update

.streamlit/config.toml

@@ -1,4 +1,7 @@
 [server]
 # Max size, in megabytes, for files uploaded with the file_uploader.
 # Default: 200
-maxUploadSize = 10
+maxUploadSize = 20
+
+[browser]
+gatherUsageStats = false

main.py

@@ -0,0 +1,56 @@
+import streamlit as st
+from src.utils import get_logo, read_file, convert_df
+from src.components import form, faq, presentation, footer, about
+
+
+# app configs
+st.set_page_config(
+    page_title="Wordify",
+    initial_sidebar_state="expanded",
+    layout="centered",
+    page_icon="./assets/logo.png",
+    menu_items={
+        'Get Help': "https://github.com/MilaNLProc/wordify-webapp-streamlit/issues/new",
+        'Report a Bug': "https://github.com/MilaNLProc/wordify-webapp-streamlit/issues/new",
+        'About': about(),
+    }
+)
+
+# logo
+st.sidebar.image(get_logo("./assets/logo.png"))
+
+# title
+st.title("Wordify")
+
+# file uploader
+uploaded_fl = st.sidebar.file_uploader(
+    label="Choose a file",
+    type=["csv", "parquet", "tsv", "xlsx"],
+    accept_multiple_files=False,
+    help="""
+        Supported formats:
+        - CSV
+        - TSV 
+        - PARQUET
+        - XLSX (do not support [Strict Open XML Spreadsheet format](https://stackoverflow.com/questions/62800822/openpyxl-cannot-read-strict-open-xml-spreadsheet-format-userwarning-file-conta))
+    """,
+)
+
+if not uploaded_fl:
+    presentation()
+    faq()
+else:
+    df = read_file(uploaded_fl)
+    new_df = form(df)
+    if new_df is not None:
+        payload = convert_df(new_df)
+        st.download_button(
+            label="Download data as CSV",
+            data=payload,
+            file_name="wordify_results.csv",
+            mime="text/csv",
+        )
+
+
+# footer
+footer()

