Skip to content

Instantly share code, notes, and snippets.

@jrjames83

jrjames83/query_string_clusters.ipynb

Forked from JnBrymn-EB/query_string_clusters.ipynb
Created December 20, 2021 13:59
Show Gist options
  • Save jrjames83/e8d7ce5f97b939aa44f89a501e703f0d to your computer and use it in GitHub Desktop.
Save jrjames83/e8d7ce5f97b939aa44f89a501e703f0d to your computer and use it in GitHub Desktop.
Download ZIP
Display the source blob
Display the rendered blob
Raw
query_string_clusters.ipynb
{
"cells": [
{
"cell_type": "code",
"execution_count": 35,
"metadata": {},
"outputs": [],
"source": [
"import pandas as pd\n",
"import numpy as np\n",
"\n",
"pd.set_option('display.max_colwidth',1000)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Get the data\n",
"`node_a` and `node_b` are query strings and `edge_weight` is the number of times that they occurred in the same session\n",
"\n",
"#### _Note: method here adapted from work by colleague Ryan Carr (thanks!)_"
]
},
{
"cell_type": "code",
"execution_count": 25,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>node_a</th>\n",
" <th>node_b</th>\n",
" <th>edge_weight</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>memorial day weekend</td>\n",
" <td>memorial day weekend events</td>\n",
" <td>4976</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>memorial day weekend events</td>\n",
" <td>memorial day weekend</td>\n",
" <td>4976</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>memorial day events</td>\n",
" <td>memorial day weekend events</td>\n",
" <td>3164</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>memorial day weekend events</td>\n",
" <td>memorial day events</td>\n",
" <td>3164</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>memorial day events</td>\n",
" <td>memorial day weekend</td>\n",
" <td>1969</td>\n",
" </tr>\n",
" <tr>\n",
" <th>5</th>\n",
" <td>memorial day weekend</td>\n",
" <td>memorial day events</td>\n",
" <td>1969</td>\n",
" </tr>\n",
" <tr>\n",
" <th>6</th>\n",
" <td>job fair</td>\n",
" <td>job fairs</td>\n",
" <td>1331</td>\n",
" </tr>\n",
" <tr>\n",
" <th>7</th>\n",
" <td>job fairs</td>\n",
" <td>job fair</td>\n",
" <td>1331</td>\n",
" </tr>\n",
" <tr>\n",
" <th>8</th>\n",
" <td>car show</td>\n",
" <td>car shows</td>\n",
" <td>1287</td>\n",
" </tr>\n",
" <tr>\n",
" <th>9</th>\n",
" <td>car shows</td>\n",
" <td>car show</td>\n",
" <td>1287</td>\n",
" </tr>\n",
" <tr>\n",
" <th>10</th>\n",
" <td>car show</td>\n",
" <td>classic car show</td>\n",
" <td>1231</td>\n",
" </tr>\n",
" <tr>\n",
" <th>11</th>\n",
" <td>classic car show</td>\n",
" <td>car show</td>\n",
" <td>1231</td>\n",
" </tr>\n",
" <tr>\n",
" <th>12</th>\n",
" <td>job fair</td>\n",
" <td>career fair</td>\n",
" <td>1171</td>\n",
" </tr>\n",
" <tr>\n",
" <th>13</th>\n",
" <td>career fair</td>\n",
" <td>job fair</td>\n",
" <td>1171</td>\n",
" </tr>\n",
" <tr>\n",
" <th>14</th>\n",
" <td>cinco de mayo</td>\n",
" <td>cinco de mayo party</td>\n",
" <td>1073</td>\n",
" </tr>\n",
" <tr>\n",
" <th>15</th>\n",
" <td>cinco de mayo party</td>\n",
" <td>cinco de mayo</td>\n",
" <td>1073</td>\n",
" </tr>\n",
" <tr>\n",
" <th>16</th>\n",
" <td>party</td>\n",
" <td>day party</td>\n",
" <td>998</td>\n",
" </tr>\n",
" <tr>\n",
" <th>17</th>\n",
" <td>day party</td>\n",
" <td>party</td>\n",
" <td>998</td>\n",
" </tr>\n",
" <tr>\n",
" <th>18</th>\n",
" <td>cinco de mayo party</td>\n",
" <td>cinco de mayo events</td>\n",
" <td>948</td>\n",
" </tr>\n",
" <tr>\n",
" <th>19</th>\n",
" <td>cinco de mayo events</td>\n",
" <td>cinco de mayo party</td>\n",
" <td>948</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" node_a node_b edge_weight\n",
"0 memorial day weekend memorial day weekend events 4976\n",
