Update app.py

app.py

@@ -9,13 +9,13 @@ nlp = spacy.load('en_core_web_lg')
 
 
 # Define earthquake-related keywords
-earthquake_single_keywords = ['earthquake', 'seismic', 'tremor', 'quake', 'aftershock', 'seismology', 'tectonic', 'plate', 'seismometer', 'shake', 'temblor', 'trembler', 'seism', 'shock', 'vibration', 'groundswell']
+earthquake_single_keywords = ['earthquake', 'seismic', 'tremor', 'quake', 'aftershock', 'seismology', 'tectonic', 'plate', 'seismometer', 'temblor', 'trembler', 'seism', 'shock', 'vibration', 'shake', 'groundswell', 'earthquakes', 'seismics', 'tremors', 'quakes', 'aftershocks', 'seismologies', 'tectonics', 'plates', 'seismometers', 'temblors', 'tremblers', 'seisms', 'shocks', 'vibrations', 'shakes', 'groundswells']
 
 # Compute embeddings for single-word keywords
 earthquake_single_embeddings = [nlp(keyword).vector for keyword in earthquake_single_keywords]
 
 # Define multi-word earthquake-related keywords
-earthquake_multi_keywords = ['seismic activity', 'earthquake risk', 'earthquake zone', 'seismic waves', 'earthquake damage', 'seismic shift', 'tectonic plates', 'fault line', 'seismic retrofitting', 'seismic hazard', 'aftershock sequence', 'earthquake drill', 'seismic reflection', 'plate tectonics', 'seismic reflection imaging', 'seismic tomography', 'seismic profiling', 'seismic energy release', 'seismicity pattern', 'earthquake swarm', 'seismic gap', 'seismic inversion', 'seismic reflection', 'seismic scattering', 'seismic attenuation', 'seismic imaging', 'seismic map', 'seismic data', 'earthquake monitoring', 'seismic data analysis', 'earth shaking']
+earthquake_multi_keywords = ['seismic activity', 'earthquake risk', 'earthquake zone', 'seismic wave', 'earthquake damage', 'seismic shift', 'tectonic plate', 'fault line', 'seismic retrofitting', 'seismic hazard', 'aftershock sequence', 'earthquake drill', 'seismic reflection', 'plate tectonic', 'seismic tomography', 'seismic profiling', 'seismicity pattern', 'earthquake swarm', 'seismic gap', 'seismic inversion', 'seismic reflection', 'seismic scattering', 'seismic attenuation', 'seismic imaging', 'seismic map', 'seismic data', 'earthquake monitoring', 'earth shaking', 'seismic activities', 'earthquake risks', 'earthquake zones', 'seismic waves', 'earthquake damages', 'seismic shifts', 'tectonic plates', 'fault lines', 'seismic retrofittings', 'seismic hazards', 'aftershock sequences', 'earthquake drills', 'seismic reflections', 'plate tectonics', 'seismic tomographies', 'seismic profilings', 'seismicity patterns', 'earthquake swarms', 'seismic gaps', 'seismic inversions', 'seismic reflections', 'seismic scatterings', 'seismic attenuations', 'seismic imagings', 'seismic maps', 'earth shakings']
 
 # Compute embeddings for multi-word keywords
 earthquake_multi_embeddings = []
@@ -25,6 +25,17 @@ for keyword in earthquake_multi_keywords:
     earthquake_multi_embeddings.append(combined_emb)
 
 
+def straight_pattern_matching(ngram):
+  """
+  Function to compute a straightforward similarity between a word and the pre-defined references
+  """
+
+  if ngram in earthquake_single_keywords or ngram in earthquake_multi_keywords:
+    return ngram
+  else:
+    return False
+
+
 # Define a function to compute the semantic similarity between a word and a set of embeddings
 def compute_similarity_earthquake(word, embeddings, excluded_keywords):
     """
@@ -41,8 +52,12 @@ def compute_similarity_earthquake(word, embeddings, excluded_keywords):
     # Compute the cosine similarity between the word embedding and the keyword embeddings
     similarity_scores = [np.dot(word_emb, emb) / (np.linalg.norm(word_emb) * np.linalg.norm(emb)) for emb in embeddings]
 
-    # Return True if the maximum similarity score is above a certain threshold, else False
-    return max(similarity_scores) > 0.65
+    # Return if the maximum similarity score is above a certain threshold
+    if max(similarity_scores) > 0.65:
+      return word
+
+    else:
+      return False
 
 
 def identify_earthquake_event(input_sentence):
@@ -50,44 +65,94 @@ def identify_earthquake_event(input_sentence):
   Compute the semantic similarity for earthquaqe events
   """
 