notebooks/wordifier_nb.ipynb

@@ -1,67 +1,589 @@
 {
- "metadata": {
-  "language_info": {
-   "codemirror_mode": {
-    "name": "ipython",
-    "version": 3
-   },
-   "file_extension": ".py",
-   "mimetype": "text/x-python",
-   "name": "python",
-   "nbconvert_exporter": "python",
-   "pygments_lexer": "ipython3",
-   "version": "3.8.3"
-  },
-  "orig_nbformat": 2,
-  "kernelspec": {
-   "name": "python383jvsc74a57bd01cb9a1c850fd1d16c5b98054247a74b7b7a12849bcfa00436ba202c2a9e2bbb2",
-   "display_name": "Python 3.8.3 64-bit ('py38': conda)"
-  }
- },
- "nbformat": 4,
- "nbformat_minor": 2,
  "cells": [
   {
    "cell_type": "code",
-   "execution_count": 1,
+   "execution_count": 65,
    "metadata": {},
    "outputs": [],
    "source": [
     "import sys\n",
-    "nb_dir = os.path.split(os.getcwd())[0]\n",
-    "if nb_dir not in sys.path:\n",
-    "    sys.path.append(nb_dir)\n",
+    "sys.path.insert(0, \"..\")\n",
+    "import vaex\n",
+    "from vaex.ml import LabelEncoder\n",
+    "import spacy\n",
+    "import pandas as pd\n",
+    "from tqdm import tqdm\n",
+    "import os\n",
+    "import multiprocessing as mp\n",
+    "from src.preprocessing import PreprocessingPipeline, encode\n",
+    "from src.wordifier import ModelConfigs\n",
+    "from sklearn.pipeline import Pipeline\n",
+    "from sklearn.linear_model import LogisticRegression\n",
+    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
+    "import numpy as np"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 67,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "pipe = PreprocessingPipeline(\n",
+    "    language=\"English\",\n",
+    "    pre_steps=list(PreprocessingPipeline.pipeline_components().keys()),\n",
+    "    lemmatization_step=list(PreprocessingPipeline.lemmatization_component().keys())[1],\n",
+    "    post_steps=list(PreprocessingPipeline.pipeline_components().keys()),\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 68,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def fn(t):\n",
+    "    return pipe.post(pipe.lemma(pipe.nlp(pipe.pre(t))))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 69,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "vdf = vaex.from_pandas(df)\n",
+    "vdf[\"processed_text\"] = vdf.apply(fn, arguments=[vdf[\"text\"]], vectorize=False)\n",
+    "df = vdf.to_pandas_df()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 71,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "2021-11-28 17:01:36.883 \n",
+      "  \u001b[33m\u001b[1mWarning:\u001b[0m to view this Streamlit app on a browser, run it with the following\n",
+      "  command:\n",
+      "\n",
+      "    streamlit run /Users/pietrolesci/miniconda3/envs/wordify/lib/python3.7/site-packages/ipykernel_launcher.py [ARGUMENTS]\n"
+     ]
+    }
+   ],
+   "source": [
+    "import streamlit as st\n",
+    "pbar = st.progress(0)\n",
+    "N = 100\n",
+    "for i, _ in enumerate(range(N)):\n",
+    "    if i % N == 0:\n",
+    "        pbar.progress(1)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 24,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "configs = ModelConfigs\n",
+    "clf = Pipeline(\n",
+    "    [\n",
+    "        (\"tfidf\", TfidfVectorizer()),\n",
+    "        (\n",
+    "            \"classifier\",\n",
+    "            LogisticRegression(\n",
+    "                penalty=\"l1\",\n",
+    "                C=configs.PENALTIES.value[np.random.randint(len(configs.PENALTIES.value))],\n",
+    "                solver=\"liblinear\",\n",
+    "                multi_class=\"auto\",\n",
+    "                max_iter=500,\n",
+    "                class_weight=\"balanced\",\n",
+    "            ),\n",
+    "        ),\n",
+    "    ]\n",
+    ")\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 29,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "Pipeline(steps=[('tfidf', TfidfVectorizer()),\n",
+       "                ('classifier',\n",
+       "                 LogisticRegression(C=1, class_weight='balanced', max_iter=500,\n",
+       "                                    penalty='l1', solver='liblinear'))])"
+      ]
+     },
+     "execution_count": 29,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "clf.fit(df[\"text\"], df[\"label\"])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 39,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "array(['00', '000', '00001', ..., 'ís', 'über', 'überwoman'], dtype=object)"
+      ]
+     },
+     "execution_count": 39,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": []
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 40,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def wordifier(df, text_col, label_col, configs=ModelConfigs):\n",
+    "\n",
+    "    n_instances, n_features = X.shape\n",
+    "    n_classes = np.unique(y)\n",
+    "\n",
+    "    # NOTE: the * 10 / 10 trick is to have \"nice\" round-ups\n",
+    "    sample_fraction = np.ceil((n_features / n_instances) * 10) / 10\n",
+    "\n",
+    "    sample_size = min(\n",
+    "        # this is the maximum supported\n",
+    "        configs.MAX_SELECTION.value,\n",
+    "        # at minimum you want MIN_SELECTION but in general you want\n",
+    "        # n_instances * sample_fraction\n",
+    "        max(configs.MIN_SELECTION.value, int(n_instances * sample_fraction)),\n",
+    "        # however if previous one is bigger the the available instances take\n",
+    "        # the number of available instances\n",
+    "        n_instances,\n",
+    "    )\n",
+    "\n",
+    "    # TODO: might want to try out something to subsample features at each iteration\n",
+    "\n",
+    "    # initialize coefficient matrices\n",
+    "    pos_scores = np.zeros((n_classes, n_features), dtype=int)\n",
+    "    neg_scores = np.zeros((n_classes, n_features), dtype=int)\n",
+    "\n",
+    "    for _ in range(configs.NUM_ITERS.value):\n",
+    "\n",
+    "        # run randomized regression\n",
+    "        clf = Pipeline([\n",
+    "            ('tfidf', TfidfVectorizer()), \n",
+    "            ('classifier', LogisticRegression(\n",
+    "                penalty=\"l1\",\n",
+    "                C=configs.PENALTIES.value[\n",
+    "                    np.random.randint(len(configs.PENALTIES.value))\n",
+    "                ],\n",
+    "                solver=\"liblinear\",\n",
+    "                multi_class=\"auto\",\n",
+    "                max_iter=500,\n",
+    "                class_weight=\"balanced\",\n",
+    "            ))]\n",
+    "        )\n",
+    "\n",
+    "        # sample indices to subsample matrix\n",
+    "        selection = resample(\n",
+    "            np.arange(n_instances), replace=True, stratify=y, n_samples=sample_size\n",
+    "        )\n",
+    "\n",
+    "        # fit\n",
+    "        try:\n",
+    "            clf.fit(X[selection], y[selection])\n",
+    "        except ValueError:\n",
+    "            continue\n",
+    "\n",
+    "        # record coefficients\n",
+    "        if n_classes == 2:\n",
+    "            pos_scores[1] = pos_scores[1] + (clf.coef_ > 0.0)\n",
+    "            neg_scores[1] = neg_scores[1] + (clf.coef_ < 0.0)\n",
+    "            pos_scores[0] = pos_scores[0] + (clf.coef_ < 0.0)\n",
+    "            neg_scores[0] = neg_scores[0] + (clf.coef_ > 0.0)\n",
+    "        else:\n",
+    "            pos_scores += clf.coef_ > 0\n",
+    "            neg_scores += clf.coef_ < 0\n",
+    "\n",
+    "\n",
+    "        # normalize\n",
+    "        pos_scores = pos_scores / configs.NUM_ITERS.value\n",
+    "        neg_scores = neg_scores / configs.NUM_ITERS.value\n",
+    "\n",
+    "        # get only active features\n",
+    "        pos_positions = np.where(\n",
+    "            pos_scores >= configs.SELECTION_THRESHOLD.value, pos_scores, 0\n",
+    "        )\n",
+    "        neg_positions = np.where(\n",
+    "            neg_scores >= configs.SELECTION_THRESHOLD.value, neg_scores, 0\n",
+    "        )\n",
+    "\n",
+    "        # prepare DataFrame\n",
+    "        X_names = clf.steps[0][1].get_feature_names_out()\n",
+    "        pos = [\n",
+    "            (X_names[i], pos_scores[c, i], y_names[c])\n",
+    "            for c, i in zip(*pos_positions.nonzero())\n",
+    "        ]\n",
+    "        neg = [\n",
+    "            (X_names[i], neg_scores[c, i], y_names[c])\n",
+    "            for c, i in zip(*neg_positions.nonzero())\n",
+    "        ]\n",
+    "\n",
+    "    posdf = pd.DataFrame(pos, columns=\"word score label\".split()).sort_values(\n",
+    "        [\"label\", \"score\"], ascending=False\n",
+    "    )\n",
+    "    negdf = pd.DataFrame(neg, columns=\"word score label\".split()).sort_values(\n",
+    "        [\"label\", \"score\"], ascending=False\n",
+    "    )\n",
+    "\n",
+    "    return posdf, negdf"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 41,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "res = vdf.apply(wordifier, arguments=[vdf.processed_text, vdf.encoded_label], vectorize=False)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 45,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from vaex.ml.sklearn import Predictor"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 60,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "clf = Pipeline(\n",
+    "    [\n",
+    "        (\n",
+    "            \"tfidf\",\n",
+    "            TfidfVectorizer(\n",
+    "                input=\"content\",  # default: file already in memory\n",
+    "                encoding=\"utf-8\",  # default\n",
+    "                decode_error=\"strict\",  # default\n",
+    "                strip_accents=None,  # do nothing\n",
+    "                lowercase=False,  # do nothing\n",
+    "                preprocessor=None,  # do nothing - default\n",
+    "                tokenizer=None,  # default\n",
+    "                stop_words=None,  # do nothing\n",
+    "                analyzer=\"word\",\n",
+    "                ngram_range=(1, 3),  # maximum 3-ngrams\n",
+    "                min_df=0.001,\n",
+    "                max_df=0.75,\n",
+    "                sublinear_tf=True,\n",
+    "            ),\n",
+    "        ),\n",
+    "        (\n",
+    "            \"classifier\",\n",
+    "            LogisticRegression(\n",
+    "                penalty=\"l1\",\n",
+    "                C=configs.PENALTIES.value[np.random.randint(len(configs.PENALTIES.value))],\n",
+    "                solver=\"liblinear\",\n",
+    "                multi_class=\"auto\",\n",
+    "                max_iter=500,\n",
+    "                class_weight=\"balanced\",\n",
+    "            ),\n",
+    "        ),\n",
+    "    ]\n",
+    ")\n",
     "\n",
+    "vaex_model = Predictor(\n",
+    "    features=[\"processed_text\"],\n",
+    "    target=\"encoded_label\",\n",
+    "    model=clf,\n",
+    "    prediction_name=\"prediction\",\n",
+    ")\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 61,
+   "metadata": {},
+   "outputs": [
+    {
+     "ename": "TypeError",
+     "evalue": "unhashable type: 'list'",
+     "output_type": "error",
+     "traceback": [
+      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
+      "\u001b[0;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
+      "\u001b[0;32m/var/folders/b_/m81mmt0s6gv48kdvk44n2l740000gn/T/ipykernel_52217/687453386.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mvaex_model\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mvdf\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
+      "\u001b[0;32m~/miniconda3/envs/wordify/lib/python3.7/site-packages/vaex/ml/sklearn.py\u001b[0m in \u001b[0;36mfit\u001b[0;34m(self, df, **kwargs)\u001b[0m\n\u001b[1;32m    103\u001b[0m         '''\n\u001b[1;32m    104\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 105\u001b[0;31m         \u001b[0mX\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mdf\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfeatures\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    106\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtarget\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    107\u001b[0m             \u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mdf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mevaluate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtarget\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
+      "\u001b[0;32m~/miniconda3/envs/wordify/lib/python3.7/site-packages/vaex/dataframe.py\u001b[0m in \u001b[0;36mvalues\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m   6897\u001b[0m         \u001b[0mIf\u001b[0m \u001b[0many\u001b[0m \u001b[0mof\u001b[0m \u001b[0mthe\u001b[0m \u001b[0mcolumns\u001b[0m \u001b[0mcontain\u001b[0m \u001b[0mmasked\u001b[0m \u001b[0marrays\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mthe\u001b[0m \u001b[0mmasks\u001b[0m \u001b[0mare\u001b[0m \u001b[0mignored\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0me\u001b[0m\u001b[0;34m.\u001b[0m \u001b[0mthe\u001b[0m \u001b[0mmasked\u001b[0m \u001b[0melements\u001b[0m \u001b[0mare\u001b[0m \u001b[0mreturned\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mwell\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   6898\u001b[0m         \"\"\"\n\u001b[0;32m-> 6899\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__array__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   6900\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   6901\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
