Update to latest version of DialogueKit and Python 3.11

.github/workflows/build.yaml

@@ -23,7 +23,7 @@ jobs:
 
       - uses: actions/setup-python@v4
         with:
-          python-version: "3.9"
+          python-version: "3.11"
           cache: "pip"
 
       - name: Install Dependencies
@@ -54,7 +54,7 @@ jobs:
 
       - uses: actions/setup-python@v4
         with:
-          python-version: "3.9"
+          python-version: "3.11"
           cache: "pip"
 
       - name: Install Dependencies
@@ -66,7 +66,7 @@ jobs:
         run: pre-commit run mypy --all-file
 
   build-and-test:
-    name: "Build and Test Python 3.9"
+    name: "Build and Test Python 3.11"
     runs-on: ubuntu-latest
     if: always()
     timeout-minutes: 20
@@ -76,7 +76,7 @@ jobs:
       - name: Setup python
         uses: actions/setup-python@v4
         with:
-          python-version: "3.9"
+          python-version: "3.11"
           cache: "pip"
 
       - name: Install Dependencies

.github/workflows/build_docs.yaml

@@ -20,7 +20,7 @@ jobs:
       - name: Setup Python
         uses: actions/setup-python@v4
         with:
-          python-version: 3.9.15
+          python-version: 3.11
           cache: "pip"
       - name: Installing the Documentation requirements
         run: |

.github/workflows/merge.yaml

@@ -6,7 +6,7 @@ on:
 
 jobs:
   build-and-test:
-    name: "Build and Test Python 3.9"
+    name: "Build and Test Python 3.11"
     runs-on: ubuntu-latest
     if: always()
     timeout-minutes: 20
@@ -16,7 +16,7 @@ jobs:
       - name: Setup python
         uses: actions/setup-python@v4
         with:
-          python-version: "3.9"
+          python-version: "3.11"
           cache: "pip"
 
       - name: Install Dependencies

.pre-commit-config.yaml

@@ -3,7 +3,7 @@ repos:
     rev: 22.6.0
     hooks:
     - id: black
-      language_version: python3.9
+      language_version: python3.11
 -   repo: https://github.com/pycqa/flake8
     rev: 5.0.4
     hooks:

README.md

@@ -4,7 +4,7 @@
 ![docs](https://img.shields.io/github/actions/workflow/status/iai-group/UserSimCRS/build_docs.yaml?label=docs&branch=main)
 ![Tests](https://img.shields.io/github/actions/workflow/status/iai-group/UserSimCRS/merge.yaml?label=Tests&branch=main)
 ![Coverage Badge](https://img.shields.io/endpoint?url=https://gist.githubusercontent.com/NoB0/cd558f4b76df656b67277f8ae214b7e0/raw/coverage.UserSimCRS.main.json)
-![Python version](https://img.shields.io/badge/python-3.9-blue)
+![Python version](https://img.shields.io/badge/python-3.11-blue)
 
 UserSimCRS is an extensible user simulation toolkit for evaluating conversational recommender systems.  
 
@@ -29,7 +29,7 @@ We refer to the [documentation](https://iai-group.github.io/UserSimCRS/main) for
 
