update stats

README.md

@@ -11,18 +11,18 @@ Details about the inspec dataset can be found in the original paper [(Hulth, 200
 Reference (indexer-assigned) keyphrases are also categorized under the PRMU (<u>P</u>resent-<u>R</u>eordered-<u>M</u>ixed-<u>U</u>nseen) scheme as proposed in [(Boudin and Gallina, 2021)][boudin-2021].
 
 Text pre-processing (tokenization) is carried out using `spacy` (`en_core_web_sm` model) with a special rule to avoid splitting words with hyphens (e.g. graph-based is kept as one token).
-Stemming (Porter's stemmer implementation provided in `nltk`) is performed before reference keyphrases are matched against the source text.
+Stemming (Porter's stemmer implementation provided in `nltk`) is applied before reference keyphrases are matched against the source text.
 Details about the process can be found in `prmu.py`.
 
 ## Content and statistics
 
 The dataset is divided into the following three splits:
 
-| Split      | # documents | # keyphrases | % Present | % Reordered | % Mixed | % Unseen |
-| :--------- | ----------: | -----------: | --------: | ----------: | ------: | -------: |
-| Train      | 1,000       | 9.79         | 77.83     | 9.90        | 6.30    | 5.98     |
-| Validation | 500         | 9.15         | 77.90     | 9.82        | 6.74    | 5.54     |
-| Test       | 500         | 9.83         | 78.49     | 9.82        | 6.76    | 4.92     |
+| Split      | # documents | #words | # keyphrases | % Present | % Reordered | % Mixed | % Unseen |
+| :--------- | ----------: | -----: | -----------: | --------: | ----------: | ------: | -------: |
+| Train      | 1,000       | 141.7  | 9.79         | 78.00     | 9.85        | 6.22    | 5.93     |
+| Validation | 500         | 132.2  | 9.15         | 77.96     | 9.82        | 6.75    | 5.47     |
+| Test       | 500         | 134.8  | 9.83         | 78.70     | 9.92        | 6.48    | 4.91     |
 
 The following data fields are available :
 

stats.ipynb

@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "code",
-   "execution_count": 4,
+   "execution_count": 1,
    "id": "eba2ee81",
    "metadata": {},
    "outputs": [
@@ -17,7 +17,7 @@
     {
      "data": {
       "application/vnd.jupyter.widget-view+json": {
-       "model_id": "1ee1af5876804725adcd149763bd27b8",
+       "model_id": "6b71557b1f2b4fb48282f7d5d4a4fe37",
        "version_major": 2,
        "version_minor": 0
       },
@@ -37,14 +37,14 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 20,
+   "execution_count": 2,
    "id": "4ba72244",
    "metadata": {},
    "outputs": [
     {
      "data": {
       "application/vnd.jupyter.widget-view+json": {
-       "model_id": "a14738b9d72b45d29d24cd764e272fd3",
+       "model_id": "5ece754125e04f6ca8856ab3d924eec6",
        "version_major": 2,
        "version_minor": 0
       },
@@ -61,16 +61,16 @@
      "text": [
       "statistics for train\n",
       "# keyphrases: 9.79\n",
-      "% P: 77.83\n",
-      "% R: 9.90\n",
-      "% M: 6.30\n",
-      "% U: 5.98\n"
+      "% P: 78.00\n",
+      "% R: 9.85\n",
+      "% M: 6.22\n",
+      "% U: 5.93\n"
      ]
     },
     {
      "data": {
       "application/vnd.jupyter.widget-view+json": {
-       "model_id": "59a4352f94d84ada80beb34607d63425",
+       "model_id": "f607af2e66a94ece874b8c5ebea207f4",
        "version_major": 2,
        "version_minor": 0
       },
@@ -87,16 +87,16 @@
      "text": [
       "statistics for validation\n",
       "# keyphrases: 9.15\n",
-      "% P: 77.90\n",
+      "% P: 77.96\n",
       "% R: 9.82\n",
-      "% M: 6.74\n",
-      "% U: 5.54\n"
+      "% M: 6.75\n",
+      "% U: 5.47\n"
      ]
     },
     {
      "data": {
       "application/vnd.jupyter.widget-view+json": {
-       "model_id": "77346b747c5248fb9566d8665c7017bb",
+       "model_id": "54e7f57783e74ca9b5f7815eb6413031",
        "version_major": 2,
        "version_minor": 0
       },
@@ -112,19 +112,17 @@
      "output_type": "stream",
      "text": [
       "statistics for test\n",
-      "# keyphrases: 9.83\n",
-      "% P: 78.49\n",
-      "% R: 9.82\n",
-      "% M: 6.76\n",
-      "% U: 4.92\n"
+      "# keyphrases: 9.82\n",
+      "% P: 78.70\n",
+      "% R: 9.92\n",
+      "% M: 6.48\n",
+      "% U: 4.91\n"
      ]
     }
    ],
    "source": [
     "from tqdm.notebook import tqdm\n",
     "\n",
-    "\n",
-    "\n",
     "for split in ['train', 'validation', 'test']:\n",
     "    \n",
     "    P, R, M, U, nb_kps = [], [], [], [], []\n",
@@ -144,10 +142,79 @@
     "    print(\"% U: {:.2f}\".format(sum(U)/len(U)*100))"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "id": "4e08f80f",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import spacy\n",
+    "\n",
+    "nlp = spacy.load(\"en_core_web_sm\")\n",
+    "\n",
+    "# https://spacy.io/usage/linguistic-features#native-tokenizer-additions\n",
+    "\n",
+    "from spacy.lang.char_classes import ALPHA, ALPHA_LOWER, ALPHA_UPPER\n",
+    "from spacy.lang.char_classes import CONCAT_QUOTES, LIST_ELLIPSES, LIST_ICONS\n",
+    "from spacy.util import compile_infix_regex\n",
+    "\n",
+    "# Modify tokenizer infix patterns\n",
+    "infixes = (\n",
+    "    LIST_ELLIPSES\n",
+    "    + LIST_ICONS\n",
+    "    + [\n",
+    "        r\"(?<=[0-9])[+\\-\\*^](?=[0-9-])\",\n",
+    "        r\"(?<=[{al}{q}])\\.(?=[{au}{q}])\".format(\n",
+    "            al=ALPHA_LOWER, au=ALPHA_UPPER, q=CONCAT_QUOTES\n",
+    "        ),\n",
+    "        r\"(?<=[{a}]),(?=[{a}])\".format(a=ALPHA),\n",
+    "        # ✅ Commented out regex that splits on hyphens between letters:\n",
+    "        # r\"(?<=[{a}])(?:{h})(?=[{a}])\".format(a=ALPHA, h=HYPHENS),\n",
+    "        r\"(?<=[{a}0-9])[:<>=/](?=[{a}])\".format(a=ALPHA),\n",
+    "    ]\n",
+    ")\n",
+    "\n",
+    "infix_re = compile_infix_regex(infixes)\n",
+    "nlp.tokenizer.infix_finditer = infix_re.finditer"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "6f219825",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "cecaef36d18c43caa6c8386cb67cdd96",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "  0%|          | 0/1000 [00:00<?, ?it/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "for split in ['train', 'validation', 'test']:\n",
+    "    doc_len = []\n",
+    "    for sample in tqdm(dataset[split]):\n",
+    "        doc_len.append(len(nlp(sample[\"title\"])) + len(nlp(sample[\"abstract\"])))\n",
+    "        \n",
+    "    print(\"statistics for {}\".format(split))\n",
+    "    print(\"avg doc len: {:.1f}\".format(sum(doc_len)/len(doc_len)))\n",
+    "        "
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "4e4dcdab",
+   "id": "9cdd2319",
    "metadata": {},
    "outputs": [],
    "source": []