Merge pull request #7 from Canadian-Geospatial-Platform/dev

v1-1-1-release

aws-lambda/Preprocessing_lambda/Dockerfile

@@ -0,0 +1,24 @@
+# Use a base image provided by AWS for Python 3.9 runtime 
+FROM public.ecr.aws/lambda/python:3.9-x86_64
+
+# Set the working directory
+WORKDIR /var/task
+
+# Copy requirements.txt
+COPY requirements.txt ./
+
+# Copy function code
+COPY app.py ./
+
+# Install the specified packages
+RUN pip install -r requirements.txt
+
+# Download the NLTK 'punkt' tokenizer and 'stopwords' data
+RUN python -m nltk.downloader -d /var/task/nltk_data punkt stopwords
+
+# Set the NLTK_DATA environment variable to use the included data
+ENV NLTK_DATA=/var/task/nltk_data
+
+# Set the CMD to your handler (could also be done as a parameter override outside of the Dockerfile)
+CMD [ "app.lambda_handler" ]
+

aws-lambda/Preprocessing_lambda/app.py

@@ -8,31 +8,39 @@
 import nltk
 from nltk.corpus import stopwords          # module for stop words that come with NLTK
 from nltk.stem import PorterStemmer        # module for stemming
-from nltk.tokenize import word_tokenize   # module for tokenizing strings       
-nltk.download('punkt')
-nltk.download('stopwords')
+from nltk.tokenize import word_tokenize   # module for tokenizing strings  
+
+#nltk.download('punkt')
+#nltk.download('stopwords')
 
 import re
 
-# Environment variables
+# environment variables for lambda
+file_name = os.environ['FILE_NAME']
+bucket_name = os.environ['BUCKET_NAME']
+bucket_name_nlp = os.environ['BUCKET_NAME_NLP']
+
+"""
+#dev setting  -- comment out for release
 file_name = "records.parquet"
 bucket_name = "webpresence-geocore-geojson-to-parquet-dev"
 bucket_name_nlp='nlp-data-preprocessing'
-
-selected_var = ['features_properties_id', 'features_properties_title_en', 'features_properties_title_en','features_properties_description_en','features_properties_keywords_en']
+"""
 
 def lambda_handler(event, context):
+
     #Change directory to /tmp folder
     os.chdir('/tmp')    #This is important
+    """
     #Make a directory
     if not os.path.exists(os.path.join('mydir')):
         os.makedirs('mydir')
-        
+    """  
     df = open_S3_file_as_df(bucket_name, file_name)
     print(f'The shape of the raw metadata parquet dataset is {df.shape}')
 
     # Select key columns, currently only english
-    df_en = df[selected_var]
+    df_en = df[['features_properties_id', 'features_properties_title_en','features_properties_title_fr','features_properties_description_en','features_properties_keywords_en']]
     # Replace NaN and "Not Available; Indisponible" with empty string 
     print("The NaN values in the English columns are \n") 
     df_en = df_en.fillna('')
@@ -58,7 +66,7 @@ def lambda_handler(event, context):
     #print(duplicateRowsDF['features_properties_id'].unique())
 
     # Save to temp  folder, see https://iotespresso.com/temporary-storage-during-aws-lambda-runtime-python/
-    save_path = os.path.join(os.getcwd(), 'mydir', 'duplicateRowsDF')
+    save_path = os.path.join(os.getcwd(), 'duplicateRowsDF')
     duplicateRowsDF.to_csv(save_path)
     df_fetched= pd.read_csv(save_path)
 

aws-lambda/Preprocessing_lambda/requirements.txt

@@ -2,4 +2,5 @@ pandas
 nltk
 pyarrow
 fastparquet
-requests
+requests
+urllib3<2

aws-lambda/Preprocessing_lambda/zip-deployment.sh

@@ -0,0 +1,17 @@
+#!/bin/bash
+# Build Cloud Formation Zip Package for Similarity Engine using Data Preprocessing
+# Author: Xinli Cai 
+# Date: August 29, 2023 
+
+
+# Navigate to the directory containing `requirements.txt`
+cd similarity-engine-data-process/
+
+# Install required packages
+pip install -t similarity-engine-data-preprocess-20230822-2200/ -r requirements.txt
+
+# Change to the build directory
+cd similarity-engine-data-preprocess-20230822-2200/
+
+# Zip the necessary files
+zip -r similarity-engine-data-preprocess-20230822-2200.zip ../app.py ../__init.py__ ./*

aws-lambda/Word2Vec_lambda/Dockerfile

@@ -0,0 +1,19 @@
+# Use a base image provided by AWS for Python 3.9 runtime 
+FROM public.ecr.aws/lambda/python:3.9-x86_64
+
+# Set the working directory
+WORKDIR /var/task
+
+# Copy requirements.txt
+COPY requirements.txt ./
+
+# Copy function code
+COPY app.py ./
+COPY dynamodb.py ./
+
+# Install the specified packages
+RUN pip install -r requirements.txt
+
+# Set the CMD to your handler (could also be done as a parameter override outside of the Dockerfile)
+CMD [ "app.lambda_handler" ]
+