+      "\u001b[0;32m~/miniconda3/envs/wordify/lib/python3.7/site-packages/vaex/dataframe.py\u001b[0m in \u001b[0;36m__array__\u001b[0;34m(self, dtype, parallel)\u001b[0m\n\u001b[1;32m   5989\u001b[0m                 \u001b[0;32mif\u001b[0m \u001b[0mcolumn_type\u001b[0m \u001b[0;34m!=\u001b[0m \u001b[0mdtype\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   5990\u001b[0m                     \u001b[0;32mraise\u001b[0m \u001b[0mValueError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"Cannot cast %r (of type %r) to %r\"\u001b[0m \u001b[0;34m%\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mname\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdata_type\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mname\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdtype\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 5991\u001b[0;31m         \u001b[0mchunks\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mevaluate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcolumn_names\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mparallel\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mparallel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0marray_type\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m'numpy'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   5992\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0many\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mma\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0misMaskedArray\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mchunk\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mchunk\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mchunks\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   5993\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mma\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0marray\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mchunks\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdtype\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mdtype\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mT\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
+      "\u001b[0;32m~/miniconda3/envs/wordify/lib/python3.7/site-packages/vaex/dataframe.py\u001b[0m in \u001b[0;36mevaluate\u001b[0;34m(self, expression, i1, i2, out, selection, filtered, array_type, parallel, chunk_size, progress)\u001b[0m\n\u001b[1;32m   2962\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mevaluate_iterator\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mexpression\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0ms1\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mi1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0ms2\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mi2\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mout\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mout\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mselection\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mselection\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfiltered\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mfiltered\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0marray_type\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0marray_type\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mparallel\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mparallel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mchunk_size\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mchunk_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mprogress\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mprogress\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2963\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2964\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_evaluate_implementation\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mexpression\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mi1\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mi1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mi2\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mi2\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mout\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mout\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mselection\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mselection\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfiltered\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mfiltered\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0marray_type\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0marray_type\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mparallel\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mparallel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mchunk_size\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mchunk_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mprogress\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mprogress\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2965\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2966\u001b[0m     \u001b[0;34m@\u001b[0m\u001b[0mdocsubst\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
+      "\u001b[0;32m~/miniconda3/envs/wordify/lib/python3.7/site-packages/vaex/dataframe.py\u001b[0m in \u001b[0;36m_evaluate_implementation\u001b[0;34m(self, expression, i1, i2, out, selection, filtered, array_type, parallel, chunk_size, raw, progress)\u001b[0m\n\u001b[1;32m   6207\u001b[0m             \u001b[0;31m# TODO: For NEP branch: dtype -> dtype_evaluate\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   6208\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 6209\u001b[0;31m             \u001b[0mexpression_to_evaluate\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlist\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mset\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mexpressions\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m  \u001b[0;31m# lets assume we have to do them all\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   6210\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   6211\u001b[0m             \u001b[0;32mfor\u001b[0m \u001b[0mexpression\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mset\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mexpressions\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
+      "\u001b[0;31mTypeError\u001b[0m: unhashable type: 'list'"
+     ]
+    }
+   ],
+   "source": [
+    "vaex_model.fit(vdf)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 52,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "b'\\x80\\x03c__main__\\nwordifier\\nq\\x00.'"
+      ]
+     },
+     "execution_count": 52,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "import pickle\n",
+    "pickle.dumps(wordifier)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 47,
+   "metadata": {},
+   "outputs": [
+    {
+     "ename": "TypeError",
+     "evalue": "unhashable type: 'list'",
+     "output_type": "error",
+     "traceback": [
+      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
+      "\u001b[0;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
+      "\u001b[0;32m/var/folders/b_/m81mmt0s6gv48kdvk44n2l740000gn/T/ipykernel_52217/687453386.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mvaex_model\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mvdf\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
+      "\u001b[0;32m~/miniconda3/envs/wordify/lib/python3.7/site-packages/vaex/ml/sklearn.py\u001b[0m in \u001b[0;36mfit\u001b[0;34m(self, df, **kwargs)\u001b[0m\n\u001b[1;32m    103\u001b[0m         '''\n\u001b[1;32m    104\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 105\u001b[0;31m         \u001b[0mX\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mdf\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfeatures\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    106\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtarget\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    107\u001b[0m             \u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mdf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mevaluate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtarget\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
+      "\u001b[0;32m~/miniconda3/envs/wordify/lib/python3.7/site-packages/vaex/dataframe.py\u001b[0m in \u001b[0;36mvalues\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m   6897\u001b[0m         \u001b[0mIf\u001b[0m \u001b[0many\u001b[0m \u001b[0mof\u001b[0m \u001b[0mthe\u001b[0m \u001b[0mcolumns\u001b[0m \u001b[0mcontain\u001b[0m \u001b[0mmasked\u001b[0m \u001b[0marrays\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mthe\u001b[0m \u001b[0mmasks\u001b[0m \u001b[0mare\u001b[0m \u001b[0mignored\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0me\u001b[0m\u001b[0;34m.\u001b[0m \u001b[0mthe\u001b[0m \u001b[0mmasked\u001b[0m \u001b[0melements\u001b[0m \u001b[0mare\u001b[0m \u001b[0mreturned\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mwell\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   6898\u001b[0m         \"\"\"\n\u001b[0;32m-> 6899\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__array__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   6900\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   6901\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
+      "\u001b[0;32m~/miniconda3/envs/wordify/lib/python3.7/site-packages/vaex/dataframe.py\u001b[0m in \u001b[0;36m__array__\u001b[0;34m(self, dtype, parallel)\u001b[0m\n\u001b[1;32m   5989\u001b[0m                 \u001b[0;32mif\u001b[0m \u001b[0mcolumn_type\u001b[0m \u001b[0;34m!=\u001b[0m \u001b[0mdtype\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   5990\u001b[0m                     \u001b[0;32mraise\u001b[0m \u001b[0mValueError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"Cannot cast %r (of type %r) to %r\"\u001b[0m \u001b[0;34m%\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mname\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdata_type\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mname\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdtype\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 5991\u001b[0;31m         \u001b[0mchunks\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mevaluate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcolumn_names\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mparallel\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mparallel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0marray_type\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m'numpy'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   5992\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0many\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mma\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0misMaskedArray\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mchunk\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mchunk\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mchunks\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   5993\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mma\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0marray\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mchunks\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdtype\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mdtype\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mT\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
+      "\u001b[0;32m~/miniconda3/envs/wordify/lib/python3.7/site-packages/vaex/dataframe.py\u001b[0m in \u001b[0;36mevaluate\u001b[0;34m(self, expression, i1, i2, out, selection, filtered, array_type, parallel, chunk_size, progress)\u001b[0m\n\u001b[1;32m   2962\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mevaluate_iterator\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mexpression\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0ms1\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mi1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0ms2\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mi2\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mout\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mout\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mselection\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mselection\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfiltered\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mfiltered\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0marray_type\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0marray_type\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mparallel\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mparallel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mchunk_size\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mchunk_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mprogress\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mprogress\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2963\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2964\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_evaluate_implementation\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mexpression\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mi1\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mi1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mi2\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mi2\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mout\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mout\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mselection\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mselection\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfiltered\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mfiltered\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0marray_type\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0marray_type\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mparallel\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mparallel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mchunk_size\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mchunk_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mprogress\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mprogress\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2965\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2966\u001b[0m     \u001b[0;34m@\u001b[0m\u001b[0mdocsubst\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
