diff --git a/requirements.tsx b/requirements.txt
similarity index 94%
rename from requirements.tsx
rename to requirements.txt
index 71f6223..5a418a5 100644
--- a/requirements.tsx
+++ b/requirements.txt
@@ -1,7 +1,8 @@
-pandas >= 2.2.2
-seaborn >= 0.13.2
-sentence-transformers >= 3.0.1
-tensorflow >= 2.17.0
-tf_keras >= 2.17.0
-torch >= 2.3.1
-transformers >= 4.43.1
+pandas >= 2.2.2
+seaborn >= 0.13.2
+sentence-transformers >= 3.0.1
+tensorflow >= 2.17.0
+tf_keras >= 2.17.0
+torch >= 2.3.1
+transformers >= 4.43.1
+doxa-cli
\ No newline at end of file
diff --git a/submission_bag_of_words/competition.py b/submission_bag_of_words/competition.py
new file mode 100644
index 0000000..5679f03
--- /dev/null
+++ b/submission_bag_of_words/competition.py
@@ -0,0 +1,27 @@
+import os
+from typing import Any, Generator
+
+import pandas as pd
+
+
+class BaseEvaluator:
+    def predict(self, df: pd.DataFrame) -> Generator[int, Any, None]:
+        raise NotImplementedError
+
+    def run(self):
+        stream_directory = os.environ.get("DOXA_STREAMS")
+
+        in_file = f"{stream_directory}/in" if stream_directory else "train.csv"
+        out_file = f"{stream_directory}/out" if stream_directory else "predictions.csv"
+
+        with (
+            open(in_file, "r", encoding="utf8") as r,
+            open(out_file, "w") as w,
+        ):
+            w.write(f"OK\n")
+            w.flush()
+
+            df = pd.read_csv(r)
+            for prediction in self.predict(df):
+                w.write(f"{prediction}\n")
+                w.flush()
diff --git a/submission_bag_of_words/doxa.yaml b/submission_bag_of_words/doxa.yaml
new file mode 100644
index 0000000..c440078
--- /dev/null
+++ b/submission_bag_of_words/doxa.yaml
@@ -0,0 +1,4 @@
+competition: harmony-matching
+environment: cpu
+language: python
+entrypoint: run.py
diff --git a/submission_bag_of_words/run.py b/submission_bag_of_words/run.py
new file mode 100644
index 0000000..6bf9a76
--- /dev/null
+++ b/submission_bag_of_words/run.py
@@ -0,0 +1,68 @@
+import pathlib
+import sys
+from typing import Any, Generator
+
+import pandas as pd
+
+directory = pathlib.Path(__file__).parent
+sys.path.insert(0, str(directory.resolve()))
+
+from competition import BaseEvaluator
+from sentence_transformers import SentenceTransformer
+
+#################################################################################
+#                                                                               #
+#   This file gets run when you submit your work for evaluation on the DOXA     #
+#   AI platform. Modify the predict() method to implement your own strategy!    #
+#                                                                               #
+#################################################################################
+
+import re
+
+re_tok = re.compile(r"(?i)([a-z']+)")
+
+def to_bag_of_words(sentence):
+    return set(re_tok.findall(sentence.lower()))
+
+def jaccard_similarity(bag_of_words_1: set, bag_of_words_2: set):
+    tokens_in_any_sentence = len(bag_of_words_1.union(bag_of_words_2))
+    tokens_in_both_sentences = len(bag_of_words_1.intersection(bag_of_words_2))
+    if tokens_in_both_sentences == 0:
+        return 0
+    ratio = tokens_in_both_sentences / tokens_in_any_sentence
+    return ratio
+
+class Evaluator(BaseEvaluator):
+    def predict(self, df: pd.DataFrame) -> Generator[int, Any, None]:
+        """Write all the code you need to generate predictions for the test set here!
+
+        Args:
+            df (pd.DataFrame): This is a dataframe containing `sentence_1` and `sentence_`, just as in the training data
+
+        Yields:
+            Generator[int, Any, None]: For each pair of sentences in `df`, yield your similarity prediction,
+                                       which should be an integer in the range [0, 100].
+        """
+
+        #model = SentenceTransformer(str(directory / "model"))
+
+        sentences = list(set(df["sentence_1"]) | set(df["sentence_2"]))
+
+        bags_of_words = {}
+        for sentence in sentences:
+            bags_of_words[sentence] = to_bag_of_words(sentence)
+
+        bags_of_words_1 = df["sentence_1"].map(bags_of_words)
+        bags_of_words_2 = df["sentence_2"].map(bags_of_words)
+        predictions = [0] * len(df)
+        for idx in range(len(df)):
+            predictions[idx] = jaccard_similarity(bags_of_words_1.iloc[idx], bags_of_words_2.iloc[idx])
+        df["prediction"] = predictions
+        df["prediction"] = (100 * df["prediction"]).apply(int).clip(0, 100)
+
+        for _, row in df.iterrows():
+            yield row["prediction"]
+
+
+if __name__ == "__main__":
+    Evaluator().run()
diff --git a/tw_test.py b/tw_test.py
new file mode 100644
index 0000000..b797d2b
--- /dev/null
+++ b/tw_test.py
@@ -0,0 +1,9 @@
+from submission_bag_of_words.run import Evaluator
+import pandas as pd
+
+df = pd.DataFrame()
+df["sentence_1"] = ["I feel sad", "I feel happy"]
+df["sentence_2"] = ["I feel sad", "I feel sad"]
+
+for result in Evaluator().predict(df):
+    print (result)
\ No newline at end of file