"1 memorial day weekend events memorial day weekend 4976\n",
"2 memorial day events memorial day weekend events 3164\n",
"3 memorial day weekend events memorial day events 3164\n",
"4 memorial day events memorial day weekend 1969\n",
"5 memorial day weekend memorial day events 1969\n",
"6 job fair job fairs 1331\n",
"7 job fairs job fair 1331\n",
"8 car show car shows 1287\n",
"9 car shows car show 1287\n",
"10 car show classic car show 1231\n",
"11 classic car show car show 1231\n",
"12 job fair career fair 1171\n",
"13 career fair job fair 1171\n",
"14 cinco de mayo cinco de mayo party 1073\n",
"15 cinco de mayo party cinco de mayo 1073\n",
"16 party day party 998\n",
"17 day party party 998\n",
"18 cinco de mayo party cinco de mayo events 948\n",
"19 cinco de mayo events cinco de mayo party 948"
]
},
"execution_count": 25,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"synonym_graph = pd.read_csv('data/synonyms_by_session.csv', names=['node_a', 'node_b', 'edge_weight'])\n",
"synonym_graph.head(20)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Normalize the edge weight\n",
"Goal: create a notion of edge_weight that discounts trivial \"popular query\" relationships.\n",
"\n",
"`norm_edge_weight = edge_weight / node_b_count`"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>node_a</th>\n",
" <th>node_b</th>\n",
" <th>edge_weight</th>\n",
" <th>node_b_count</th>\n",
" <th>norm_edge_weight</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>8431</th>\n",
" <td>memorial day events</td>\n",
" <td>memorial day weekend events</td>\n",
" <td>3164</td>\n",
" <td>10</td>\n",
" <td>316.400000</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4596</th>\n",
" <td>earth day festival</td>\n",
" <td>earth day events</td>\n",
" <td>412</td>\n",
" <td>2</td>\n",
" <td>206.000000</td>\n",
" </tr>\n",
" <tr>\n",
" <th>8821</th>\n",
" <td>mother</td>\n",
" <td>mothers day</td>\n",
" <td>617</td>\n",
" <td>3</td>\n",
" <td>205.666667</td>\n",
" </tr>\n",
" <tr>\n",
" <th>8432</th>\n",
" <td>memorial day events</td>\n",
" <td>memorial day weekend</td>\n",
" <td>1969</td>\n",
" <td>10</td>\n",
" <td>196.900000</td>\n",
" </tr>\n",
" <tr>\n",
" <th>9045</th>\n",
" <td>mothers day brunch</td>\n",
" <td>mothers day</td>\n",
" <td>534</td>\n",
" <td>3</td>\n",
" <td>178.000000</td>\n",
" </tr>\n",
" <tr>\n",
" <th>5229</th>\n",
" <td>father</td>\n",
" <td>fathers day</td>\n",
" <td>174</td>\n",
" <td>1</td>\n",
" <td>174.000000</td>\n",
" </tr>\n",
" <tr>\n",
" <th>14369</th>\n",
" <td>yog</td>\n",
" <td>yoga</td>\n",
" <td>159</td>\n",
" <td>1</td>\n",
" <td>159.000000</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2474</th>\n",
" <td>carnival dates</td>\n",
" <td>carnival</td>\n",
" <td>158</td>\n",
" <td>1</td>\n",
" <td>158.000000</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3793</th>\n",
" <td>curl fest</td>\n",
" <td>curlfest</td>\n",
" <td>149</td>\n",
" <td>1</td>\n",
" <td>149.000000</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2977</th>\n",
" <td>cinco de mayo festival</td>\n",
" <td>cinco de mayo events</td>\n",
" <td>742</td>\n",
" <td>5</td>\n",
" <td>148.400000</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" node_a node_b edge_weight \\\n",
"8431 memorial day events memorial day weekend events 3164 \n",
"4596 earth day festival earth day events 412 \n",
"8821 mother mothers day 617 \n",
"8432 memorial day events memorial day weekend 1969 \n",
"9045 mothers day brunch mothers day 534 \n",
"5229 father fathers day 174 \n",
"14369 yog yoga 159 \n",
"2474 carnival dates carnival 158 \n",
"3793 curl fest curlfest 149 \n",
"2977 cinco de mayo festival cinco de mayo events 742 \n",
"\n",
