app.py

from transformers import AutoModelForMaskedLM , AutoModelForSequenceClassification
from transformers import AutoTokenizer
from sklearn.metrics.pairwise import cosine_similarity
import streamlit as st
import torch
import pickle
import numpy as np
import itertools

model = AutoModelForMaskedLM.from_pretrained("vives/distilbert-base-uncased-finetuned-cvent-2019_2022", output_hidden_states=True)
tokenizer = AutoTokenizer.from_pretrained("vives/distilbert-base-uncased-finetuned-cvent-2019_2022")
kp_dict_checkpoint = "kp_dict_merged.pickle"
kp_cosine_checkpoint = "cosine_kp.pickle"
 
model_finbert = AutoModelForSequenceClassification.from_pretrained("ProsusAI/finbert", output_hidden_states=True)
tokenizer_finbert = AutoTokenizer.from_pretrained("ProsusAI/finbert")
kp_dict_finbert_checkpoint = "kp_dict_finbert.pickle"
kp_cosine_finbert_checkpoint = "cosine_kp_finbert.pickle"

text = st.text_input("Enter word or key-phrase")
exclude_words = st.radio("exclude_words",[True,False], help="Exclude results that contain any words in the query")
exclude_text = st.radio("exclude_text",[True,False], help="Exclude results that contain the query (i.e exclude 'tomato soup recipe' if the query is 'tomato soup')")
k = st.number_input("Top k nearest key-phrases",1,10,5)

with st.sidebar:
  diversify_box = st.checkbox("Diversify results",True)
  if diversify_box:
    k_diversify = st.number_input("Set of key-phrases to diversify from",10,30,20)

#columns
col1, col2 = st.columns(2)
#load kp dicts
with open(kp_dict_checkpoint,'rb') as handle:
  kp_dict = pickle.load(handle)
keys = list(kp_dict.keys())

with open(kp_dict_finbert_checkpoint,'rb') as handle:
  kp_dict_finbert = pickle.load(handle)
keys_finbert = list(kp_dict_finbert.keys())

#load cosine distances of kp dict
with open(kp_cosine_checkpoint,'rb') as handle:
  cosine_kp = pickle.load(handle)
with open(kp_cosine_finbert_checkpoint,'rb') as handle:
  cosine_finbert_kp = pickle.load(handle)
  
def calculate_top_k(out, tokens,text,kp_dict,exclude_text=False,exclude_words=False, k=5):
  sim_dict = {}
  pools = pool_embeddings(out, tokens).detach().numpy()
  for key in kp_dict.keys():
    if key == text:
      continue
    if exclude_text and text in key:
      continue
    if exclude_words and True in [x in key for x in text.split(" ")]:
      continue
    sim_dict[key] = cosine_similarity(
        pools,
        [kp_dict[key]]
    )[0][0]
  sims = sorted(sim_dict.items(), key= lambda x: x[1], reverse = True)[:k]
  return {x:y for x,y in sims}
def concat_tokens(sentences, tokenizer):
  tokens = {'input_ids': [], 'attention_mask': [], 'KPS': []}
  for sentence in sentences:
      # encode each sentence and append to dictionary
      new_tokens = tokenizer.encode_plus(sentence, max_length=64,
                                         truncation=True, padding='max_length',
                                         return_tensors='pt')
      tokens['input_ids'].append(new_tokens['input_ids'][0])
      tokens['attention_mask'].append(new_tokens['attention_mask'][0])
      tokens['KPS'].append(sentence)
  # reformat list of tensors into single tensor
  tokens['input_ids'] = torch.stack(tokens['input_ids'])
  tokens['attention_mask'] = torch.stack(tokens['attention_mask'])
  return tokens

def pool_embeddings(out, tok):
  embeddings = out["hidden_states"][-1]
  attention_mask = tok['attention_mask']
  mask = attention_mask.unsqueeze(-1).expand(embeddings.size()).float()
  masked_embeddings = embeddings * mask
  summed = torch.sum(masked_embeddings, 1)
  summed_mask = torch.clamp(mask.sum(1), min=1e-9)
  mean_pooled = summed / summed_mask
  return mean_pooled
  
def extract_idxs(top_dict, kp_dict):
  idxs = []
  c = 0
  for i in list(kp_dict.keys()):
    if i in top_dict.keys():
      idxs.append(c)
    c+=1
  return idxs
  
if text:
  text = text.lower()
  new_tokens = concat_tokens([text], tokenizer)
  new_tokens.pop("KPS")
  new_tokens_finbert = concat_tokens([text], tokenizer_finbert)
  new_tokens_finbert.pop("KPS")
  with torch.no_grad():
    outputs = model(**new_tokens)
    outputs_finbert = model_finbert(**new_tokens_finbert)
  sim_dict = calculate_top_k(outputs, new_tokens, text, kp_dict, exclude_text=exclude_text,exclude_words=exclude_words,k=k)
  sim_dict_finbert = calculate_top_k(outputs_finbert, new_tokens_finbert, text, kp_dict_finbert, exclude_text=exclude_text,exclude_words=exclude_words,k=k)
  if not diversify_box:
    with col1:
      st.write("distilbert-cvent")
      st.json(sim_dict)
    with col2:
      st.write("finbert")
      st.json(sim_dict_finbert)
  else:
    idxs = extract_idxs(sim_dict, kp_dict)
    idxs_finbert = extract_idxs(sim_dict_finbert, kp_dict_finbert)
    distances_candidates = cosine_kp[np.ix_(idxs, idxs)]
    distances_candidates_finbert = cosine_finbert_kp[np.ix_(idxs_finbert, idxs_finbert)]
    #first do distilbert
    candidate = None
    min_sim = np.inf
    for combination in itertools.combinations(range(len(idxs)), k):
      sim = sum([distances_candidates[i][j] for i in combination for j in combination if i != j])
      if sim < min_sim:
        candidate = combination
        min_sim = sim
    #then do finbert
    candidate_finbert = None
    min_sim = np.inf
    for combination in itertools.combinations(range(len(idxs_finbert)), k):
      sim = sum([distances_candidates_finbert[i][j] for i in combination for j in combination if i != j])
      if sim < min_sim:
        candidate_finbert = combination
        min_sim = sim
    #distilbert
    ret = {keys[idxs[idx]]:sim_dict[keys[idxs[idx]]] for idx in candidate}
    ret = sorted(ret.items(), key= lambda x: x[1], reverse = True)
    ret = {x:y for x,y in ret}
    #finbert
    ret_finbert = {keys_finbert[idxs_finbert[idx]]:sim_dict_finbert[keys_finbert[idxs_finbert[idx]]] for idx in candidate_finbert}
    candidate_finbert = sorted(candidate_finbert.items(), key= lambda x: x[1], reverse = True)
    candidate_finbert = {x:y for x,y in candidate_finbert}
    with col1:
      st.write("distilbert-cvent")
      st.json(ret)
    with col2:
      st.write("finbert")
      st.json(ret_finbert)