+      "\u001b[0;32m~/miniconda3/envs/wordify/lib/python3.7/site-packages/vaex/dataframe.py\u001b[0m in \u001b[0;36m_evaluate_implementation\u001b[0;34m(self, expression, i1, i2, out, selection, filtered, array_type, parallel, chunk_size, raw, progress)\u001b[0m\n\u001b[1;32m   6207\u001b[0m             \u001b[0;31m# TODO: For NEP branch: dtype -> dtype_evaluate\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   6208\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 6209\u001b[0;31m             \u001b[0mexpression_to_evaluate\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlist\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mset\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mexpressions\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m  \u001b[0;31m# lets assume we have to do them all\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   6210\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   6211\u001b[0m             \u001b[0;32mfor\u001b[0m \u001b[0mexpression\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mset\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mexpressions\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
+      "\u001b[0;31mTypeError\u001b[0m: unhashable type: 'list'"
+     ]
+    }
+   ],
+   "source": []
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "res = []\n",
+    "with tqdm(total=len(df)) as pbar:\n",
+    "    for doc in tqdm(nlp.pipe(df[\"text\"].values, batch_size=500, n_process=n_cpus)):\n",
+    "        res.append([i.lemma_ for i in doc])\n",
+    "        pbar.update(1)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import pickle"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def fn(t):\n",
+    "    return "
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "%%timeit\n",
+    "with mp.Pool(mp.cpu_count()) as pool:\n",
+    "    new_s = pool.map(nlp, df[\"text\"].values)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from typing import List\n",
     "import numpy as np\n",
     "import pandas as pd\n",
-    "# import modin.pandas as mpd\n",
-    "import spacy\n",
-    "from src.configs import ModelConfigs, Languages\n",
-    "from src.utils import wordifier, TextPreprocessor, encode\n",
-    "\n",
-    "from textacy.preprocessing import make_pipeline, remove, replace, normalize\n",
-    "from tqdm import trange\n",
-    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
+    "import streamlit as st\n",
     "from sklearn.linear_model import LogisticRegression\n",
-    "from sklearn.preprocessing import LabelEncoder\n",
     "from sklearn.utils import resample\n",
-    "import multiprocessing as mp\n",
-    "# import dask.dataframe as dask_df\n",
-    "from stqdm import stqdm\n",
-    "stqdm.pandas()\n",
     "\n",
-    "from tqdm import trange\n",
+    "from src.configs import ModelConfigs\n",
     "\n",
-    "import os\n",
-    "# os.environ[\"MODIN_ENGINE\"] = \"ray\"  # Modin will use Ray\n",
     "\n",
-    "import vaex\n",
-    "pd.set_option(\"display.max_colwidth\", None)"
+    "def wordifier(X, y, X_names: List[str], y_names: List[str], configs=ModelConfigs):\n",
+    "\n",
+    "    n_instances, n_features = X.shape\n",
+    "    n_classes = len(y_names)\n",
+    "\n",
+    "    # NOTE: the * 10 / 10 trick is to have \"nice\" round-ups\n",
+    "    sample_fraction = np.ceil((n_features / n_instances) * 10) / 10\n",
+    "\n",
+    "    sample_size = min(\n",
+    "        # this is the maximum supported\n",
+    "        configs.MAX_SELECTION.value,\n",
+    "        # at minimum you want MIN_SELECTION but in general you want\n",
+    "        # n_instances * sample_fraction\n",
+    "        max(configs.MIN_SELECTION.value, int(n_instances * sample_fraction)),\n",
+    "        # however if previous one is bigger the the available instances take\n",
+    "        # the number of available instances\n",
+    "        n_instances,\n",
+    "    )\n",
+    "\n",
+    "    # TODO: might want to try out something to subsample features at each iteration\n",
+    "\n",
+    "    # initialize coefficient matrices\n",
+    "    pos_scores = np.zeros((n_classes, n_features), dtype=int)\n",
+    "    neg_scores = np.zeros((n_classes, n_features), dtype=int)\n",
+    "\n",
+    "    with st.spinner(\"Wordifying!\"):\n",
+    "        pbar = st.progress(0)\n",
+    "\n",
+    "        for i, _ in enumerate(range(configs.NUM_ITERS.value)):\n",
+    "\n",
+    "            # run randomized regression\n",
+    "            clf = LogisticRegression(\n",
+    "                penalty=\"l1\",\n",
+    "                C=configs.PENALTIES.value[\n",
+    "                    np.random.randint(len(configs.PENALTIES.value))\n",
+    "                ],\n",
+    "                solver=\"liblinear\",\n",
+    "                multi_class=\"auto\",\n",
+    "                max_iter=500,\n",
+    "                class_weight=\"balanced\",\n",
+    "            )\n",
+    "\n",
+    "            # sample indices to subsample matrix\n",
+    "            selection = resample(\n",
+    "                np.arange(n_instances), replace=True, stratify=y, n_samples=sample_size\n",
+    "            )\n",
+    "\n",
+    "            # fit\n",
+    "            try:\n",
+    "                clf.fit(X[selection], y[selection])\n",
+    "            except ValueError:\n",
+    "                continue\n",
+    "\n",
+    "            # record coefficients\n",
+    "            if n_classes == 2:\n",
+    "                pos_scores[1] = pos_scores[1] + (clf.coef_ > 0.0)\n",
+    "                neg_scores[1] = neg_scores[1] + (clf.coef_ < 0.0)\n",
+    "                pos_scores[0] = pos_scores[0] + (clf.coef_ < 0.0)\n",
+    "                neg_scores[0] = neg_scores[0] + (clf.coef_ > 0.0)\n",
+    "            else:\n",
+    "                pos_scores += clf.coef_ > 0\n",
+    "                neg_scores += clf.coef_ < 0\n",
+    "\n",
+    "            pbar.progress(i + 1)\n",
+    "\n",
+    "        # normalize\n",
+    "        pos_scores = pos_scores / configs.NUM_ITERS.value\n",
+    "        neg_scores = neg_scores / configs.NUM_ITERS.value\n",
+    "\n",
+    "        # get only active features\n",
+    "        pos_positions = np.where(\n",
+    "            pos_scores >= configs.SELECTION_THRESHOLD.value, pos_scores, 0\n",
+    "        )\n",
+    "        neg_positions = np.where(\n",
+    "            neg_scores >= configs.SELECTION_THRESHOLD.value, neg_scores, 0\n",
+    "        )\n",
+    "\n",
+    "        # prepare DataFrame\n",
+    "        pos = [\n",
+    "            (X_names[i], pos_scores[c, i], y_names[c])\n",
+    "            for c, i in zip(*pos_positions.nonzero())\n",
+    "        ]\n",
+    "        neg = [\n",
+    "            (X_names[i], neg_scores[c, i], y_names[c])\n",
+    "            for c, i in zip(*neg_positions.nonzero())\n",
+    "        ]\n",
+    "\n",
+    "    posdf = pd.DataFrame(pos, columns=\"word score label\".split()).sort_values(\n",
+    "        [\"label\", \"score\"], ascending=False\n",
+    "    )\n",
+    "    negdf = pd.DataFrame(neg, columns=\"word score label\".split()).sort_values(\n",
+    "        [\"label\", \"score\"], ascending=False\n",
+    "    )\n",
+    "\n",
+    "    return posdf, negdf\n"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 4,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -70,7 +592,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 28,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -79,7 +601,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 29,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -91,7 +613,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 30,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -104,24 +626,16 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 31,
+   "execution_count": null,
    "metadata": {},
-   "outputs": [
-    {
-     "output_type": "stream",
-     "name": "stderr",
-     "text": [
-      "100%|██████████| 9939/9939 [00:06<00:00, 1431.09it/s]\n"
-     ]
-    }
-   ],
+   "outputs": [],
    "source": [
     "df[\"p_text\"] = prep.fit_transform(df[\"text\"])"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 32,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -130,7 +644,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 21,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -146,28 +660,9 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 22,
+   "execution_count": null,
    "metadata": {},
-   "outputs": [
-    {
-     "output_type": "stream",
-     "name": "stdout",
-     "text": [
-      "CPU times: user 1.45 s, sys: 10.6 ms, total: 1.46 s\nWall time: 1.46 s\n"
-     ]
-    },
-    {
-     "output_type": "execute_result",
-     "data": {
-      "text/plain": [
-       "LogisticRegression(C=0.05, class_weight='balanced', max_iter=500, penalty='l1',\n",
-       "                   solver='liblinear')"
-      ]
-     },
-     "metadata": {},
-     "execution_count": 22
-    }
-   ],
+   "outputs": [],
    "source": [
     "%%time\n",
     "clf.fit(X, y)"
@@ -182,32 +677,9 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 14,
+   "execution_count": null,
    "metadata": {},
-   "outputs": [
-    {
-     "output_type": "stream",
-     "name": "stderr",
-     "text": [
-      "  6%|▌         | 28/500 [01:01<27:33,  3.50s/it]/Users/49796/miniconda3/envs/py38/lib/python3.8/site-packages/sklearn/svm/_base.py:976: ConvergenceWarning: Liblinear failed to converge, increase the number of iterations.\n",
-      "  warnings.warn(\"Liblinear failed to converge, increase \"\n",
-      " 31%|███       | 156/500 [06:18<13:54,  2.43s/it]\n"
-     ]
-    },
-    {
-     "output_type": "error",
-     "ename": "KeyboardInterrupt",
-     "evalue": "",
-     "traceback": [
-      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
-      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
-      "\u001b[0;32m<ipython-input-14-1fef5b7ccf45>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m     39\u001b[0m     \u001b[0;31m# fit\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     40\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 41\u001b[0;31m         \u001b[0mclf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mselection\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mselection\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     42\u001b[0m     \u001b[0;32mexcept\u001b[0m \u001b[0mValueError\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     43\u001b[0m         \u001b[0;32mcontinue\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;32m~/miniconda3/envs/py38/lib/python3.8/site-packages/sklearn/linear_model/_logistic.py\u001b[0m in \u001b[0;36mfit\u001b[0;34m(self, X, y, sample_weight)\u001b[0m\n\u001b[1;32m   1354\u001b[0m                               \u001b[0;34m\" 'solver' is set to 'liblinear'. Got 'n_jobs'\"\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m                               \" = {}.\".format(effective_n_jobs(self.n_jobs)))\n\u001b[0;32m-> 1356\u001b[0;31m             self.coef_, self.intercept_, n_iter_ = _fit_liblinear(\n\u001b[0m\u001b[1;32m   1357\u001b[0m                 \u001b[0mX\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mC\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfit_intercept\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mintercept_scaling\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1358\u001b[0m                 \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mclass_weight\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpenalty\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdual\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mverbose\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;32m~/miniconda3/envs/py38/lib/python3.8/site-packages/sklearn/svm/_base.py\u001b[0m in \u001b[0;36m_fit_liblinear\u001b[0;34m(X, y, C, fit_intercept, intercept_scaling, class_weight, penalty, dual, verbose, max_iter, tol, random_state, multi_class, loss, epsilon, sample_weight)\u001b[0m\n\u001b[1;32m    964\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    965\u001b[0m     \u001b[0msolver_type\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_get_liblinear_solver_type\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmulti_class\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpenalty\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mloss\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdual\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 966\u001b[0;31m     raw_coef_, n_iter_ = liblinear.train_wrap(\n\u001b[0m\u001b[1;32m    967\u001b[0m         \u001b[0mX\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_ind\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0misspmatrix\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msolver_type\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtol\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbias\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mC\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    968\u001b[0m         \u001b[0mclass_weight_\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmax_iter\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrnd\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrandint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0miinfo\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'i'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmax\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
-     ]
-    }
-   ],
+   "outputs": [],
    "source": [
     "n_instances, n_features = X.shape\n",
     "n_classes = len(y_names)\n",
@@ -293,5 +765,30 @@
    "outputs": [],
    "source": []
   }
- ]
-}
+ ],
+ "metadata": {
+  "interpreter": {
+   "hash": "aa7efd0b3ada76bb0689aa8ed0b61d7de788847e3d11d2d142fc5800c765982f"
+  },
+  "kernelspec": {
+   "display_name": "Python 3.8.3 64-bit ('py38': conda)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.7.11"
+  },
+  "orig_nbformat": 2
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}