" node_b_count norm_edge_weight \n",
"8431 10 316.400000 \n",
"4596 2 206.000000 \n",
"8821 3 205.666667 \n",
"8432 10 196.900000 \n",
"9045 3 178.000000 \n",
"5229 1 174.000000 \n",
"14369 1 159.000000 \n",
"2474 1 158.000000 \n",
"3793 1 149.000000 \n",
"2977 5 148.400000 "
]
},
"execution_count": 4,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"num_times_node_b_is_with_another_query = synonym_graph.groupby('node_a').agg({'node_b':'count'})\n",
"synonym_graph_norm = synonym_graph.set_index('node_a')\\\n",
" .join(num_times_node_b_is_with_another_query, rsuffix='_count')\\\n",
" .reset_index()\n",
"synonym_graph_norm['norm_edge_weight'] = synonym_graph_norm.edge_weight / (synonym_graph_norm.node_b_count)\n",
"synonym_graph_norm.sort_values('norm_edge_weight', ascending=False).head(10)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Create adjacency matrix of query text\n",
"Rows and columns corresponding to every term. Values corresponding to the normalized edge weight."
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"(2649, 2649)"
]
},
"execution_count": 5,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"query_mat = pd.crosstab(\n",
" synonym_graph_norm['node_a'],\n",
" synonym_graph_norm['node_b'],\n",
" synonym_graph_norm['norm_edge_weight'],\n",
" aggfunc='sum',\n",
").fillna(0)\n",
"query_mat.shape"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Let's look at a subset of the matrix."
]
},
{
"cell_type": "code",
"execution_count": 26,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th>node_b</th>\n",
" <th>memorial day weekend events</th>\n",
" <th>memorial day weekend</th>\n",
" <th>memorial day events</th>\n",
" <th>mother</th>\n",
" <th>mothers day</th>\n",
" <th>mothers day brunch</th>\n",
" </tr>\n",
" <tr>\n",
" <th>node_a</th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>memorial day weekend events</th>\n",
" <td>0.000000</td>\n",
" <td>105.872340</td>\n",
" <td>67.319149</td>\n",
" <td>0.000000</td>\n",
" <td>1.234043</td>\n",
" <td>0.000000</td>\n",
" </tr>\n",
" <tr>\n",
" <th>memorial day weekend</th>\n",
" <td>73.176471</td>\n",
" <td>0.000000</td>\n",
" <td>28.955882</td>\n",
" <td>0.000000</td>\n",
" <td>0.794118</td>\n",
" <td>0.000000</td>\n",
" </tr>\n",
" <tr>\n",
" <th>memorial day events</th>\n",
" <td>316.400000</td>\n",
" <td>196.900000</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>2.100000</td>\n",
" <td>0.000000</td>\n",
" </tr>\n",
" <tr>\n",
" <th>mother</th>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>205.666667</td>\n",
" <td>20.666667</td>\n",
" </tr>\n",
" <tr>\n",
" <th>mothers day</th>\n",
" <td>0.271028</td>\n",
" <td>0.252336</td>\n",
" <td>0.098131</td>\n",
" <td>2.883178</td>\n",
" <td>0.000000</td>\n",
" <td>2.495327</td>\n",
" </tr>\n",
" <tr>\n",
" <th>mothers day brunch</th>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>20.666667</td>\n",
" <td>178.000000</td>\n",
" <td>0.000000</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
"node_b memorial day weekend events \\\n",
"node_a \n",
"memorial day weekend events 0.000000 \n",
"memorial day weekend 73.176471 \n",
"memorial day events 316.400000 \n",
"mother 0.000000 \n",
"mothers day 0.271028 \n",
"mothers day brunch 0.000000 \n",
"\n",
"node_b memorial day weekend memorial day events \\\n",
"node_a \n",
"memorial day weekend events 105.872340 67.319149 \n",
"memorial day weekend 0.000000 28.955882 \n",
"memorial day events 196.900000 0.000000 \n",
"mother 0.000000 0.000000 \n",
"mothers day 0.252336 0.098131 \n",
"mothers day brunch 0.000000 0.000000 \n",
"\n",
"node_b mother mothers day mothers day brunch \n",
"node_a \n",
"memorial day weekend events 0.000000 1.234043 0.000000 \n",
