Nitin Solanki JanBask Expert Latest answer
Answered on Nov 9, 2019
There are multiple approaches to automatically perform the extraction,
1. Rule-based Approach: In this approach we already predefined rules for the syntax and other grammatical characteristics of the language which plays an important part in extracting info from text.
2. Supervised: Consider a sentence A. It has 2 entities A and B. Our supervised machine learning model has the duty to find is there any relationship between the two entities. So, in other words, the task of extraction turns into the detecting relationship in a supervised ML model. The only con of this approach is that it requires a lot of data(labeled) to train.
3. Semi-supervised: When we don’t have enough data(labeled), we can use a set of triples to formulate high-precision patterns that can be used to extract more relations from the text.
4. Distant supervision: It is a type of ML in which a model is learned given a very weak labeled training data (training data is labeled automatically based on heuristics/rules)