aws-lambda/Word2Vec_lambda/app.py

@@ -1,52 +1,54 @@
-
 import boto3
 import logging 
 from botocore.exceptions import ClientError
 import pandas as pd 
-
-
 import numpy as np
 import json
 import datetime
-
 import io
 import os 
 
 from gensim.models import Word2Vec
 from sklearn.metrics.pairwise import cosine_similarity
 from gensim import matutils
-
-from tqdm import tqdm
 from dynamodb import * 
 
 
-#dev setting 
+# environment variables for lambda
+file_name = os.environ['FILE_NAME']
+file_name_origianl = os.environ['FILE_NAME_ORIGINAL']
+bucket_name_nlp = os.environ['BUCKET_NAME_NLP']
+bucket_name = os.environ['BUCKET_NAME']
+
+"""
+#dev setting  -- comment out for release
 file_name = "Processed_records.parquet"
 bucket_name_nlp = "nlp-data-preprocessing"
 file_name_origianl = "records.parquet"
 bucket_name = "webpresence-geocore-geojson-to-parquet-dev"
+"""
 
 def lambda_handler(event, context):
+    """
     #Change directory to /tmp folder, this is required if new files are created for lambda 
     os.chdir('/tmp')    #This is important
     #Make a directory
     if not os.path.exists(os.path.join('mydir')):
         os.makedirs('mydir')
-
+    """    
     # Read the preprocessed data from S3 
-    df_en = open_S3_file_as_df(bucket_name_nlp, file_name)
+    try:
+        df_en = open_S3_file_as_df(bucket_name_nlp, file_name)
+    except ClientError as e:
+        print('Accessing the S3 was failed on line 47 when calling  df_en = open_S3_file_as_df(bucket_name_nlp, file_name)')
+        print(e.response['Error']['Message'])
 
-    # Get a sample of 500 rows as the training data 
-    df = df_en[['features_properties_id', 'features_properties_title_en', 'metadata_en_processed']]
-    #df = df.sample(n=500, random_state=1)
-    # Use all data to train the model
-    df.head()
-    print(df.shape)
-
-
     # Use all data to train the model
     df = df_en[['features_properties_id', 'features_properties_title_en', 'features_properties_title_fr','metadata_en_processed']]
     print(f'The shape of the preprocessed df is {df.shape}')
+        # Replace the missing value in the 'features_properties_title_en' column with an empty string
+    df['features_properties_title_en'].fillna('', inplace=True)
+
 
     # Prepare the input for the Word2Vec model
     sentences = df['metadata_en_processed'].apply(lambda x: x.split(' ')).tolist()
@@ -58,55 +60,21 @@ def lambda_handler(event, context):
 
     # Convert each sentence in 'metadata_preprocessed' into a vector
     vectors = df['metadata_en_processed'].apply(sentence_to_vector, model=model)
-    # Replace the missing value in the 'features_properties_title_en' column with an empty string
-    df['features_properties_title_en'].fillna('', inplace=True)
+
 
     # Calculate similarity between each vector and all others
     similarity_matrix = cosine_similarity(np.array(vectors.tolist()))
 
-
-    # Initialize new columns for the top 5 similar texts
-    df['sim1'], df['sim2'], df['sim3'], df['sim4'], df['sim5'] = "", "", "", "", ""
-
-    # For each text, find the top 5 most similar texts and append their 'features_properties_title_en' as new columns
-    df.reset_index(drop=True, inplace=True)
-    for i in tqdm(range(similarity_matrix.shape[0])):
-        top_5_similar = np.argsort(-similarity_matrix[i, :])[1:6]  # Exclude the text itself
-        df.loc[i, ['sim1', 'sim2', 'sim3', 'sim4', 'sim5']] = df.loc[top_5_similar, 'features_properties_title_en'].values
-
-    # Read the original parquet file and merge by features_properties_id
-    df_original = open_S3_file_as_df(bucket_name, file_name_origianl)
-    merged_df = df_original.merge(df[['features_properties_id', 'sim1', 'sim2', 'sim3', 'sim4', 'sim5']], on='features_properties_id', how='left')
-    """ Option 1: merge the similar results with records.parquet directly 
-    # Initialize new columns for the top 10 similar texts
-    df['sim1'], df['sim2'], df['sim3'], df['sim4'], df['sim5'],df['sim6'], df['sim7'], df['sim8'], df['sim9'], df['sim10'] = "", "", "", "", "","", "", "", "", ""
 