src/components.py

@@ -0,0 +1,232 @@
+import streamlit as st
+from src.preprocessing import PreprocessingPipeline
+from src.wordifier import input_transform, wordifier, output_transform
+from src.configs import PreprocessingConfigs, SupportedFiles, Languages
+
+
+@st.experimental_memo
+def form(df):
+    with st.form("my_form"):
+        col1, col2 = st.columns([1, 2])
+        with col1:
+
+            cols = [""] + df.columns.tolist()
+            label_column = st.selectbox(
+                "Select label column", cols, index=0, help="Select the column containing the labels"
+            )
+            text_column = st.selectbox(
+                "Select text column", cols, index=0, help="Select the column containing the text"
+            )
+            language = st.selectbox(
+                "Select language",
+                [i.name for i in Languages],
+                help="""
+                    Select the language of your texts amongst the supported one. If we currently do
+                    not support it, feel free to open an issue
+                """,
+            )
+
+        with col2:
+            steps_options = list(PreprocessingPipeline.pipeline_components().keys())
+            pre_steps = st.multiselect(
+                "Select pre-lemmatization processing steps (ordered)",
+                options=steps_options,
+                default=[steps_options[i] for i in PreprocessingConfigs.DEFAULT_PRE.value],
+                format_func=lambda x: x.replace("_", " ").title(),
+                help="Select the processing steps to apply before the text is lemmatized",
+            )
+
+            lammatization_options = list(PreprocessingPipeline.lemmatization_component().keys())
+            lemmatization_step = st.selectbox(
+                "Select lemmatization",
+                options=lammatization_options,
+                index=PreprocessingConfigs.DEFAULT_LEMMA.value,
+                help="Select lemmatization procedure",
+            )
+
+            post_steps = st.multiselect(
+                "Select post-lemmatization processing steps (ordered)",
+                options=steps_options,
+                default=[steps_options[i] for i in PreprocessingConfigs.DEFAULT_POST.value],
+                format_func=lambda x: x.replace("_", " ").title(),
+                help="Select the processing steps to apply after the text is lemmatized",
+            )
+
+        # Every form must have a submit button.
+        submitted = st.form_submit_button("Submit")
+        if submitted:
+
+            # preprocess
+            with st.spinner("Step 1/4: Preprocessing text"):
+                pipe = PreprocessingPipeline(language, pre_steps, lemmatization_step, post_steps)
+                df = pipe.vaex_process(df, text_column)
+
+            # prepare input
+            with st.spinner("Step 2/4: Preparing inputs"):
+                input_dict = input_transform(df[text_column], df[label_column])
+
+            # wordify
+            with st.spinner("Step 3/4: Wordifying"):
+                pos, neg = wordifier(**input_dict)
+
+            # prepare output
+            with st.spinner("Step 4/4: Preparing outputs"):
+                new_df = output_transform(pos, neg)
+
+            return new_df
+
+
+def faq():
+    st.subheader("Frequently Asked Questions")
+    with st.expander("What is Wordify?"):
+        st.markdown(
+            """
+            __Wordify__ is a way to find out which n-grams (i.e., words and concatenations of words) are most indicative for each of your dependent
+            variable values.
+            """
+        )
+
+    with st.expander("What happens to my data?"):
+        st.markdown(
+            """
+            Nothing. We never store the data you upload on disk: it is only kept in memory for the
+            duration of the modeling, and then deleted. We do not retain any copies or traces of
+            your data.
+            """
+        )
+
+    with st.expander("What input formats do you support?"):
+        st.markdown(
+            f"""
+            We currently support {", ".join([i.name for i in SupportedFiles])}.
+            """
+        )
+
+    with st.expander("What languages are supported?"):
+        st.markdown(
+            f"""
+            Currently we support: {", ".join([i.name for i in Languages])}.
+            """
+        )
+
+    with st.expander("How does it work?"):
+        st.markdown(
+            """
+            It uses a variant of the Stability Selection algorithm
+            [(Meinshausen and Bühlmann, 2010)](https://rss.onlinelibrary.wiley.com/doi/full/10.1111/j.1467-9868.2010.00740.x)
+            to fit hundreds of logistic regression models on random subsets of the data, using
+            different L1 penalties to drive as many of the term coefficients to 0. Any terms that
+            receive a non-zero coefficient in at least 30% of all model runs can be seen as stable
+            indicators.
+            """
+        )
+
+    with st.expander("What libraries do you use?"):
+        st.markdown(
+            """
+            We leverage the power of many great libraries in the Python ecosystem:
+            - `Streamlit`
+            - `Pandas`
+            - `Numpy`
+            - `Spacy`
+            - `Scikit-learn`
+            - `Vaex`
+            """
+        )
+
+    with st.expander("How much data do I need?"):
+        st.markdown(
+            """
+            We recommend at least 2000 instances, the more, the better. With fewer instances, the
+            results are less replicable and reliable.
+            """
+        )
+
+    with st.expander("Is there a paper I can cite?"):
+        st.markdown(
+            """
+            Yes, please! Cite [Wordify: A Tool for Discovering and Differentiating Consumer Vocabularies](https://academic.oup.com/jcr/article/48/3/394/6199426)
+            ```
+            @article{10.1093/jcr/ucab018,
+                author = {Hovy, Dirk and Melumad, Shiri and Inman, J Jeffrey},
+                title = "{Wordify: A Tool for Discovering and Differentiating Consumer Vocabularies}",
+                journal = {Journal of Consumer Research},
+                volume = {48},
+                number = {3},
+                pages = {394-414},
+                year = {2021},
+                month = {03},
+                abstract = "{This work describes and illustrates a free and easy-to-use online text-analysis tool for understanding how consumer word use varies across contexts. The tool, Wordify, uses randomized logistic regression (RLR) to identify the words that best discriminate texts drawn from different pre-classified corpora, such as posts written by men versus women, or texts containing mostly negative versus positive valence. We present illustrative examples to show how the tool can be used for such diverse purposes as (1) uncovering the distinctive vocabularies that consumers use when writing reviews on smartphones versus PCs, (2) discovering how the words used in Tweets differ between presumed supporters and opponents of a controversial ad, and (3) expanding the dictionaries of dictionary-based sentiment-measurement tools. We show empirically that Wordify’s RLR algorithm performs better at discriminating vocabularies than support vector machines and chi-square selectors, while offering significant advantages in computing time. A discussion is also provided on the use of Wordify in conjunction with other text-analysis tools, such as probabilistic topic modeling and sentiment analysis, to gain more profound knowledge of the role of language in consumer behavior.}",
+                issn = {0093-5301},
+                doi = {10.1093/jcr/ucab018},
+                url = {https://doi.org/10.1093/jcr/ucab018},
+                eprint = {https://academic.oup.com/jcr/article-pdf/48/3/394/40853499/ucab018.pdf},
+            }
+            ```
+            """
+        )
+
+    with st.expander("How can I reach out to the Wordify team?"):
+        st.markdown(contacts(), unsafe_allow_html=True)
+
+
+def presentation():
+    st.markdown(
+        """
+        Wordify makes it easy to identify words that discriminate categories in textual data.
+        
+        :point_left: Start by uploading a file. *Once you upload the file, __Wordify__ will
+        show an interactive UI*.
+        """
+    )
+
+    st.subheader("Input format")
+    st.markdown(
+        """
+        Please note that your file must have a column with the texts and a column with the labels,
+        for example
+        """
+    )
+    st.table(
+        {"text": ["A review", "Another review", "Yet another one", "etc"], "label": ["Good", "Bad", "Good", "etc"]}
+    )
+
+    st.subheader("Output format")
+    st.markdown(
+        """
+        As a result of the process, you will get a file containing 4 columns:
+        - `Word`: the n-gram (i.e., a word or a concatenation of words) considered
+        - `Score`: the wordify score, between 0 and 1, of how important is `Word` to discrimitate `Label`
+        - `Label`: the label that `Word` is discriminating
+        - `Correlation`: how `Word` is correlated with `Label` (e.g., "negative" means that if `Word` is present in the text then the label is less likely to be `Label`)
+        """
+    )
+
+
+def footer():
+    st.sidebar.markdown(
+        """
+        <span style="font-size: 0.75em">Built with &hearts; by [`Pietro Lesci`](https://pietrolesci.github.io/) and [`MilaNLP`](https://twitter.com/MilaNLProc?ref_src=twsrc%5Egoogle%7Ctwcamp%5Eserp%7Ctwgr%5Eauthor).</span>
+        """,
+        unsafe_allow_html=True,
+    )
+
+
+def contacts():
+    return """
+    You can reach out to us via email, phone, or via mail
+    
+    - :email: wordify@unibocconi.it
+    
+    - :telephone_receiver: +39 02 5836 2604
+    
+    - :postbox: Via Röntgen n. 1, Milan 20136 (ITALY)
+
+
+    <iframe src="https://www.google.com/maps/embed?pb=!1m18!1m12!1m3!1d2798.949796165441!2d9.185730115812493!3d45.450667779100726!2m3!1f0!2f0!3f0!3m2!1i1024!2i768!4f13.1!3m3!1m2!1s0x4786c405ae6543c9%3A0xf2bb2313b36af88c!2sVia%20Guglielmo%20R%C3%B6ntgen%2C%201%2C%2020136%20Milano%20MI!5e0!3m2!1sit!2sit!4v1569325279433!5m2!1sit!2sit" frameborder="0" style="border:0; width: 100%; height: 312px;" allowfullscreen></iframe>
+    """
+
+def about():
+    return """
+    The wordify team
+    """

