, because intervening phrases and words are likely to result in a bad translation. For instance, in the sentence ‘The sample from the lot that he tested was defective’, the Deep Parser may assume that he tested a ‘lot’ (and not a ‘sample’), thereby resulting in a defective translation if that assumption was incorrect.
Multiple wh-pronouns such as what, why, where, how, and when in a single sentence (for e.g., ‘I will tell you why I did what I did.’) are likely to cause problems in translation.
Avoid sentential subjects (such as ‘That he would win the match easily was obvious to all.’). Such sentences usually have a simpler form (for e.g., ‘It was obvious to all that he would win the match easily.’) that can be used instead.
It should be clear from the above that translating a sentence from English to Hindi can be tricky because a number of transformations can go wrong in each step – for e.g., Case, or the right form of the Verb, or inflections for other components in the sentence (Nouns, Prepositions, Adjectives, Determiners, Adjuncts, etc.).
If English text is to be translated automatically to Hindi reasonably well, content editors should keep the following points in mind: