ecommerce_hybrid_search queries are not realistic? #4473
Description
Queries are generated as a sample of the first 8 words of product titles (or less if the title has less than 8 words). As an example, here are the first 50 queries that have been generated when I ran the data preparation scripts:
"Disney Sum Panel No Sew Fleece Throw Kit,"
"Seattle Mariners 2016 Topps MLB Baseball Regular Issue"
"GreenStar Black Sign Vinyl 24 x 10Yd, Graphics"
"Kids Satin Bonnet Sleep Cap Wide Elastic Band"
"Great Papers! Gold Foil Stars & Streamers Flat"
"FAT BIG CAT Love Notes Framelits Clear Stamp"
"3 Colors Door Handle Frame Trim For Range"
"MOOSUN Front Traction Suspension Control Lower Tie Bar"
"F-sport OEM Factory AIR Filter 2006-2013 IS250 IS350"
"Oil Pastels Set,24 Assorted Colors Non Toxic Professional"
"Organ Home Sewing Machine Needle"
"Kato USA Model Train Products N Bethgon Coalporter"
"16 Sheets Valentine Heat Transfer Vinyls HTV Iron"
"K&N RF-1037DK Black Drycharger Filter Wrap - For"
"Sublimation Coating Spray with High Gloss, Super Adhesion,"
"Free People Berlin Mini Burnt Orange XS (Womens"
"3 PCS Medium Type Stainless Steel Scissors Black"
"X AUTOHAUX Master Driver Side Power Window Switch"
"Darice 2014-102 Embossing Essentials Dies-Flowers"
"Female French Stand Dress Form White/Burgundy"
"Caltric Carburetor Compatible with Yamaha Grizzly 350 2X4"
"Sterling Silver Pave Bezel Gemstone Pendant - Moonstone"
"England Navy Blue Soccer Hoodie (Hooded Sweatshirt). Brand"
"2019 Topps Pro Debut #51 Ronny Mauricio"
"MAXXHAUL 50214 Spare Tire Carrier for Utility Trailer"
"JP Auto Side Door View Mirror Compatible With"
"Cyleto Front Fork Oil Seal and Dust Seal"
"APS NB-T3103S Mirror Polished Nerf Bar Bolt Over"
"Ghostbusters Stantz Name Tag Iron On Uniform Embroidered"
"We R Memory Keepers Mini Tool Kit Aqua"
"1984-1989 911 Carrera Genuine Leather Seats Cover Custom"
"C-Lon Bead Cord, White"
"My Favorite Things Die-Namics Die, 2.5 by 2.75-Inch,"
"Outland Models Train Railway Platform/Loading Dock (Wide) w"
"On-The-Go Ladder Telescoping Ladder for 5th Wheel RVs,"
"LG Electronics 5210JA3005W Refrigerator Water Line Tubing"
"Ft. Hood III Corps Challenge Coin"
"4L60E 4L65E Remanfactured Transmission Solenoid Kit W/Harness Compatible"
"Rockwell RW9149 Sonicrafter Sanding Finger Pad, 1-Piece"
"Premium Tailor Scissors Heavy Duty Multi-Purpose Titanium Scissors"
"Julie Wang 10 Sets Oval Bezel Pendant Kit"
"Stamping Bella Cling Stamps, Oddball Sparkle Fairy"
"Swingline 2-3 Hole Punch, Adjustable, Heavy Duty Hole"
"Sewer Hose Connector For Connecting Sewer Hose To"
"Echo Park Paper Company Adventure Awaits Collection Kit"
"Dorman M630120 Brake Master Cylinder Compatible with Select"
"Janome Large Spool Holder Cap"
"Prickly Pear Cactus Stencil, 12 x 14 inch"
"Belmor BS-1881-1 Black Bug Screen Truck Grille Cover"
"SixSixOne 401 Chevron Black/Purple 2011 (Size: L) Full"
This doesn't look very realistic to me, as these queries look more specific than what I would expect average queries to look like on an e-commerce website. For instance, there are barely any queries with one or two common words, and the share of queries that contain rare terms like product identifiers is higher than what I'd expect in an ecommerce query log?