src/configs.py

@@ -10,6 +10,19 @@ class ModelConfigs(Enum):
     MIN_SELECTION = 10_000
 
 
+class InputTransformConfigs(Enum):
+    NGRAM_RANGE = (1, 3)
+    MIN_DF = 0.001
+    MAX_DF = 0.75
+    SUBLINEAR = True
+
+
+class PreprocessingConfigs(Enum):
+    DEFAULT_PRE = [1, 3, 5, 15, 21, 22, 18, 19, 0, 20, -1]
+    DEFAULT_LEMMA = 1
+    DEFAULT_POST = [20, -1]
+
+
 class Languages(Enum):
     English = "en_core_web_sm"
     Italian = "it_core_news_sm"

src/preprocessing.py

@@ -1,56 +1,20 @@
+import multiprocessing as mp
+import os
 import re
 import string
 from collections import OrderedDict
-from typing import Callable, List, Optional, Tuple
+from typing import Callable, List, Optional
 
-import numpy as np
 import pandas as pd
+from pandas.core.frame import DataFrame
 import spacy
 import streamlit as st
+import vaex
 from pandas.core.series import Series
-from sklearn.feature_extraction.text import TfidfVectorizer
-from sklearn.preprocessing import LabelEncoder
-from stqdm import stqdm
 from textacy.preprocessing import make_pipeline, normalize, remove, replace
 
 from .configs import Languages
 
-stqdm.pandas()
-
-
-def encode(text: pd.Series, labels: pd.Series):
-    """
-    Encodes text in mathematical object ameanable to training algorithm
-    """
-    tfidf_vectorizer = TfidfVectorizer(
-        input="content",  # default: file already in memory
-        encoding="utf-8",  # default
-        decode_error="strict",  # default
-        strip_accents=None,  # do nothing
-        lowercase=False,  # do nothing
-        preprocessor=None,  # do nothing - default
-        tokenizer=None,  # default
-        stop_words=None,  # do nothing
-        analyzer="word",
-        ngram_range=(1, 3),  # maximum 3-ngrams
-        min_df=0.001,
-        max_df=0.75,
-        sublinear_tf=True,
-    )
-    label_encoder = LabelEncoder()
-
-    with st.spinner("Encoding text using TF-IDF and Encoding labels"):
-        X = tfidf_vectorizer.fit_transform(text.values)
-        y = label_encoder.fit_transform(labels.values)
-
-    return {
-        "X": X,
-        "y": y,
-        "X_names": np.array(tfidf_vectorizer.get_feature_names()),
-        "y_names": label_encoder.classes_,
-    }
-
-
 # more [here](https://github.com/fastai/fastai/blob/master/fastai/text/core.py#L42)
 # and [here](https://textacy.readthedocs.io/en/latest/api_reference/preprocessing.html)
 # fmt: off
@@ -87,118 +51,101 @@ def normalize_repeating_words(t):
     return _re_wrep.sub(_replace_wrep, t)
 
 
-# fmt: on
-class Lemmatizer:
-    """Creates lemmatizer based on spacy"""
+def lowercase(t: str) -> str:
+    return t.lower()
 
-    def __init__(
-        self, language: str, remove_stop: bool = True, lemmatization: bool = True
-    ) -> None:
-        self.language = language
-        self.nlp = spacy.load(
-            Languages[language].value, exclude=["parser", "ner", "pos", "tok2vec"]
-        )
-        self._lemmatizer_fn = self._get_lemmatization_fn(remove_stop, lemmatization)
-        self.lemmatization = lemmatization
 
-    def _get_lemmatization_fn(
-        self, remove_stop: bool, lemmatization: bool
-    ) -> Optional[Callable]:
-        """Return the correct spacy Doc-level lemmatizer"""
-        if remove_stop and lemmatization:
+def strip(t: str) -> str:
+    return t.strip()
 
-            def lemmatizer_fn(doc: spacy.tokens.doc.Doc) -> str:
-                return " ".join(
-                    [t.lemma_ for t in doc if t.lemma_ != "-PRON-" and not t.is_stop]
-                )
 
-        elif remove_stop and not lemmatization:
+def lemmatize_remove_stopwords(doc: spacy.tokens.doc.Doc) -> str:
+    return " ".join(
+        [t.lemma_ for t in doc if t.lemma_ != "-PRON-" and not t.is_stop]
+    )
 
-            def lemmatizer_fn(doc: spacy.tokens.doc.Doc) -> str:
-                return " ".join([t.text for t in doc if not t.is_stop])
 
-        elif lemmatization and not remove_stop:
+def remove_stopwords(doc: spacy.tokens.doc.Doc) -> str:
+    return " ".join([t.text for t in doc if not t.is_stop])
 
-            def lemmatizer_fn(doc: spacy.tokens.doc.Doc) -> str:
-                return " ".join([t.lemma_ for t in doc if t.lemma_ != "-PRON-"])
 
