Computational Modeling of Morphological Effects in Bangla Visual Word Recognition

Area of Science:

• Psycholinguistics
• Computational Linguistics
• Cognitive Science

Background:

• The mental lexicon's organization of polymorphemic words is debated.
• Understanding word processing in morphologically rich languages like Bangla is crucial.

Purpose of the Study:

• To model the organization and processing of Bangla polymorphemic words.
• To determine if words are accessed holistically or via decomposition.
• To develop a computational model explaining Bangla derived word processing.

Main Methods:

• A masked priming experiment with native Bangla speakers.
• Analysis of reaction times (RT) and error rates.
• Development of computational models based on linguistic features and RT data.

Main Results:

• Morphologically derived words are generally accessed through decomposition.
• Word processing depends on factors like base/surface frequency, suffix characteristics, and productivity.
• A novel computational model integrating linguistic features achieved ~80% accuracy.

Conclusions:

• The mental lexicon likely decomposes Bangla polymorphemic words.
• The proposed computational model effectively explains processing phenomena.
• This model offers improved accuracy over existing frequency-based models.