 ## Installation
 
-The recommended version of Python is 3.9.  
+UserSimCRS requires Python 3.11+.
 The easiest way to install UserSimCRS and all of its dependencies is by using pip:
 
 ```shell

config/default/config_default.yaml

@@ -36,6 +36,4 @@ historical_ratings_ratio: 0.8
 
 dialogues: data/datasets/moviebot/annotated_dialogues.json
 intent_classifier: "cosine"
-# If using the DIET classifier the following file needs to be provided. 
-# rasa_dialogues: data/agents/moviebot/annotated_dialogues_rasa_agent.yml
 

data/datasets/README.md

@@ -1,16 +1,10 @@
 # Dialogue files
 
-The YAML files required to train the Rasa DIET classifier can be generated from the annotated dialogues saved in the correct format.
-The generation of these files can be done with this command:
+This folder contains dialogue files used for configuring a user simulator. For each dataset supported, annotated dialogues in JSON format are provided, along with prompts for LLM-based NLU and NLG components.
 
-```shell
-  cd usersimcrs/utils
-  python -m annotation_converter_rasa -source PathToAnnotatedDialoguesFile -destination PathToDestinationFolder
-```
+The datasets currently supported are:
 
-It creates the following files:
-
-  - `<originalname>_reformat.yaml`: The original file saved as a yaml file
-  - `<originalname>_types_w_examples.yaml`: Slots and example values extracted from the dialogues
-  - `<originalname>_rasa_agent.yaml`: Examples of agent utterances for all possible intents/actions that the agent can take
-  - `<originalname>_rasa_user.yaml`: Similar to the agent file, but for users
+  * MovieBot: dialogues related to movie recommendations obtained with IAI MovieBot v1.
+  * [ReDial](https://redialdata.github.io/website/): dialogues related to movie recommendations obtained with crowdsourcing.
+  * [IARD](https://github.com/wanlingcai1997/umap_2020_IARD): subset of ReDial dialogues related to movie recommendations.
+  * [INSPIRED](https://github.com/sweetpeach/Inspired): dialogues related to movie recommendations obtained with crowdsourcing.

docs/source/configuration.rst

@@ -33,7 +33,6 @@ Simulator parameters
   * `historical_ratings_ratio`: Ratio ([0..1]) of ratings to be used as historical data.
   * `dialogues`: Path to domain config file.
   * `intent_classifier`: Intent classifier model to be used. Only supports DialogueKit intent classifiers.
-  * `rasa_dialogues`: File with Rasa annotated dialogues. Only needed when using a DIET intent classifier.
   * `debug`: Flag (boolean) to activate debug mode.
 
 Configuration example
@@ -71,8 +70,6 @@ Below is the default configuration to run simulation with the IAI MovieBot as th
 
   dialogues: data/agents/moviebot/annotated_dialogues.json
   intent_classifier: "cosine"
-  # If using the DIET classifier the following file needs to be provided. 
-  # rasa_dialogues: data/agents/moviebot/annotated_dialogues_rasa_agent.yml
 
   debug: False
 

docs/source/installation.rst

@@ -1,7 +1,7 @@
 Installation
 ============
 
-The recommended version of Python is 3.9.  
+UserSimCRS requires Python 3.11+.  
 The easiest way to install UserSimCRS and all of its dependencies is by using pip:
 
 .. code-block:: shell

requirements/requirements.txt

@@ -17,5 +17,6 @@ torch==2.0.1
 openai==1.30.5
 ollama==0.3.1
 pandas==2.2.2
-git+https://github.com/iai-group/DialogueKit.git
+dialoguekit==0.1.1
+scikit-learn >= 0.24
 setuptools<=70.0.0

usersimcrs/core/information_need.py

@@ -41,14 +41,14 @@ def generate_random_information_need(
         target_item.properties.keys()
     )
     num_constraints = random.randint(1, len(informable_slots))
-    for slot in random.sample(informable_slots, num_constraints):
+    for slot in random.sample(list(informable_slots), num_constraints):
         constraints[slot] = target_item.get_property(slot)
 
     requestable_slots = set(
         domain.get_requestable_slots()
     ).symmetric_difference(constraints.keys())
     num_requests = random.randint(1, len(requestable_slots))
-    requests = random.sample(requestable_slots, num_requests)
+    requests = random.sample(list(requestable_slots), num_requests)
 
     return InformationNeed([target_item], constraints, requests)
 

usersimcrs/items/ratings.py

@@ -31,7 +31,7 @@ def user_item_sampler(
     """
     # Determine the number of items to use as historical data for a given user.
     nb_historical_items = int(historical_ratio * len(item_ratings))
-    return random.sample(item_ratings.keys(), nb_historical_items)
+    return random.sample(list(item_ratings.keys()), nb_historical_items)
 
 
 class Ratings:

usersimcrs/run_simulation.py

@@ -122,11 +122,6 @@ def parse_args() -> argparse.Namespace:
         choices=["cosine", "diet"],
         help="Intent classifier model to be used. Defaults to cosine.",
     )
-    parser.add_argument(
-        "--rasa_dialogues",
-        type=str,
-        help="Path to the Rasa annotated dialogues file.",
-    )
     parser.add_argument(
         "-d",
         "--debug",

usersimcrs/utils/simulation_utils.py

@@ -1,10 +1,9 @@
 """Utility functions to run the simulation."""
 
 import json
-from typing import Any, Dict, Set, Tuple, Type
+from typing import Any, Dict, Tuple, Type
 
 import confuse
-import yaml
 from dialoguekit.core.intent import Intent
 from dialoguekit.core.utterance import Utterance
 from dialoguekit.nlg import ConditionalNLG
@@ -16,7 +15,6 @@
     DisjointDialogueActExtractor,
 )
 from dialoguekit.nlu.intent_classifier import IntentClassifier
-from dialoguekit.nlu.models.diet_classifier_rasa import IntentClassifierRasa
 from dialoguekit.nlu.models.intent_classifier_cosine import (
     IntentClassifierCosine,
 )
@@ -131,7 +129,7 @@ def _get_agenda_based_simulator_config(
 
     ratings = Ratings(item_collection)
     ratings.load_ratings_csv(file_path=config["ratings"].get())
-    historical_ratings, ground_truth_ratings = ratings.create_split(
+    historical_ratings, _ = ratings.create_split(
         config["historical_ratings_ratio"].get(0.8)
     )
 
@@ -200,9 +198,6 @@ def get_NLU(config: confuse.Configuration) -> NLU:
         return NLU(
             DisjointDialogueActExtractor(classifier, slot_value_annotators=[])
         )
-    elif intent_classifier == "diet":
-        classifier = train_rasa_diet_classifier(config)
-        return NLU(DisjointDialogueActExtractor(classifier, [classifier]))
     raise ValueError(
         "Unsupported intent classifier. Check DialogueKit intent"
         " classifiers."
@@ -242,35 +237,3 @@ def train_cosine_classifier(
     intent_classifier = IntentClassifierCosine(intents=gt_intents)
     intent_classifier.train_model(utterances=utterances, labels=gt_intents)
     return intent_classifier
-
-
-def train_rasa_diet_classifier(
-    config: confuse.Configuration,
-) -> IntentClassifierRasa:
-    """Trains a DIET classifier on Rasa annotated dialogues for NLU module.
-
-    Args:
-        config: Configuration generated from YAML configuration file.
-
-    Returns:
-        A trained Rasa DIET model for intent classification.
-    """
-    # TODO: Move to DialogueKit as util function.
-    # See: https://github.com/iai-group/UserSimCRS/issues/92
-    intent_schema_file = config["intents"].get()
-    intent_schema = yaml.load(open(intent_schema_file), Loader=yaml.FullLoader)
-
-    agent_intents_str: Set[str] = set()
-    for v in intent_schema["user_intents"].values():
-        intents = v.get("expected_agent_intents", []) or []
-        agent_intents_str.update(intents)
-    # agent_intents_str = intent_schema["agent_elicit_intents"]
-    # agent_intents_str.extend(intent_schema["agent_set_retrieval"])
-    agent_intents = [Intent(intent) for intent in agent_intents_str]
-    intent_classifier = IntentClassifierRasa(
-        agent_intents,
-        config["rasa_dialogues"].get(),
-        ".rasa",
-    )
-    intent_classifier.train_model()
-    return intent_classifier