"memorial day weekend 0.000000 0.794118 0.000000 \n",
"memorial day events 0.000000 2.100000 0.000000 \n",
"mother 0.000000 205.666667 20.666667 \n",
"mothers day 2.883178 0.000000 2.495327 \n",
"mothers day brunch 20.666667 178.000000 0.000000 "
]
},
"execution_count": 26,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"queries = ['memorial day weekend events', 'memorial day weekend', 'memorial day events', 'mother', 'mothers day', 'mothers day brunch']\n",
"query_mat.loc[queries][queries]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Identify query clusters using \"Affinity Propagation\""
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"labels: [138 3 4 ... 679 680 41]\n"
]
}
],
"source": [
"from sklearn.cluster import AffinityPropagation\n",
"from sklearn.externals import joblib\n",
"\n",
"query_affinity_file = 'data/query_string_clusters/query_affinity.mdl'\n",
"try:\n",
" aff = joblib.load(query_affinity_file)\n",
" labels = aff.labels_\n",
"except Exception as e: \n",
" aff = AffinityPropagation(\n",
" damping=.8, \n",
" max_iter=200, \n",
" convergence_iter=20, \n",
" affinity='precomputed',\n",
" )\n",
" labels = aff.fit_predict(query_mat)\n",
" joblib.dump(aff, query_affinity_file)\n",
" \n",
"print('labels:', labels)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"every query string gets a cluster number that is stored in `labels`\n",
"\n",
"let's collect all the queries together according to their label"
]
},
{
"cell_type": "code",
"execution_count": 19,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>queries</th>\n",
" </tr>\n",
" <tr>\n",
" <th>cluster_number</th>\n",
" <th></th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>38</th>\n",
" <td>[april 27, april 28, april 29]</td>\n",
" </tr>\n",
" <tr>\n",
" <th>277</th>\n",
" <td>[fluxx, parq]</td>\n",
" </tr>\n",
" <tr>\n",
" <th>328</th>\n",
" <td>[hip ho, hip hop, hiphop, rap, rb]</td>\n",
" </tr>\n",
" <tr>\n",
" <th>425</th>\n",
" <td>[advertising, brand, branding, communication, ...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>238</th>\n",
" <td>[conferencia dunamis, conferência dunamis, dun...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>442</th>\n",
" <td>[method man]</td>\n",
" </tr>\n",
" <tr>\n",
" <th>529</th>\n",
" <td>[bb kings, reggae fest, reggaefest]</td>\n",
" </tr>\n",
" <tr>\n",
" <th>422</th>\n",
" <td>[make up classes, makeup class, makeup classes...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>34</th>\n",
" <td>[anime north 2018]</td>\n",
" </tr>\n",
" <tr>\n",
" <th>290</th>\n",
" <td>[game, games]</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" queries\n",
"cluster_number \n",
"38 [april 27, april 28, april 29]\n",
"277 [fluxx, parq]\n",
"328 [hip ho, hip hop, hiphop, rap, rb]\n",
"425 [advertising, brand, branding, communication, ...\n",
"238 [conferencia dunamis, conferência dunamis, dun...\n",
"442 [method man]\n",
"529 [bb kings, reggae fest, reggaefest]\n",
"422 [make up classes, makeup class, makeup classes...\n",
"34 [anime north 2018]\n",
"290 [game, games]"
]
},
"execution_count": 19,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"query_families = pd.DataFrame(\n",
" list((zip(labels, query_mat.index))),\n",
" columns=['cluster_number', 'queries'],\n",
").groupby(\n",
" 'cluster_number'\n",
").agg({'queries':lambda x: list(x)})\n",
"query_families.sample(10)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"`aff.cluster_centers_indices_` indicates which query is the \"center\" or \"exemplar\" of each cluster.\n",
"\n",
"Pull the exemplar cluster into a new column.\n",
"\n",
"Notice:\n",
"* The queries in the cluster make sense.\n",
"* The exemplars are the \"best\" of the cluster."