-        else:
-            self.status = False
-            return
-
-        return lemmatizer_fn
-
-    def __call__(self, series: Series) -> Series:
-        """
-        Apply spacy pipeline to transform string to spacy Doc and applies lemmatization
-        """
-        res = []
-        pbar = stqdm(total=len(series), desc="Lemmatizing")
-        for doc in self.nlp.pipe(series, batch_size=500):
-            res.append(self._lemmatizer_fn(doc))
-            pbar.update(1)
-        pbar.close()
-        return pd.Series(res)
+def lemmatize_keep_stopwords(doc: spacy.tokens.doc.Doc) -> str:
+    return " ".join([t.lemma_ for t in doc if t.lemma_ != "-PRON-"])
 
 
+# fmt: on
 class PreprocessingPipeline:
     def __init__(
-        self, pre_steps: List[str], lemmatizer: Lemmatizer, post_steps: List[str]
+        self,
+        language: str,
+        pre_steps: Optional[List[str]],
+        lemmatization_step: Optional[str],
+        post_steps: Optional[List[str]],
     ):
+        self.language = language
+        self.pre_steps = pre_steps
+        self.lemmatization_step = lemmatization_step
+        self.post_steps = post_steps
 
-        # build pipeline
-        self.pre_pipeline, self.lemmatizer, self.post_pipeline = self.make_pipeline(
-            pre_steps, lemmatizer, post_steps
-        )
+        self.nlp = spacy.load(Languages[language].value, disable=["parser", "ner"])
+        self.pre = self.make_pre_post_component(self.pre_steps)
+        self.post = self.make_pre_post_component(self.post_steps)
+        self.lemma = self.lemmatization_component()[self.lemmatization_step]
 
-    def __call__(self, series: Series) -> Series:
-        with st.spinner("Pre-lemmatization cleaning"):
-            res = series.progress_map(self.pre_pipeline)
+    def apply_multiproc(fn, series):
+        with mp.Pool(mp.cpu_count()) as pool:
+            new_series = pool.map(fn, series)
+
+        return new_series
 
-        with st.spinner("Lemmatizing"):
-            res = self.lemmatizer(series)
+    def vaex_process(self, df: DataFrame, text_column: str) -> DataFrame:
+        def fn(t):
+            return self.post(self.lemma(self.nlp(self.pre(t))))
 
-        with st.spinner("Post-lemmatization cleaning"):
-            res = series.progress_map(self.post_pipeline)
+        vdf = vaex.from_pandas(df)
+        vdf["processed_text"] = vdf.apply(fn, arguments=[vdf[text_column]], vectorize=False)
 
-        return res
+        return vdf.to_pandas_df()
 
-    def make_pipeline(
-        self, pre_steps: List[str], lemmatizer: Lemmatizer, post_steps: List[str]
-    ) -> Tuple[Callable]:
+    def __call__(self, series: Series) -> Series:
+        if self.pre:
+            series = series.map(self.pre)
+
+        if self.lemma:
+            total_steps = len(series) // 100
+            res = []
+            pbar = st.progress(0)
+            for i, doc in enumerate(self.nlp.pipe(series, batch_size=500, n_process=os.cpu_count())):
+                res.append(self.lemma(doc))
 
-        # pre-lemmatization steps
-        pre_steps = [
-            self.pipeline_components()[step]
-            for step in pre_steps
-            if step in self.pipeline_components()
-        ]
-        pre_steps = make_pipeline(*pre_steps) if pre_steps else lambda x: x
+                if i % total_steps == 0:
+                    pbar.progress(1)
 
-        # lemmatization
-        lemmatizer = lemmatizer if lemmatizer.lemmatization else lambda x: x
+            series = pd.Series(res)
 
-        # post lemmatization steps
-        post_steps = [
-            self.pipeline_components()[step]
-            for step in post_steps
-            if step in self.pipeline_components()
-        ]
-        post_steps = make_pipeline(*post_steps) if post_steps else lambda x: x
+        if self.post:
+            series = series.map(self.post)
 
-        return pre_steps, lemmatizer, post_steps
+        return series
+
+    def make_pre_post_component(self, steps: Optional[List[str]]) -> Optional[Callable]:
+        if not steps:
+            return
+        components = [self.pipeline_components()[step] for step in steps]
+
+        return make_pipeline(*components)
 
     @staticmethod
     def pipeline_components() -> "OrderedDict[str, Callable]":
         """Returns available cleaning steps in order"""
         return OrderedDict(
             [
-                ("lower", lambda x: x.lower()),
+                ("lowercase", lowercase),
                 ("normalize_unicode", normalize.unicode),
                 ("normalize_bullet_points", normalize.bullet_points),
                 ("normalize_hyphenated_words", normalize.hyphenated_words),
                 ("normalize_quotation_marks", normalize.quotation_marks),
-                ("normalize_whitespace", normalize.whitespace),
+                ("normalize_whitespaces", normalize.whitespace),
                 ("replace_urls", replace.urls),
                 ("replace_currency_symbols", replace.currency_symbols),
                 ("replace_emails", replace.emails),
@@ -216,6 +163,17 @@ class PreprocessingPipeline:
                 ("normalize_useless_spaces", normalize_useless_spaces),
                 ("normalize_repeating_chars", normalize_repeating_chars),
                 ("normalize_repeating_words", normalize_repeating_words),
-                ("strip", lambda x: x.strip()),
+                ("strip", strip),
+            ]
+        )
+
+    @staticmethod
+    def lemmatization_component() -> "OrderedDict[str, Optional[Callable]]":
+        return OrderedDict(
+            [
+                ("Spacy lemmatizer (keep stopwords)", lemmatize_keep_stopwords),
+                ("Spacy lemmatizer (no stopwords)", lemmatize_remove_stopwords),
+                ("Disable lemmatizer", None),
+                ("Remove stopwords", remove_stopwords),
             ]
         )

src/utils.py

@@ -3,11 +3,9 @@ import altair as alt
 import pandas as pd
 import streamlit as st
 from PIL import Image
-from stqdm import stqdm
 
 from .configs import SupportedFiles
 
-stqdm.pandas()
 
 
 @st.cache
@@ -15,20 +13,19 @@ def get_logo(path):
     return Image.open(path)
 
 
-# @st.cache(suppress_st_warning=True)
 @st.cache(allow_output_mutation=True)
 def read_file(uploaded_file) -> pd.DataFrame:
-
     file_type = uploaded_file.name.split(".")[-1]
-    if file_type in set(i.name for i in SupportedFiles):
-        read_f = SupportedFiles[file_type].value[0]
-        df = read_f(uploaded_file)
-        # remove any NA
-        df = df.dropna()
-        return df
+    read_fn = SupportedFiles[file_type].value[0]
+    df = read_fn(uploaded_file)
+    df = df.dropna()
+    return df
+
 
-    else:
-        st.error("File type not supported")
+@st.cache
+def convert_df(df):
+    # IMPORTANT: Cache the conversion to prevent computation on every rerun
+    return df.to_csv(index=False, sep=";").encode("utf-8")
 
 
 def download_button(dataframe: pd.DataFrame, name: str):
@@ -55,12 +52,7 @@ def plot_labels_prop(data: pd.DataFrame, label_column: str):
 
         return
 
-    source = (
-        data[label_column]
-        .value_counts()
-        .reset_index()
-        .rename(columns={"index": "Labels", label_column: "Counts"})
-    )
+    source = data[label_column].value_counts().reset_index().rename(columns={"index": "Labels", label_column: "Counts"})
     source["Props"] = source["Counts"] / source["Counts"].sum()
     source["Proportions"] = (source["Props"].round(3) * 100).map("{:,.2f}".format) + "%"
 
@@ -73,9 +65,7 @@ def plot_labels_prop(data: pd.DataFrame, label_column: str):
         )
     )
 
-    text = bars.mark_text(align="center", baseline="middle", dy=15).encode(
-        text="Proportions:O"
-    )
+    text = bars.mark_text(align="center", baseline="middle", dy=15).encode(text="Proportions:O")
 
     return (bars + text).properties(height=300)
 
