Updated readme of demo

examples/die_vs_data_rest_api/README.md

@@ -9,10 +9,11 @@ By default, A Flask server will listen to port 5000. The endpoint is `/`.
 First install the dependencies from the requirements.txt file using `pip install -r requirements.txt`
 
 ```shell script
-$ python app.py
+$ python app.py --model-path DTAI-KULeuven/robbertje-shuffled-dutch-die-vs-dat --fast-model-path pdelobelle/robbert-v2-dutch-base
 ```
 
-And then make a http POST request to `/` with the parameter `sentence` filled in:
+## Classification model
+Simply make a http POST request to `/` with the parameter `sentence` filled in:
 
 ```shell script
 $ curl --data "sentence=Daar loopt _die_ meisje." localhost:5000
@@ -22,9 +23,29 @@ This should give you the following response:
 
 ```json
 {
- "rating": 1, 
- "interpretation": "incorrect", 
- "confidence": 5.222124099731445, 
- "sentence": "Daar loopt _die_ meisje."
+    "rating": 1, 
+    "interpretation": "incorrect", 
+    "confidence": 5.222124099731445, 
+    "sentence": "Daar loopt _die_ meisje."
 }
 ```
+
+## Zero-shot model
+We also have a faster zero-shot model (using RobBERT base), which might be faster and easier to use. There is a small drop in accuracy, but that should be quite limited.
+
+To use the faster zero-shot model, just make a http POST request to `/fast` with the same parameter `sentence` filled in:
+
+```shell script
+$ curl --data "sentence=Daar loopt _die_ meisje." localhost:5000/fast
+```
+
+This should give you the following response, which is similar, but also provides `die`, `dat`, `Die` or `Dat` as the rating value:
+
+```json
+{
+    "rating": "dat", 
+    "interpretation": "incorrect", 
+    "confidence": 2.567270278930664, 
+    "sentence": "Daar loopt _die_ meisie."
+}                       
+```