{ "cells": [ { "cell_type": "code", "execution_count": 17, "id": "2b881572b62f8ce1", "metadata": { "ExecuteTime": { "end_time": "2024-10-09T13:27:12.237540800Z", "start_time": "2024-10-09T13:26:27.495918500Z" }, "collapsed": false }, "outputs": [], "source": [ "import json\n", "path = \"Movies_and_TV_5.json\"\n", "dict_edge = {} #example: 8842281e1d1347389f2ab93d60773d4d|23310161 : One of my favorite books.\n", "dict_num_to_id = {} # reorder the node's id\n", "edge_score = []\n", "count = 0\n", "review_text = \"Reviewer [reviewerName] left a review on [reviewTime], giving the product [rating] stars. In his/her review, he/she wrote: [reviewText]. His/Her summary was [summary].\"\n", "with open(path) as f:\n", " for line in f:\n", " d = json.loads(line)\n", " edge = d[\"reviewerID\"] + \"|\" + d[\"asin\"]\n", " try:\n", " reviewtext = review_text.replace(\"[reviewerName]\", d[\"reviewerName\"])\n", " except:\n", " reviewtext = review_text.replace(\"[reviewerName]\", \"\")\n", " if d[\"reviewTime\"] == \"\":\n", " reviewtext = reviewtext.replace(\"[reviewTime]\", \"Unknown reviewtime\")\n", " else:\n", " reviewtext = reviewtext.replace(\"[reviewTime]\", d[\"reviewTime\"])\n", " if d[\"overall\"] == \"\":\n", " reviewtext = reviewtext.replace(\"[rating]\", \"Unknown\")\n", " else:\n", " reviewtext = reviewtext.replace(\"[rating]\", str(d[\"overall\"]))\n", " reviewtext = reviewtext.replace(\"[reviewText]\", d[\"reviewText\"])\n", " if d[\"summary\"] == \"\":\n", " reviewtext = reviewtext.replace(\"[summary]\", \"Unknown\")\n", " else:\n", " reviewtext = reviewtext.replace(\"[summary]\", d[\"summary\"])\n", " dict_edge[edge] = reviewtext\n", " edge_score.append(d[\"overall\"])\n", " if d[\"reviewerID\"] not in dict_num_to_id: # user node \n", " dict_num_to_id[d[\"reviewerID\"]] = count\n", " count += 1\n", " if d[\"asin\"] not in dict_num_to_id: # goods node\n", " dict_num_to_id[d[\"asin\"]] = count\n", " count += 1\n", " " ] }, { "cell_type": "code", "execution_count": 27, "id": "acb9e595af870544", "metadata": { "ExecuteTime": { "start_time": "2024-10-09T13:27:12.279999300Z" }, "collapsed": false, "is_executing": true }, "outputs": [], "source": [ "import json\n", "dict_id_to_text = {}\n", "dictid_to_label = {}\n", "nodes_texts = \"The product titled '[title]'. It features [feature] and is about [description], making it an excellent choice for [fit]. This product is priced at [price] and comes from the brand [brand]. It ranks [rank] and was released on [date].\"\n", "# nodes_texts = \"The product titled '[title]' falls under the Movies & TV category. It features [feature] and is about [description], making it a great fit for [fit]. This product is sold for [price] and is from the brand [brand]. This product has a rank of [rank] and was released on [date]. For more details, check out the [imageURL] or the high-resolution image [imageURLHighRes].\"\n", "with open(\"meta_Movies_and_TV.json\") as f:\n", " for line in f:\n", " d = json.loads(line)\n", " label_list = []\n", " for x in d[\"category\"]:\n", " label_list.append(x)\n", " dictid_to_label[d[\"asin\"]] = label_list\n", " product_text = nodes_texts.replace(\"[title]\", d[\"title\"])\n", " category_text = \", \".join(label_list[1:])\n", " product_text = product_text.replace(\"[category]\", category_text)\n", " if d[\"feature\"] == []:\n", " product_text = product_text.replace(\"[feature]\",\"Unknown feature\")\n", " else:\n", " feature_text = \", \".join(d[\"feature\"])\n", " product_text = product_text.replace(\"[feature]\",feature_text)\n", " if