-    # For each text, find the top 10 most similar texts and append their 'features_properties_title_en' as new columns
-    df.reset_index(drop=True, inplace=True)
-    for i in tqdm(range(similarity_matrix.shape[0])):
-        top_10_similar = np.argsort(-similarity_matrix[i, :])[1:11]  # Exclude the text itself
-        df.loc[i, ['sim1', 'sim2', 'sim3', 'sim4', 'sim5','sim6', 'sim7', 'sim8', 'sim9', 'sim10']] = df.loc[top_10_similar, 'features_properties_id'].values
-
-    # Read the original parquet file and merge by features_properties_id
-    df_original = open_S3_file_as_df(bucket_name, file_name_origianl)
-    merged_df = df_original.merge(df[['features_properties_id', 'sim1', 'sim2', 'sim3', 'sim4', 'sim5','sim6', 'sim7', 'sim8', 'sim9', 'sim10']], 
-    on='features_properties_id', how='left')
-    print(f'the shape of original parquet file is {df_original.shape}')
-    
-    # Save to temp  folder, see https://iotespresso.com/temporary-storage-during-aws-lambda-runtime-python/
-    save_path = os.path.join(os.getcwd(), 'mydir', 'merged_df')
-    merged_df.to_csv(save_path)
-    df_fetched= pd.read_csv(save_path)
-    print(f'the shape of merged parquet file is {merged_df.shape}')
-    # upload merged dataframe to S3
-    upload_dataframe_to_s3_as_parquet(df=df_fetched, bucket_name=bucket_name_nlp, file_key='sim_word2vec_records.parquet')
-    """ 
+    # Upload the similar results as a AWS dynamodb 
+    """
+    The parquet lambda function has been modified to merge the similairy table with records.parquet everytime when records.parquet is updated. 
     
-    #Option 2: upload the similar results as a dynamodb, and merge the tabke with records.parquet everytime when records.parquet is updated 
+    """
     df['similarity'] = np.nan  # Initialize the column
     # For each text, find the top 10 most similar texts and save them as a JSON array object in the 'similarity' column
     df.reset_index(drop=True, inplace=True)
-    for i in tqdm(range(similarity_matrix.shape[0])):
+    for i in range(similarity_matrix.shape[0]):
         top_10_similar = np.argsort(-similarity_matrix[i, :])[1:11]  # Exclude the text itself
         sim_array = []
         for j, idx in enumerate(top_10_similar):
@@ -137,7 +105,6 @@ def lambda_handler(event, context):
             delete_table(TableName='similarity')
             waiter = client.get_waiter('table_not_exists')
             waiter.wait(TableName='similarity')
-            print('Before create')
         except ClientError as e:
             print(e)
     #Create table         
@@ -148,8 +115,20 @@ def lambda_handler(event, context):
         waiter.wait(TableName='similarity')
     except ClientError as e:
             print(e)  
+
+    """DEBUG
+    #Check if empty string in the primary key before scan the table 
+    empty_string_rows = df[df['features_properties_id'] == '']
+    print(f'Number of NA values in the df id column is \n {empty_string_rows}')
+    """
+    #Remove rows with empty string in 'features_properties_id', primary key can not be empty in DynamoDB table 
+    df_cleaned = df[df['features_properties_id']!='']
+    rows_removed = df.shape[0] - df_cleaned.shape[0]
+    print(f'Removed {rows_removed} rows with empyt string in features_properties_id')
     #Batch write to table 
-    batch_write_items_into_table(df, TableName='similarity')
+    batch_write_items_into_table(df_cleaned, TableName='similarity')
+
+
 
 # Function to read the parquet file as pandas dataframe 
 def open_S3_file_as_df(bucket_name, file_name):

aws-lambda/Word2Vec_lambda/requirements.txt

@@ -1,5 +1,7 @@
-gensim==4.1.2
-scikit-learn==1.0
-pyarrow
-fastparquet
-tqdm
+pandas
+numpy
+scikit-learn
+transformers
+torch
+pyarrow 
+fastparquet 

aws-lambda/Word2Vec_lambda/zip-deployment.sh

@@ -0,0 +1,22 @@
+#!/bin/bash
+# Build Cloud Formation Package for Similarity Engine using Word2Vec
+# Author: Xinli Cai 
+# Date: August 29, 2023 
+
+
+# Cloud Formation Package for Similarity Engine using Word2Vec
+
+# Step 1: Navigate to the directory where the `requirements.txt` is
+cd similarity-engine-word2vec-model-dev/
+
+# Step 2: Install the required packages
+pip install -t similarity-engine-word2vec-model-build/ -r requirements.txt
+
+# Step 3: Change to the build directory
+cd similarity-engine-word2vec-model-build/
+
+# Step 4: Zip the necessary files
+zip -r similarity-engine-word2vec-model.zip ../app.py ../dynamodb.py ../__init.py__ ./*
+
+# Restore to the initial directory
+cd ..