Created
January 23, 2014 15:20
-
-
Save RinkeHoekstra/8580324 to your computer and use it in GitHub Desktop.
Small script to evaluate whether the output of SampLD based on the triple-guessing approach makes any sense compared to firing queries against the samples. Prerequisites:
* `pip install beautifulsoup4`
* `pip install SPARQLWrapper`
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| # -*- coding: utf-8 -*- | |
| # <nbformat>3.0</nbformat> | |
| # <codecell> | |
| from bs4 import BeautifulSoup | |
| html_doc = open('results.html','r').read() | |
| soup = BeautifulSoup(html_doc) | |
| # <codecell> | |
| import re | |
| cols = 15 | |
| # The array of graph names that we intend to check | |
| # Needs to be in the same order as the columns in results.html | |
| graphs = ["http://swdf_resourceContext_indegree"] | |
| # The graphs with the old names | |
| #graphs = ["http://df_so-so_unweighted_directed_indegree_max-50-47.nt", | |
| # "http://df_so-so_unweighted_directed_outdegree_max-50-49.nt", | |
| # "http://df_so-so_unweighted_directed_pagerank_max-50-50.nt", | |
| # "http://df_s-o-litWithPred_unweighted_directed_indegree_max_max-50-50.nt", | |
| # "http://df_s-o-litWithPred_unweighted_directed_outdegree_max_max-50-48.nt", | |
| # "http://df_s-o-litWithPred_unweighted_directed_pagerank_max_max-50-50.nt", | |
| # "http://df_s-o-litAsNode_unweighted_directed_indegree_max_max-50-48.nt", | |
| # "http://df_s-o-litAsNode_unweighted_directed_outdegree_max_max-50-48.nt", | |
| # "http://df_s-o-litAsNode_unweighted_directed_pagerank_max_max-50-50.nt", | |
| # "http://df_s-o-litAsLit_unweighted_directed_indegree_max_max-50-47.nt", | |
| # "http://df_s-o-litAsLit_unweighted_directed_outdegree_max_max-50-48.nt", | |
| # "http://df_s-o-litAsLit_unweighted_directed_pagerank_max_max-50-48.nt", | |
| # "http://df_s-o-noLit_unweighted_directed_indegree_max_max-50-47.nt", | |
| # "http://df_s-o-noLit_unweighted_directed_outdegree_max_max-50-46.nt", | |
| # "http://df_s-o-noLit_unweighted_directed_pagerank_max_max-50-50.nt"] | |
| regex = r"FROM .*\nFROM .*?\n" | |
| endpoint = "http://ops.few.vu.nl:8890/sparql" | |
| # <codecell> | |
| rows = soup.find_all('tr') | |
| queries = [] | |
| for row in rows: | |
| if len(row.find_all('td')) == 0 : | |
| continue | |
| query_id = row.find_all('td')[0].text | |
| # The cols array should contain all columns that we want to check | |
| # These should correspond 1:1 to the graphs array defined above!!! | |
| # The below is to get only a single column. For some odd reason the find_all function does not return a list of length one, | |
| # but just a single element. Really annoying, but in any case, hence the extra brackets. | |
| cols = [row.find_all('td')[11]] | |
| # The below is to get all sample-columns, excluding the random sample and the other one whose name I keep forgetting | |
| #cols = row.find_all('td')[7:10] | |
| #cols.extend(row.find_all('td')[11:]) | |
| for col in cols : | |
| index = cols.index(col) | |
| recall = float(col.find('span').text) | |
| query = col['title'] | |
| specific_query = re.sub(regex,'FROM <{}>\n'.format(graphs[index]),query) | |
| queries.append({'id': query_id, 'sample': graphs[index], 'recall': recall, 'query': query, 'specific_query': specific_query}) | |
| # <codecell> | |
| print "Found {} query/sample combinations".format(len(queries)) | |
| # <codecell> | |
| from SPARQLWrapper import SPARQLWrapper, JSON | |
| import csv | |
| sparql = SPARQLWrapper(endpoint) | |
| sparql.setReturnFormat(JSON) | |
| w = csv.writer(open('evaluation.csv','w'),delimiter=';',quotechar='"',quoting=csv.QUOTE_NONNUMERIC) | |
| w.writerow(['id','reported recall','original results','sample results','expected recall','sample graph','query']) | |
| for q in queries: | |
| sparql.setQuery(q['query']) | |
| results = sparql.query().convert()['results']['bindings'] | |
| sparql.setQuery(q['specific_query']) | |
| specific_results = sparql.query().convert()['results']['bindings'] | |
| row = [q['id'],q['recall'], len(results), len(specific_results), float(len(specific_results))/float(len(results)), q['sample'], q['specific_query']] | |
| w.writerow(row) | |
| # Prints a row, but leaves out the query, as that makes things hard to understand. | |
| print row[:-1] | |
| # <codecell> | |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment