We are thinking over a NLP solution to allow site customer's discover merchandise by entering plain English words, though they may not conform to the language rules necessarily. In the process we have found the part-of-speech(POS) tagger component which identifies the POS for each term. The default tagger sometimes is not able to identify proper nouns for brand names properly. To identify the correct POS for such words/terms, the model needs to be re-trained.
The idea is to then identify patterns of words/phrases to map them to search engine queries for eventually fetch search results. In light of the above , i had a couple of questions. 1. Does this approach sound reasonable? 2. Is there a number of things to be done, while retraining the model for newer content? -Vikas
