The Alexa Skill ‘Turf Expert’ vocabulary of terms is currently in version 2 and developing well, it’s now time to develop a decision tree skill; this will be based on turfgrass identification to help determine the general flow of a voice decision tree.
I’ve sketched out the general technical approach, see below and now need to work this into the voice design code.
The aim is to be able to use the voice skill to fairly easily narrow down probable options as to what it is someone is looking at. The final options will give from 2 to 5 possibilities with a brief description of each.
The decision options should work out as follows:
- B1: Chewings Fescue | Sheep’s Fescue | Fine Leaved Sheep’s Fescue | Hard Fescue | Blue Fescue
- C1: No Hairs on Leaves : Strong Creeping Red Fescue | Slender CRF | Bloomed Fescue | Mat Grass | Wavy Hairgrass
- C1: Lightly Hairy Leaves : Sand Fescue | Crested Hairgrass | Fine Fescues x 2 (i.e. Festuca rubra ssp megastachys, and F.r. ssp multiflora)
- B2: Auricle: Perennial Ryegrass | Italian Ryegrass | Tall Fescue | Meadow Fescue | Couch Grass
- C2: Leaf blade upper with ribs:
- D2: Ligule fringe hairs: Heath grass | Purple Moor Grass
- E2: Ligule membranous: Spreading by:
- Stolons: Creeping Bent (x 2 sub-species) | Velvet Bent
- Rhizomes: Common Bent | Highland Bent | Black Bent | Brown Bent
- Tufted only: Crested Dogs-Tail | Tufted Hairgrass
- C2: Smooth Blade:
- D3: Sheath lightly hairy: Yorkshire Fog | Annual Vernal Grass | Sweet Vernal Grass | Soft Brome | Yellow Oat-Grass
- E3: Sheath mostly smooth: Spread by:
- Stolons: Rough Stalked Meadow Grass | Smaller Timothy Grass | Common Salt Marsh Grass
- Rhizomes: Smooth Stalked Meadow Grass |Creeping Soft-grass | Narrow Leaved Meadow Grass | Spreading Meadow Grass
- Tufted only: F3
- F3: Leaf Width (mostly):
- 2mm – 5mm: Annual Meadow Grass | Wood Meadow Grass | Reflexed Salt Marsh Grass (x 2 sub-species)
- 6mm – 14mm: Timothy Grass | Cock’s Foot | False Oat-Grass | Meadow Foxtail
A more advanced development would help narrow it down further, but for now, the aim is to understand the principles behind the coding to create a working model and a reasonable attempt after no more than 6 questions to have a reasonable probability of a correct grass. Clearly though, you will need to still refer to a suitable grass ID book, but this Skill is aiming to support learning and a learning programme.
The target date is to have this completed in May.
Once this is completed then the concept will then be applied to more complex decision processes, including 1. career options and 2. technical problem solver for turf surfaces. This will give a good basis for further developments.
Other possibilities could be:
- ‘What fertiliser should I apply?’;
- ‘What Weed is this?’;
- ‘What Pest is this?’;
- ‘What Disease is this?’;
- ‘What Flower / Plant is this?’;
- ‘How many games can we realistically play on our pitch?’;
- ‘Why is my lawn in a poor condition?’;
- ‘What qualification can I do next?’;
- or maybe some non-turf specific apps;
- …. etc.
Where my current Skills are in English, others will also be in French to start with, so future projects are already lining up. I’m just revisiting my pretty rusty French in a structured learning programme, so this will help ensure it does actually make sense to a French speaker.
Plenty of ideas, so let’s see how it goes!
Chris Gray, 27th April 2018