d[\"description\"] == []:\n", " product_text = product_text.replace(\"[description]\",\"Unknown description\")\n", " else:\n", " description_text = \", \".join(d[\"description\"])\n", " product_text = product_text.replace(\"[description]\",description_text)\n", " if d[\"fit\"] == \"\":\n", " product_text = product_text.replace(\"[fit]\",\"Unknown fit\")\n", " else:\n", " product_text = product_text.replace(\"[fit]\",d[\"fit\"])\n", " if d[\"price\"] == \"\" or d[\"price\"][0] != \"$\":\n", " product_text = product_text.replace(\"[price]\",\"Unknown price\")\n", " else:\n", " product_text = product_text.replace(\"[price]\",d[\"price\"])\n", " if d[\"brand\"] == \"\":\n", " product_text = product_text.replace(\"[brand]\",\"Unknown brand\")\n", " else:\n", " product_text = product_text.replace(\"[brand]\",d[\"brand\"])\n", " if d[\"rank\"] == \"\":\n", " product_text = product_text.replace(\"[rank]\",\"Unknown rank\")\n", " else:\n", " try:\n", " product_text = product_text.replace(\"[rank]\",d[\"rank\"])\n", " product_text = product_text.replace(\"in Movies & TV (\",\"\")\n", " except:\n", " product_text = product_text.replace(\"[rank]\",\"Unknown rank\")\n", " if d[\"date\"] == \"\":\n", " product_text = product_text.replace(\"[date]\",\"Unknown date\")\n", " else:\n", " product_text = product_text.replace(\"[date]\",d[\"date\"])\n", " if d[\"imageURL\"] == []:\n", " product_text = product_text.replace(\"[imageURL]\",\"Unknown imageURL\")\n", " else:\n", " imageURL_text = \", \".join(d[\"imageURL\"])\n", " product_text = product_text.replace(\"[imageURL]\",imageURL_text)\n", " if d[\"imageURLHighRes\"] == []:\n", " product_text = product_text.replace(\"[imageURLHighRes]\",\"Unknown imageURLHighRes\")\n", " else:\n", " imageURLHighRes_text = \", \".join(d[\"imageURLHighRes\"])\n", " product_text = product_text.replace(\"[imageURLHighRes]\",imageURLHighRes_text)\n", " dict_id_to_text[d[\"asin\"]] = product_text" ] }, { "cell_type": "code", "execution_count": 28, "id": "5e69e274cb42bf36", "metadata": { "collapsed": false, "is_executing": true }, "outputs": [], "source": [ "edge1 = [] \n", "edge2 = [] # edge1 edge2 are to generate edge_index\n", "text_nodes = [None] * len(dict_num_to_id)\n", "text_edges = []\n", "text_node_labels = [-1] * len(dict_num_to_id)" ] }, { "cell_type": "code", "execution_count": null, "id": "388c334a", "metadata": {}, "outputs": [], "source": [ "\"The product titled 'An American Christmas Carol VHS'. It features Unknown feature and is about In Depression-era New England, a miserly businessman named Benedict Slade receives a long-overdue attitude adjustment one Christmas eve when he is visited by three ghostly figures who resemble three of the people whose possessions Slade had seized to collect on unpaid loans. Assuming the roles of the Ghosts of Christmas Past, Present, and Future from Charles Dickens' classic story, the three apparitions force Slade to face the consequences of his skinflint ways, and he becomes a caring, generous, amiable man., making it an excellent choice for Unknown fit. This product is priced at .a-box-inner{background-color:#fff}#alohaBuyBoxWidget .selected{background-color:#fffbf3;border-color:#e77600;box-shadow:0 0 3px rgba(228,121,17,.5)}#alohaBuyBoxWidget .contract-not-available{color:gray}#aloha-cart-popover .aloha-cart{height:auto;overflow:hidden}#aloha-cart-popover #aloha-cartInfo{float:left}#aloha-cart-popover #aloha-cart-details{float:right;margin-top:1em}#aloha-cart-popover .deviceContainer{width:160px;float:left;padding-right:10px;border-right:1px solid #ddd}#aloha-cart-popover li:last-child{border-right:0}#aloha-cart-popover .aloha-device-title{height:3em;overflow:hidden}#aloha-cart-popover .aloha-thumbnail-container{height:100px;margin-bottom:1em;text-align:center}#aloha-cart-popover .aloha-price-container{text-align:center}#aloha-cart-popover .aloha-thumbnail-container img{height:inherit}#aloha-cart-popover .aloha-cart{border-top:1px solid #ddd;border-bottom:1px solid #ddd}#aloha-cart-popover #aloha-cart-info{margin-right:0}#alohaBuyBoxWidget .without-contract-subheading{margin-right:0}#aloha-bb-help-nodes .aloha-bb-contract-term-heading{color:gray;font-family:arial;margin-top:.5em;text-align:center;height:.7em;border-bottom:1px solid gray;margin-bottom:1.6em}#aloha-bb-help-nodes .aloha-bb-contract-term-heading span{background-color:#fff;padding:0 10px 0 10px}#alohaAvailabilityUS_feature_div .availability a{text-decoration:none}#alohaPricingWidget a{text-decoration:none}#alohaAvailabilityUS_feature_div .availability{margin-top:-4px;margin-bottom:0}#alohaBuyBoxWidget .select-transaction-alert .a-icon-alert{top:18px;left:3px}#alohaBuyBoxWidget .select-transaction-alert .a-alert-container{padding-left:39px;width:290px}#alohaBuyBoxUS_feature_div #alohaBuyBoxWidget .contract-container .contract-term-heading a{text-decoration:none}#alohaBuyBoxUS_feature_div #alohaBuyBoxWidget .annual-contract-box .a-icon-popover{display:none}#alohaBuyBoxUS_feature_div #alohaBuyBoxWidget .contract-container .annual-contract-box{cursor:pointer;cursor:hand}#alohaBuyBoxUS_feature_div #alohaBuyBoxWidget .aloha-buybox-price{font-size:15px}#alohaBuyBoxUS_feature_div #alohaBuyBoxWidget #linkOffSection a{text-decoration:none}#alohaBuyBoxUS_feature_div .lockedUsedBuyboxContainer{padding-left:3.5%}#alohaBuyBoxUS_feature_div .alohaBuyboxUtilsNoWrap{white-space:nowrap}.hidden{display:none}.simo-no-padding{padding:0}.carrier-reviews-cell{padding-left:10px}.carrier-reviews-bordered-cell{border:1px dotted #ccc}.carrier-reviews-selected-cell{background-color:#ffd}#aloha-carrier-compatibility-modal-table-description{margin-top:10px;margin-bottom:14px}.aloha-carrier-compatibility-sortable-header.carrier{min-width:97px}.aloha-carrier-compatibility-sortable-header.compatibility{min-width:156px}.aloha-carrier-compatibility-sortable-header div{float:left}.aloha-carrier-compatibility-sortable-header i.a-icon{margin-left:10px;margin-top:4px}#aloha-carrier-compatibility-overview-table.a-bordered.a-vertical-stripes td:nth-child(2n),#aloha-carrier-compatibility-overview-table.a-bordered.a-vertical-stripes th:nth-child(2n){background-color:initial}#aloha-carrier-compatibility-modal-table.a-bordered.a-vertical-stripes td:nth-child(2n),#aloha-carrier-compatibility-modal-table.a-bordered.a-vertical-stripes th:nth-child(2n){background-color:initial}#aloha-carrier-compatibility-table.a-bordered.a-vertical-stripes th:nth-child(2n),.aloha-carrier-compatibility-table.a-bordered.a-vertical-stripes td:nth-child(2n){background-color:transparent}.aloha-carrier-compatibility-column-gray{background-color:#f6f6f6}.aloha-carrier-compatibility-modal-table-row .aloha-carrier-compatibility-tech-text,.aloha-carrier-compatibility-modal-table-row .carrier-name,.aloha-carrier-compatibility-modal-table-row .carrier-rating-summary{min-height:27px;display:inline-block;cursor:default}.aloha-carrier-compatibility-modal-table-row .aloha-carrier-compatibility-tech-text:first-line,.aloha-carrier-compatibility-modal-table-row .carrier-name:first-line,.aloha-carrier-compatibility-modal-table-row .carrier-rating-summary:first-line{line-height:27px}.aloha-carrier-compatibility-modal-table-row .aloha-carrier-compatibility-icon{margin-top:6px}.aloha-carrier-compatibility-check-icon{width:30px;height:27px;background-position:-318px -35px;background-image:url(https://images-na.ssl-images-amazon.com/images/G/01/AUIClients/AmazonUIBaseCSS-sprite_2x-8e7ef370dc28a214b3f490c9620f4ac501d5a864._V2_.png);background-repeat:no-repeat;background-size:400px 650px;display:inline-block;vertical-align:top}.aloha-carrier-compatibility-hidden{display:none}.aloha-buybox-spaced-link{margin-top:12px;margin-bottom:7px;text-align:center}.popover-tab and comes from the brand Unknown brand. It ranks 704,028 in Movies & TV ( and was released on Unknown date.\"" ] }, { "cell_type": "code", "execution_count": 30, "id": "f2adedbc870feda", "metadata": { "collapsed": false, "is_executing": true }, "outputs": [], "source": [ "i = 0\n", "for edge, edge_text in dict_edge.items():\n", " node1 = edge.split(\"|\")[0]\n", " node2 = edge.split(\"|\")[1]\n", " node1_id = int(dict_num_to_id[node1])\n", " node2_id = int(dict_num_to_id[node2])\n", " edge1.append(node1_id)\n", " edge2.append(node2_id)\n", " text_nodes[node1_id] = \"reviewer\"\n", " try:\n", " text_nodes[node2_id] = dict_id_to_text[node2]\n", " except:\n", " text_nodes[node2_id] = \"item\"\n", " text_edges.append(edge_text)\n", " try:\n", " text_node_labels[node2_id] = dictid_to_label[node2]\n", " except:\n", " text_node_labels[node2_id] = \"Unknown\"" ] }, { "cell_type": "code", "execution_count": 31, "id": "3305934f1a11caa7", "metadata": { "collapsed": false, "is_executing": true }, "outputs": [], "source": [ "from torch_geometric.data import Data\n", "import torch" ] }, { "cell_type": "code", "execution_count": 32, "id": "5030fa8672f2b177", "metadata": { "collapsed": false, "is_executing": true }, "outputs": [], "source": [ "edge_index = torch.tensor([edge1,edge2])" ] }, { "cell_type": "code", "execution_count": 33, "id": "21085a8a04df7062", "metadata": { "collapsed": false, "is_executing": true }, "outputs": [], "source": [ "new_data = Data(\n", " edge_index=edge_index,\n", " text_nodes=text_nodes,\n", " text_edges=text_edges,\n", " text_node_labels=text_node_labels,\n", " edge_score=edge_score\n", ")" ] }, { "cell_type": "code", "execution_count": 35, "id": "d39601d90a0171c5", "metadata": { "collapsed": false, "is_executing": true }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Data saved to ./processed/movie.pkl\n" ] } ], "source": [ "import pickle\n", "output_file_path = '../processed/movie.pkl'\n", "with open(output_file_path, 'wb') as output_file:\n", " pickle.dump(new_data, output_file)\n", "\n", "print(f\"Data saved to {output_file_path}\")" ] }, { "cell_type": "code", "execution_count": 37, "id": "60f52e9317cfad61", "metadata": { "collapsed": false, "is_executing": true }, "outputs": [ { "data": { "text/plain": [ "Data(edge_index=[2, 1697533], text_nodes=[174012], text_edges=[1697533], text_node_labels=[174012], edge_score=[1697533])" ] }, "execution_count": 37, "metadata": {}, "output_type": "execute_result" } ], "source": [ "new_data" ] }, { "cell_type": "code", "execution_count": null, "id": "4aaa10c4d649044a", "metadata": { "collapsed": false, "is_executing": true }, "outputs": [], "source": [] } ], "metadata": { "kernelspec": { "display_name": "Python 3", "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": "3.10.12" } }, "nbformat": 4, "nbformat_minor": 5 }