def create_dictionaries_from_wiki_tables(input_file, output_folder):
dict_headers = {}
dict_page_titles = {}
dict_captions = {}
dict_section_titles = {}
dict_data = {}
with open(input_file) as json_file:
wiki_tables = json.load(json_file)
for table_id, wiki_table in wiki_tables.items():
preprocessed_page_title = preprocess_string(wiki_table['pgTitle'])
list(map(lambda x: add_to_dict(dict_page_titles, x, table_id), preprocessed_page_title))
preprocessed_section_title = preprocess_string(wiki_table['secondTitle'])
list(map(lambda x: add_to_dict(dict_section_titles, x, table_id), preprocessed_section_title))
preprocessed_caption = preprocess_string(wiki_table['caption'])
list(map(lambda x: add_to_dict(dict_captions, str(x), table_id), preprocessed_caption))
preprocessed_headers = [x for title in wiki_table['title'] for x in preprocess_string(title)]
list(map(lambda x: add_to_dict(dict_headers, str(x), table_id), preprocessed_headers))
preprocessed_data = list(map(lambda x: list(map(lambda y: preprocess_string(y), x)), wiki_table['data']))
list(map(lambda x: list(map(lambda y: list(map(lambda z: add_to_dict(dict_data, z, table_id), y)), x)),
preprocessed_data))
write_dictionary_to_file(dict_headers, output_folder + '/words_headers.json')
write_dictionary_to_file(dict_page_titles, output_folder + '/words_page_titles.json')
write_dictionary_to_file(dict_section_titles, output_folder + '/words_section_titles.json')
write_dictionary_to_file(dict_captions, output_folder + '/words_captions.json')
write_dictionary_to_file(dict_data, output_folder + '/words_data.json')
def ratio_query_terms_in_page_title(query, table):
"""
Ratio of the number of query tokens found in page title to total number of tokens
:param query:
:param table:
:return:
"""
tokenized_query = preprocess_string(query)
tokenized_page_title = preprocess_string(table['pgTitle'])
number_found = 0
for query_token in tokenized_query:
if query_token in tokenized_page_title:
number_found += 1
return number_found / len(tokenized_query)
def pmi(table):
"""
Takes the table and returns the ACSDb-based schema coherency score
:param table:
:return:
"""
average_pmi = 0
counter = 0
preprocessed_headers = list(
map(lambda x: preprocess_string(x), table['title']))
for i in range(len(preprocessed_headers) - 1):
for j in range(i + 1, len(preprocessed_headers)):
counter += 1
pmi = 0
for h1 in preprocessed_headers[i]:
for h2 in preprocessed_headers[j]:
pmi += compute_pmi(h1, h2, n_documents, dict_headers)
if pmi == 0:
average_pmi = 0
else:
average_pmi += (pmi / (len(preprocessed_headers[i]) *
len(preprocessed_headers[j])))
if counter == 0:
return 0.0
return average_pmi / counter
def tokenize_table(table, incl_headers=True):
''' All word tokens from the title, captions and heading of the table '''
pgTable_tokens = word_tokenize(table['pgTitle'])
caption_tokens = word_tokenize(table['caption'])
if incl_headers:
headers_tokens = [x for title in table['title'] for x in preprocess_string(title)]
else:
headers_tokens = []
result = [x.lower() for x in list(set(pgTable_tokens + caption_tokens + headers_tokens))]
return result
def preprocess_field(field):
field_result = field
# Preprocess the initial string
field_result = preprocess_string(field_result)
filtered_result = []
stop_words = set(stopwords.words('english'))
for w in field_result:
if w not in stop_words:
filtered_result.append(w)
return filtered_result
def idf_table_body(query):
"""
Takes the query and returns the sum of the IDF scores of the words in the table bodies
:param query:
:return:
"""
preprocessed_query = preprocess_string(query)
final_idf = 0
for term in preprocessed_query:
if term in dict_data:
final_idf += compute_idf_t(n_documents, len(dict_data[term]))
return final_idf
def idf_section_title(query):
"""
Takes the query and returns the sum of the IDF scores of the words in the section titles
:param query:
:return:
"""
preprocessed_query = preprocess_string(query)
final_idf = 0
for term in preprocessed_query:
if term in dict_section_titles:
final_idf += compute_idf_t(n_documents,
len(dict_section_titles[term]))
return final_idf
def term_frequency_query_in_table_body(query, table):
"""
Total query term frequency in the table body
:param query:
:param table:
:return:
"""
if len(table['data']) > 0:
tokenized_query = word_tokenize(query.lower())
number_found = 0
for row in table['data']:
data_row = list(map(lambda x: preprocess_string(x), row))
for cell in data_row:
for query_token in tokenized_query:
if query_token in cell:
number_found += 1
return number_found
return -1
def term_frequency_query_in_left_column(query, table):
"""
Total query term frequency in the leftmost column cells
:param query:
:param table:
:return:
"""
if len(table['data']) > 0:
tokenized_query = word_tokenize(query.lower())
first_column = [i[0] for i in table['data']]
tokenized_first_column = list(
map(lambda x: preprocess_string(x), first_column))
number_found = 0
for query_token in tokenized_query:
for cell in tokenized_first_column:
if query_token in cell:
number_found += 1
return number_found
return -1
def idf_catch_all(query):
"""
Takes the query and returns the sum of the IDF scores of the words in the all text of the tables
:param query:
:return:
"""
preprocessed_query = preprocess_string(query)
final_idf = 0
for term in preprocessed_query:
if term in dict_page_titles:
final_idf += compute_idf_t(n_documents,
len(dict_page_titles[term]))
if term in dict_section_titles:
final_idf += compute_idf_t(n_documents,
len(dict_section_titles[term]))
if term in dict_captions:
final_idf += compute_idf_t(n_documents, len(dict_captions[term]))
if term in dict_headers:
final_idf += compute_idf_t(n_documents, len(dict_headers[term]))
if term in dict_data:
final_idf += compute_idf_t(n_documents, len(dict_data[term]))
return final_idf