]
},
{
"cell_type": "code",
"execution_count": 70,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>queries</th>\n",
" <th>num_queries</th>\n",
" <th>exemplar</th>\n",
" </tr>\n",
" <tr>\n",
" <th>cluster_number</th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>459</th>\n",
" <td>[fashion week casting call, league of legends, lol, msi]</td>\n",
" <td>4</td>\n",
" <td>msi</td>\n",
" </tr>\n",
" <tr>\n",
" <th>619</th>\n",
" <td>[free yoga class, tequila]</td>\n",
" <td>2</td>\n",
" <td>tequila</td>\n",
" </tr>\n",
" <tr>\n",
" <th>479</th>\n",
" <td>[oak room, oakroom]</td>\n",
" <td>2</td>\n",
" <td>oakroom</td>\n",
" </tr>\n",
" <tr>\n",
" <th>177</th>\n",
" <td>[couples, date night, marriage]</td>\n",
" <td>3</td>\n",
" <td>couples</td>\n",
" </tr>\n",
" <tr>\n",
" <th>648</th>\n",
" <td>[free vending event for vendors, vendors needed, vendors wanted]</td>\n",
" <td>3</td>\n",
" <td>vendors needed</td>\n",
" </tr>\n",
" <tr>\n",
" <th>336</th>\n",
" <td>[hospitality, hotel]</td>\n",
" <td>2</td>\n",
" <td>hotel</td>\n",
" </tr>\n",
" <tr>\n",
" <th>455</th>\n",
" <td>[moonrise, moonrise festival]</td>\n",
" <td>2</td>\n",
" <td>moonrise festival</td>\n",
" </tr>\n",
" <tr>\n",
" <th>527</th>\n",
" <td>[red rocks, redrocks, redrocks h street]</td>\n",
" <td>3</td>\n",
" <td>red rocks</td>\n",
" </tr>\n",
" <tr>\n",
" <th>355</th>\n",
" <td>[islam, islamic, islamic event, quran]</td>\n",
" <td>4</td>\n",
" <td>islam</td>\n",
" </tr>\n",
" <tr>\n",
" <th>85</th>\n",
" <td>[ai, bitcoin, block, block chain, blockchai, blockchain, blockchain week, consensus, crypto, cryptocurrencies, cryptocurrency, eos, ethereum, hyperledger, ico]</td>\n",
" <td>15</td>\n",
" <td>blockchain</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" queries \\\n",
"cluster_number \n",
"459 [fashion week casting call, league of legends, lol, msi] \n",
"619 [free yoga class, tequila] \n",
"479 [oak room, oakroom] \n",
"177 [couples, date night, marriage] \n",
"648 [free vending event for vendors, vendors needed, vendors wanted] \n",
"336 [hospitality, hotel] \n",
"455 [moonrise, moonrise festival] \n",
"527 [red rocks, redrocks, redrocks h street] \n",
"355 [islam, islamic, islamic event, quran] \n",
"85 [ai, bitcoin, block, block chain, blockchai, blockchain, blockchain week, consensus, crypto, cryptocurrencies, cryptocurrency, eos, ethereum, hyperledger, ico] \n",
"\n",
" num_queries exemplar \n",
"cluster_number \n",
"459 4 msi \n",
"619 2 tequila \n",
"479 2 oakroom \n",
"177 3 couples \n",
"648 3 vendors needed \n",
"336 2 hotel \n",
"455 2 moonrise festival \n",
"527 3 red rocks \n",
"355 4 islam \n",
"85 15 blockchain "
]
},
"execution_count": 70,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# add in colums for the number of queries\n",
"query_families['num_queries'] = query_families['queries'].apply(lambda x: len(x))\n",
"query_families['exemplar'] = query_mat.index[aff.cluster_centers_indices_]\n",
"\n",
"query_families.sample(10)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Reshaping the data to build a better tagging model\n",
"We see that the affinity analysis appears to be working, but we need to reshape the data so that we can use it.\n",
"\n",
"Requirement: given a *raw query string* we need to know 2 things\n",
"1. What is the exemplar query string for this query?\n",
"2. How \"strong\" is this query in relation to it's exemplar?"
]
},
{
"cell_type": "code",
"execution_count": 68,
"metadata": {},
"outputs": [],
"source": [
"# get portion of query_mat that corresponds to the exemplars\n",
"exemplar_query_mat = query_mat.iloc[aff.cluster_centers_indices_]\n",
"\n",
"# get artifical max score for each query (TODO improve)\n",
"query_score = synonym_graph_norm.groupby('node_a').agg({'norm_edge_weight': 'max'}) * 1.1\n",
"\n",
"# create version of exemplar_query_mat that zeros out all the values that don't correspond to clustered queries\n",
"masked_exemplar_query_mat = exemplar_query_mat.copy()*0\n",
"\n",
"for i, row in query_families.iterrows():\n",
" masked_exemplar_query_mat.loc[row['exemplar']][row['queries']] = 1\n",
" \n",
"masked_exemplar_query_mat = masked_exemplar_query_mat * exemplar_query_mat\n",
"\n",
"for i, row in query_families.iterrows():\n",
" query_text = row['exemplar']\n",
" masked_exemplar_query_mat.loc[query_text][query_text] = query_score.loc[query_text]"
]
},
{
"cell_type": "code",
"execution_count": 85,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:3: FutureWarning: \n",
"Passing list-likes to .loc or [] with any missing label will raise\n",
"KeyError in the future, you can use .reindex() as an alternative.\n",
"\n",
"See the documentation here:\n",
"https://pandas.pydata.org/pandas-docs/stable/indexing.html#deprecate-loc-reindex-listlike\n",
" This is separate from the ipykernel package so we can avoid doing imports until\n"
]
},
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th>node_b</th>\n",
" <th>machine learning</th>\n",
" <th>artificial intelligence</th>\n",
" <th>bitcoin</th>\n",
" <th>block</th>\n",
" <th>block chain</th>\n",
" <th>blockchai</th>\n",
" <th>deep learning</th>\n",
" <th>python</th>\n",
" <th>blockchain</th>\n",
" <th>blockchain week</th>\n",
" <th>tensorflow</th>\n",
" <th>consensus</th>\n",
" <th>crypto</th>\n",
" <th>cryptocurrency</th>\n",
" <th>ethereum</th>\n",
" </tr>\n",
" <tr>\n",
" <th>node_a</th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>machine learning</th>\n",
" <td>11.733333</td>\n",
" <td>10.666667</td>\n",
" <td>0.0</td>\n",
" <td>0.000000</td>\n",
" <td>0.0</td>\n",
" <td>0.000000</td>\n",
" <td>6.8</td>\n",
" <td>6.4</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1.8</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>0.00000</td>\n",
" <td>0.000000</td>\n",
" </tr>\n",
" <tr>\n",
" <th>artificial intelligence</th>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" </tr>\n",
" <tr>\n",
" <th>bitcoin</th>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" </tr>\n",
" <tr>\n",
" <th>block</th>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" </tr>\n",
" <tr>\n",
" <th>block chain</th>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" </tr>\n",
" <tr>\n",
" <th>blockchai</th>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" </tr>\n",
" <tr>\n",
" <th>deep learning</th>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" </tr>\n",
" <tr>\n",
" <th>python</th>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" </tr>\n",
" <tr>\n",
" <th>blockchain</th>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>6.0</td>\n",
" <td>0.361905</td>\n",
" <td>0.6</td>\n",
" <td>0.466667</td>\n",
" <td>0.0</td>\n",
" <td>0.0</td>\n",
" <td>9.774286</td>\n",
" <td>0.295238</td>\n",
" <td>0.0</td>\n",
" <td>0.438095</td>\n",
" <td>8.885714</td>\n",
" <td>4.47619</td>\n",
" <td>1.171429</td>\n",
" </tr>\n",
" <tr>\n",
" <th>blockchain week</th>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" </tr>\n",
" <tr>\n",
" <th>tensorflow</th>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" </tr>\n",
" <tr>\n",
" <th>consensus</th>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" </tr>\n",
" <tr>\n",