-  # Define excluded keywords to ignore (because cases like I want bars with magnituted 6 - were given as correct)
-  excluded_keywords = ['magnitude', 'richter', 'moment', 'scale', 'intensity', 'amplitude', 'energy', 'force', 'power', 'seismicity']
+  try:
+
+    # Define excluded keywords to ignore (because cases like I want bars with magnituted 6 - were given as correct)
+    excluded_keywords = ['magnitude', 'richter', 'moment', 'scale', 'intensity', 'amplitude', 'energy', 'force', 'power', 'seismicity']
+
+    parsed_sentence = nlp(input_sentence)
+
+    # start with simple straight pattern matching of single keywords
+    for word in parsed_sentence:
+      if word.text not in excluded_keywords:
+
+        straight_matching_single = straight_pattern_matching(word.text)
+
+        if straight_matching_single:
+          return {'earthquaqe_event': [True, straight_matching_single]}
+    
+    # Continue with embeddings single matching
+    earthquaqe_keywords_single = []
+
+    # Check for single-word earthquake-related keywords
+    earthquaqe_keywords_single = [compute_similarity_earthquake(word.text.lower(), earthquake_single_embeddings, excluded_keywords) for word in parsed_sentence] 
+
+    single_keyword_flag = False
+
+    # check until you find one such reference and then break
+    for elem in earthquaqe_keywords_single:
+      if elem:
+        single_keyword_flag = True
+        target_elem_single = elem
+        break
+
+    # if there is at least one referece, we can assume that the sentence refers to earthquaqe events
+    if single_keyword_flag:
+      return {'earthquaqe_event': [True, target_elem_single]}
+
+    # otherwise we examine for 2grams multi-word straight patterns and embeddings
+
+    earthquaqe_keywords_multi = []
+
+    # check 2-grams
+    for i in range(len(parsed_sentence)-1):
+      bigram = parsed_sentence[i:i+2].text.lower()
+
+      # case of straight matching
+      straight_matching_multi = straight_pattern_matching(word.text)
+
+      if straight_matching_multi:
+          return {'earthquaqe_event': [True, straight_matching_multi]}
+
+      # if no straight matching then perform embeddings
+      else:
+        earthquaqe_keywords_multi.append(compute_similarity_earthquake(bigram, earthquake_multi_embeddings, excluded_keywords))
+    
+
+    # case that the straight multi matching did not give any output
+    multi_keyword_flag = False
+
+    # check until you find one such reference and then break
+    for elem in earthquaqe_keywords_multi:
+      if elem:
+        multi_keyword_flag = True
+        target_elem_multi = elem
+        break
 
-  # Check for single-word earthquake-related keywords
-  is_earthquake_related = any([compute_similarity_earthquake(word.text.lower(), earthquake_single_embeddings, excluded_keywords) for word in nlp(input_sentence)])
+    # if there is at least one referece, we can assume that the sentence refers to earthquaqe events
+    if multi_keyword_flag:
+      return {'earthquaqe_event': [True, target_elem_multi]}
 
-  # If no single-word keywords are found, check for multi-word keywords
-  if not is_earthquake_related:
+    # otherwise there is no reference
+    else:
+      return (0,'EARTHQUAQE_EVENT','no_earthquaqe_reference')
 
-      # Check 2-grams
-      for i in range(len(nlp(input_sentence))-1):
-          bigram = nlp(input_sentence)[i:i+2].text.lower()
-          if compute_similarity_earthquake(bigram, earthquake_multi_embeddings, excluded_keywords):
-              is_earthquake_related = True
-              break
-      
-      # Check 3-grams
-      if not is_earthquake_related:
-          for i in range(len(nlp(input_sentence))-2):
-              trigram = nlp(input_sentence)[i:i+3].text.lower()
-              if compute_similarity_earthquake(trigram, earthquake_multi_embeddings, excluded_keywords):
-                  is_earthquake_related = True
-                  break
+  except:
+    return (0,'EARTHQUAQE_EVENT','unknown_error')
 
-  return {'earthquaqe_event': is_earthquake_related}
 
 
 from transformers import pipeline
 import gradio as gr
 
-title = "Natural Language module Demo for Earthquaqe events identification"
-description = "This is a simple demo just for demonstration purposes, so that Serco team might have the chance to validate the results of the Natural Language module concerning the earthquaqe event identification, while in progress"
+title = "Earthquaqe events Demo"
+description = "This is a simple demo just for demonstration purposes for Serco team, to validate the results of the Natural Language module concerning the identification of earthquaqe events, while in progress"
 
 examples = [
-    ["I want all earthquakes larger than 5.0 that occurred in Rome during 3/5/20"],
-    ["I want all seism larger than 5.0 that occurred in Rome between January 2020 and February of the same year"],
-    ["earth shakes located in Ishkoshim"],
-    ["give me all the bars with magnitude above than 6 in the region of Athens for the month of January, 1990"]
+    ["I want all earthquakes that are located in Italy, in 01/01/23 with magnitude greater than 6.2"],
+    ["I want all seisms that are located in Italy, in 01/01/23 with magnitude greater than 6.2"],
+    ["I want all earth shakes that are located in Italy, in 01/01/23 with magnitude greater than 6.2"],
+    ["I want all bars that are located in Italy, in 01/01/23 with magnitude greater than 6.2"],
 ]