Upload 4 files

README.md

@@ -2,19 +2,8 @@
 license: cc-by-4.0
 ---
 This NER dataset comes from the following publication.
-
 ```
-@dataset{stutzmann_2021_5600884,
-  author       = {Stutzmann, Dominique and
-                  Torres Aguilar, Sergio and
-                  Chaffenet, Paul},
-  title        = {HOME-Alcar: Aligned and Annotated Cartularies},
-  month        = nov,
-  year         = 2021,
-  publisher    = {Zenodo},
-  doi          = {10.5281/zenodo.5600884},
-  url          = {https://doi.org/10.5281/zenodo.5600884}
-}
+Stutzmann, D., Torres Aguilar, S., & Chaffenet, P. (2021). HOME-Alcar: Aligned and Annotated Cartularies [Data set]. Zenodo. https://doi.org/10.5281/zenodo.5600884
 ```
 
 I have used my the Notebook `convert.ipynb` to convert it from the original format to spaCy's format. The notebook expects the Database download from the link above to be in the root directory.

convert.ipynb

@@ -2,26 +2,25 @@
  "cells": [
   {
    "cell_type": "code",
-   "execution_count": 138,
+   "execution_count": 108,
    "metadata": {},
    "outputs": [],
    "source": [
     "import pandas as pd\n",
     "import glob\n",
     "import spacy\n",
-    "from spacy.tokens import Span\n",
-    "import os"
+    "from spacy.tokens import Span, Doc\n",
+    "import os\n",
+    "from spacy.training import biluo_tags_to_offsets, biluo_tags_to_spans, iob_to_biluo\n",
+    "import srsly"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 139,
+   "execution_count": 113,
    "metadata": {},
    "outputs": [],
    "source": [
-    "def join_words(group):\n",
-    "    return ' '.join(group.astype(str))\n",
-    "\n",
     "def create_spacy_training_data(file_path):\n",
     "    # Load data from the Excel file\n",
     "    data = pd.read_excel(file_path)\n",
@@ -33,79 +32,41 @@
     "        group_col = \"Original_Act_ID\"\n",
     "    else:\n",
     "        \"unknown\"\n",
-    "    data = data[~data['Word_x'].apply(lambda x: isinstance(x, int))]\n",
-    "    data = data[~data['Word_x'].apply(lambda x: isinstance(x, float))]\n",
-    "\n",
-    "\n",
-    "\n",
-    "\n",
+    "    # data = data[~data['Word_x'].apply(lambda x: isinstance(x, int))]\n",
+    "    # data = data[~data['Word_x'].apply(lambda x: isinstance(x, float))]\n",
+    "    data['Word_x'] = data['Word_x'].astype(str).str.strip()\n",
     "    \n",
     "    # Combine words into sentences, assumed by unique 'Line_ID'\n",
-    "    grouped_data = data.groupby(group_col)['Word_x'].apply(' '.join).reset_index()\n",
+    "    grouped_data = data.groupby(group_col)\n",
     "    \n",
     "    # Prepare training data in spaCy format\n",
     "    training_data = []\n",
-    "    \n",
-    "    for _, row in grouped_data.iterrows():\n",
-    "        text = row['Word_x']\n",
-    "        entities = []\n",
-    "        current_position = 0\n",
+    "    for _, item in grouped_data:\n",
+    "        bilo_loc = item[\"LOC_x\"].tolist()\n",
+    "        bilo_person = item[\"PERS_x\"].tolist()\n",
+    "        tokens = item[\"Word_x\"].tolist()\n",
+    "        doc = Doc(nlp.vocab, words=tokens, spaces=[True for i in range(len(tokens))])\n",
+    "        # doc = nlp(\" \".join(tokens))\n",
     "\n",
-    "        # Iterate over words in the current line to build entities\n",
-    "        for _, word_data in data[data[group_col] == row[group_col]].iterrows():\n",
-    "            start = current_position\n",
-    "            end = start + len(word_data['Word_x'])\n",
-    "            # Check if there's a named entity\n",
-    "            if word_data['LOC_x'] != 'O':\n",
-    "                entities.append((start, end, 'LOC'))\n",
-    "            if word_data['PERS_x'] != 'O':\n",
-    "                entities.append((start, end, 'PER'))\n",
-    "            \n",
-    "            current_position = end + 1  # Update position, accounting for space\n",
+    "        spans = iob_to_biluo(bilo_person)\n",
+    "        spans = biluo_tags_to_spans(doc, spans)\n",
     "\n",
-    "        # Append to training data\n",
-    "        training_data.append({\"text\": text, \"entities\": entities})\n",
-    "    training_data = convert_to_spacy_docs(training_data)\n",
-    "    return training_data\n",
     "\n",
-    "def convert_to_spacy_docs(training_data):\n",
-    "    # Load spaCy model, you can change it to whatever model you'd like to use or just use a blank one\n",
-    "    nlp = spacy.blank('en')  # Assuming the data is in English; change the language code if needed\n",
-    "    \n",
-    "    # List to hold the spaCy docs\n",
-    "    spacy_docs = []\n",
-    "    \n",
-    "    for record in training_data:\n",
-    "        # Create a doc from the text\n",
-    "        doc = nlp(record['text'])\n",
-    "        \n",
-    "        # Create a list to collect entity spans\n",
-    "        spans = []\n",
-    "        \n",
-    "        for start, end, label in record['entities']:\n",
-    "            span = doc.char_span(start, end, label=label)\n",
-    "            if span is not None:  # Only add the span if it's correctly aligned with token boundaries\n",
-    "                spans.append(span)\n",
-    "        \n",
-    "        # Overwrite the doc's 'ents' with our list of spans\n",
-    "        try:\n",
-    "            doc.spans[\"sc\"] = spans\n",
-    "        except:\n",
-    "            ValueError\n",
-    "            print(spans)\n",
-    "            \n",
+    "        loc_spans = iob_to_biluo(bilo_loc)\n",
+    "        loc_spans = biluo_tags_to_spans(doc, loc_spans)\n",
+    "\n",
+    "        spans = loc_spans + spans\n",
+    "        doc.spans[\"sc\"] = spans\n",
     "        span_ents = []\n",
     "        for span in doc.spans[\"sc\"]:\n",
     "            span_ents.append({\"text\": span.text, \"label\": span.label_, \"start\": span.start, \"end\": span.end})\n",
-    "        # Append the modified doc to the list\n",
-    "        spacy_docs.append({\"text\": doc.text, 'spans': span_ents})\n",
-    "    \n",
-    "    return spacy_docs"
+    "        training_data.append({\"text\": doc.text, \"spans\": span_ents})\n",
+    "    return training_data"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 140,
+   "execution_count": 98,
    "metadata": {},
    "outputs": [
     {
@@ -123,7 +84,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 141,
+   "execution_count": 117,
    "metadata": {},
    "outputs": [
     {
@@ -163,7 +124,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 142,
+   "execution_count": 120,
    "metadata": {},
    "outputs": [],
    "source": [

home-alcar-ner.jsonl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3f7913bfceb475beb423ed3613573c69a0a4335bca2ff546331f2824d2dda427
-size 12207249
+oid sha256:28834247380066b3f52a1b46ecf3fc34ef83021e143eeabfcb060fc072c02ac0
+size 10718349