" <th>crypto</th>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" </tr>\n",
" <tr>\n",
" <th>cryptocurrency</th>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" </tr>\n",
" <tr>\n",
" <th>ethereum</th>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" <td>NaN</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
"node_b machine learning artificial intelligence bitcoin \\\n",
"node_a \n",
"machine learning 11.733333 10.666667 0.0 \n",
"artificial intelligence NaN NaN NaN \n",
"bitcoin NaN NaN NaN \n",
"block NaN NaN NaN \n",
"block chain NaN NaN NaN \n",
"blockchai NaN NaN NaN \n",
"deep learning NaN NaN NaN \n",
"python NaN NaN NaN \n",
"blockchain 0.000000 0.000000 6.0 \n",
"blockchain week NaN NaN NaN \n",
"tensorflow NaN NaN NaN \n",
"consensus NaN NaN NaN \n",
"crypto NaN NaN NaN \n",
"cryptocurrency NaN NaN NaN \n",
"ethereum NaN NaN NaN \n",
"\n",
"node_b block block chain blockchai deep learning \\\n",
"node_a \n",
"machine learning 0.000000 0.0 0.000000 6.8 \n",
"artificial intelligence NaN NaN NaN NaN \n",
"bitcoin NaN NaN NaN NaN \n",
"block NaN NaN NaN NaN \n",
"block chain NaN NaN NaN NaN \n",
"blockchai NaN NaN NaN NaN \n",
"deep learning NaN NaN NaN NaN \n",
"python NaN NaN NaN NaN \n",
"blockchain 0.361905 0.6 0.466667 0.0 \n",
"blockchain week NaN NaN NaN NaN \n",
"tensorflow NaN NaN NaN NaN \n",
"consensus NaN NaN NaN NaN \n",
"crypto NaN NaN NaN NaN \n",
"cryptocurrency NaN NaN NaN NaN \n",
"ethereum NaN NaN NaN NaN \n",
"\n",
"node_b python blockchain blockchain week tensorflow \\\n",
"node_a \n",
"machine learning 6.4 0.000000 0.000000 1.8 \n",
"artificial intelligence NaN NaN NaN NaN \n",
"bitcoin NaN NaN NaN NaN \n",
"block NaN NaN NaN NaN \n",
"block chain NaN NaN NaN NaN \n",
"blockchai NaN NaN NaN NaN \n",
"deep learning NaN NaN NaN NaN \n",
"python NaN NaN NaN NaN \n",
"blockchain 0.0 9.774286 0.295238 0.0 \n",
"blockchain week NaN NaN NaN NaN \n",
"tensorflow NaN NaN NaN NaN \n",
"consensus NaN NaN NaN NaN \n",
"crypto NaN NaN NaN NaN \n",
"cryptocurrency NaN NaN NaN NaN \n",
"ethereum NaN NaN NaN NaN \n",
"\n",
"node_b consensus crypto cryptocurrency ethereum \n",
"node_a \n",
"machine learning 0.000000 0.000000 0.00000 0.000000 \n",
"artificial intelligence NaN NaN NaN NaN \n",
"bitcoin NaN NaN NaN NaN \n",
"block NaN NaN NaN NaN \n",
"block chain NaN NaN NaN NaN \n",
"blockchai NaN NaN NaN NaN \n",
"deep learning NaN NaN NaN NaN \n",
"python NaN NaN NaN NaN \n",
"blockchain 0.438095 8.885714 4.47619 1.171429 \n",
"blockchain week NaN NaN NaN NaN \n",
"tensorflow NaN NaN NaN NaN \n",
"consensus NaN NaN NaN NaN \n",
"crypto NaN NaN NaN NaN \n",
"cryptocurrency NaN NaN NaN NaN \n",
"ethereum NaN NaN NaN NaN "
]
},
"execution_count": 85,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"queries = ['machine learning', 'artificial intelligence', 'bitcoin', 'block', 'block chain', 'blockchai', 'deep learning', 'python', 'blockchain', 'blockchain week', 'tensorflow', 'consensus', 'crypto','cryptocurrency', 'ethereum']\n",
"\n",
"masked_exemplar_query_mat.loc[queries][queries]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"save the matrices for later"
]
},
{
"cell_type": "code",
"execution_count": 76,
"metadata": {},
"outputs": [],
"source": [
"import os.path\n",
"\n",
"masked_exemplar_query_mat_file = 'data/masked_exemplar_query_mat.csv'\n",
"exemplar_query_mat_file = 'data/exemplar_query_mat.csv'\n",
"\n",
"if not os.path.isfile(masked_exemplar_query_mat_file):\n",
" masked_exemplar_query_mat.to_csv(masked_exemplar_query_mat_file)\n",
"if not os.path.isfile(exemplar_query_mat_file):\n",
" exemplar_query_mat.to_csv(exemplar_query_mat_file)"
]
}
],
"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": "3.6.5"
}
},
"nbformat": 4,
"nbformat_minor": 2
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment