Contextual surroundings of a passage

The contextual surroundings of a passage are the text that precedes and follows a paragraph, without which the passage loses its meaning. In ordinary reading this is natural, but for machine processing it is a problem, because the system selects individual passages in isolation. There are four most common obstacles: pronouns instead of naming the subject at the start of the paragraph, references such as as mentioned above, conditions stated in the preceding paragraph, and numerical data without a unit or period that are mentioned elsewhere. The solution is not to repeat the whole context in every paragraph, but to ensure that the first sentence of the passage contains the subject and that claims with a condition have the condition in the same sentence. The test is to read a random paragraph in isolation and see whether it makes sense.

See also: Content chunking, Definition Sentence at the Beginning, Snippet extraction.