Electronics specification analysis toolkit

ner4hardware.org provides toolkit for finding vendors, product brands, product families, product series, and product models in computer component / device names and specifications using named entity recognition model. 

Image depicting computer parts
Popular use cases
Inventory management
Analyse inventory records to extract product vendors, families and models for simplified merchandise grouping.
Customer support ticket management
Analyse customer requests to extract component and device names in question for quicker resolution.
E-commerce recommendations
Parse gadgets names and descriptions for more accurate product suggestions.
Product information management
Extract related hardware components from detailed product specifications.
Technical characteristics
NER model size

                                        Model is trained using unstructured product names found on the internet
                                        in English and Lithuanian languages. Compact model takes less than 35MB
                                        to store on disk.
                                    
Memory usage

                                        Using hosted services does not require any substantial memory or CPU capacity.
                                        For enterprise users, model consumes up to 150MB memory on average per deployment.
                                        Using batch processing can create spikes of up to 450MB of memory usage.
                                    
CPU usage

                                        Using batches allows you to send up to 1k queries per request.
                                        This amounts to 4-6ms per query on desktop Intel Core i7 processor at 2.30GHz.
                                        For example: 1k queries weighting ~126KB can be processed in 4-6seconds.
                                    
Model metrics

                                        From the usual model metrics, performance is as follows: Precision - 99.86%; Recall - 99.92%;
                                        F-score - 99.89%.
                                    
Currently supported product families
Samsung
Watches, phones, storage devices and other products manufactured by Samsung Electronics.
Apple
Mac computers, iPads, iPhonesand other gadgets manufactured by Apple Inc.
Lenovo
Laptops, tablets, desktops, servers, monitors and other gadgets manufactured by Lenovo.
Xiaomi
Smartphones, watches, tablets, monitors and oher products manufactured by Xiaomi.
HP
Laptops, desktops, printers, monitors, networking and oher products manufactured by HP inc.
Dell
Personal computers, laptops, servers, monitors and other devices manufactured by Dell.
ASUS
Personal computers, motherboards, monitors, network devices and other gadgets manufactured by ASUS.
Acer
Desktops, laptops, monitors, projectors and other devices manufactured by Acer Inc.
Google Pixel
Smartphones and watches manufactured by Google.