|
|
|
Incremental Evolutionary Learning for Fuzzy Text Fragments
learning fuzzy grammars
|
The ability to understand text fragments can benefit many information extraction tasks. The main challenge in this project is to recognize the multiple patterns that exist across the heterogeneous sources.
Although machine learning methods can be used to tackle this problem, they are constrained by the common train/test setting which is not efficient when immediate relearning or adaptation to a new example
is required. If the training set is not sufficiently diverse to cover all unseen examples, the training phase must be repeated each time a new example is added to improve the training set.
Thus, it is highly desirable for a method that can support incremental learning of new examples by slight modification of the trained system, without sacrificing or repeating learning on past examples.
The purpose of this study is to investigate incremental evolutionary grammar learning techniques which can recognise predefined domain specific text fragments and have the
ability to learn by performing modification to its fuzzy grammar collection when the designed grammars are found inadequate.
This ensures the ability of the grammar collection to encompass the newly learnt and without losing the previously seen example and most importantly reserve any supplementary retrain and retest phase.
This algorithm is being applied in smart text processing and applications such as intelligent response agents and recommendation system.
|
|