@@ -87,9 +77,7 @@ def plot_nchars(data: pd.DataFrame, text_column: str):
         alt.Chart(source)
         .mark_bar()
         .encode(
-            alt.X(
-                f"{text_column}:Q", bin=True, axis=alt.Axis(title="# chars per text")
-            ),
+            alt.X(f"{text_column}:Q", bin=True, axis=alt.Axis(title="# chars per text")),
             alt.Y("count()", axis=alt.Axis(title="")),
         )
     )
@@ -99,11 +87,7 @@ def plot_nchars(data: pd.DataFrame, text_column: str):
 
 def plot_score(data: pd.DataFrame, label_col: str, label: str):
 
-    source = (
-        data.loc[data[label_col] == label]
-        .sort_values("score", ascending=False)
-        .head(100)
-    )
+    source = data.loc[data[label_col] == label].sort_values("score", ascending=False).head(100)
 
     plot = (
         alt.Chart(source)

src/wordifier.py

@@ -1,17 +1,52 @@
-from typing import List
+from typing import Dict, List, Tuple
+
 import numpy as np
 import pandas as pd
 import streamlit as st
+from pandas.core.frame import DataFrame
+from sklearn.feature_extraction.text import TfidfVectorizer
 from sklearn.linear_model import LogisticRegression
+from sklearn.preprocessing import LabelEncoder
 from sklearn.utils import resample
-from stqdm import stqdm
 
-from .configs import ModelConfigs
+from .configs import InputTransformConfigs, ModelConfigs
+
+
+def input_transform(text: pd.Series, labels: pd.Series, configs=InputTransformConfigs) -> Dict[str, np.ndarray]:
+    """
+    Encodes text in mathematical object ameanable to training algorithm
+    """
+    tfidf_vectorizer = TfidfVectorizer(
+        input="content",  # default: file already in memory
+        encoding="utf-8",  # default
+        decode_error="strict",  # default
+        strip_accents=None,  # do nothing
+        lowercase=False,  # do nothing
+        preprocessor=None,  # do nothing - default
+        tokenizer=None,  # default
+        stop_words=None,  # do nothing
+        analyzer="word",
+        ngram_range=configs.NGRAM_RANGE.value,  # maximum 3-ngrams
+        min_df=configs.MIN_DF.value,
+        max_df=configs.MAX_DF.value,
+        sublinear_tf=configs.SUBLINEAR.value,
+    )
+    label_encoder = LabelEncoder()
+
+    X = tfidf_vectorizer.fit_transform(text.values)
+    y = label_encoder.fit_transform(labels.values)
 
-stqdm.pandas()
+    return {
+        "X": X,
+        "y": y,
+        "X_names": np.array(tfidf_vectorizer.get_feature_names_out()),
+        "y_names": label_encoder.classes_,
+    }
 
 
-def wordifier(X, y, X_names: List[str], y_names: List[str], configs=ModelConfigs):
+def wordifier(
+    X: np.ndarray, y: np.ndarray, X_names: List[str], y_names: List[str], configs=ModelConfigs
+) -> List[Tuple[str, float, str]]:
 
     n_instances, n_features = X.shape
     n_classes = len(y_names)
@@ -36,70 +71,62 @@ def wordifier(X, y, X_names: List[str], y_names: List[str], configs=ModelConfigs
     pos_scores = np.zeros((n_classes, n_features), dtype=int)
     neg_scores = np.zeros((n_classes, n_features), dtype=int)
 
-    with st.spinner("Wordifying!"):
-
-        for _ in stqdm(range(configs.NUM_ITERS.value)):
-
-            # run randomized regression
-            clf = LogisticRegression(
-                penalty="l1",
-                C=configs.PENALTIES.value[
-                    np.random.randint(len(configs.PENALTIES.value))
-                ],
-                solver="liblinear",
-                multi_class="auto",
-                max_iter=500,
-                class_weight="balanced",
-            )
-
-            # sample indices to subsample matrix
-            selection = resample(
-                np.arange(n_instances), replace=True, stratify=y, n_samples=sample_size
-            )
-
-            # fit
-            try:
-                clf.fit(X[selection], y[selection])
-            except ValueError:
-                continue
-
-            # record coefficients
-            if n_classes == 2:
-                pos_scores[1] = pos_scores[1] + (clf.coef_ > 0.0)
-                neg_scores[1] = neg_scores[1] + (clf.coef_ < 0.0)
-                pos_scores[0] = pos_scores[0] + (clf.coef_ < 0.0)
-                neg_scores[0] = neg_scores[0] + (clf.coef_ > 0.0)
-            else:
-                pos_scores += clf.coef_ > 0
-                neg_scores += clf.coef_ < 0
-
-        # normalize
-        pos_scores = pos_scores / configs.NUM_ITERS.value
-        neg_scores = neg_scores / configs.NUM_ITERS.value
-
-        # get only active features
-        pos_positions = np.where(
-            pos_scores >= configs.SELECTION_THRESHOLD.value, pos_scores, 0
-        )
-        neg_positions = np.where(
-            neg_scores >= configs.SELECTION_THRESHOLD.value, neg_scores, 0
+    pbar = st.progress(0)
+    for i, _ in enumerate(range(configs.NUM_ITERS.value)):
+
+        # run randomized regression
+        clf = LogisticRegression(
+            penalty="l1",
+            C=configs.PENALTIES.value[np.random.randint(len(configs.PENALTIES.value))],
+            solver="liblinear",
+            multi_class="auto",
+            max_iter=500,
+            class_weight="balanced",
         )
 
-        # prepare DataFrame
-        pos = [
-            (X_names[i], pos_scores[c, i], y_names[c])
-            for c, i in zip(*pos_positions.nonzero())
-        ]
-        neg = [
-            (X_names[i], neg_scores[c, i], y_names[c])
-            for c, i in zip(*neg_positions.nonzero())
-        ]
-
-    posdf = pd.DataFrame(pos, columns="word score label".split()).sort_values(
-        ["label", "score"], ascending=False
-    )
-    negdf = pd.DataFrame(neg, columns="word score label".split()).sort_values(
-        ["label", "score"], ascending=False
-    )
+        # sample indices to subsample matrix
+        selection = resample(np.arange(n_instances), replace=True, stratify=y, n_samples=sample_size)
+
+        # fit
+        try:
+            clf.fit(X[selection], y[selection])
+        except ValueError:
+            continue
+
+        # record coefficients
+        if n_classes == 2:
+            pos_scores[1] = pos_scores[1] + (clf.coef_ > 0.0)
+            neg_scores[1] = neg_scores[1] + (clf.coef_ < 0.0)
+            pos_scores[0] = pos_scores[0] + (clf.coef_ < 0.0)
+            neg_scores[0] = neg_scores[0] + (clf.coef_ > 0.0)
+        else:
+            pos_scores += clf.coef_ > 0
+            neg_scores += clf.coef_ < 0
+
+        pbar.progress(round(i / configs.NUM_ITERS.value, 1))
+
+    # normalize
+    pos_scores = pos_scores / configs.NUM_ITERS.value
+    neg_scores = neg_scores / configs.NUM_ITERS.value
+
+    # get only active features
+    pos_positions = np.where(pos_scores >= configs.SELECTION_THRESHOLD.value, pos_scores, 0)
+    neg_positions = np.where(neg_scores >= configs.SELECTION_THRESHOLD.value, neg_scores, 0)
+
+    # prepare DataFrame
+    pos = [(X_names[i], pos_scores[c, i], y_names[c]) for c, i in zip(*pos_positions.nonzero())]
+    neg = [(X_names[i], neg_scores[c, i], y_names[c]) for c, i in zip(*neg_positions.nonzero())]
+
+    return pos, neg
+
+
+def output_transform(pos: List[Tuple[str, float, str]], neg: List[Tuple[str, float, str]]) -> DataFrame:
+    posdf = pd.DataFrame(pos, columns="word score label".split()).sort_values(["label", "score"], ascending=False)
+    posdf["correlation"] = "positive"
+    negdf = pd.DataFrame(neg, columns="word score label".split()).sort_values(["label", "score"], ascending=False)
+    negdf["correlation"] = "negative"
+
+    output = pd.concat([posdf, negdf], ignore_index=False, axis=0)
+    output.columns = output.columns.str.title()
 
-    return posdf, negdf
+    return output