NLI Parameters
The .py file that sits in a configuration directory specifies the parameter settings of an NLI. These are all commented in the file. For example here is the initial .py file that is created for a standalone configuration:
# *** Time and Size Limits ***
cphrase.nlu_timeout = 5 # The maximum number of seconds analysis of user utterance
cphrase.row_display_limit = None # The maximum number of rows displayed in table result
cphrase.value_display_limit = None # The maximum number of values displayed in list result
# *** Logging Behaviour ***
cphrase.log_to_file = False # print log messages to .log file
cphrase.log_sql = False # log all SQL queries/answers to/from database
cphrase.log_nlu = False # deeply log natural language analysis
cphrase.log_sql_idm = True # Log all SQL insert,deletes and updates to special file.
# *** NLU Processing Flags ***
# word-ordering sensitivity ***
cphrase.action_can_match_any_table = True # SELECT *, UPDATE and DELETE can match any table in utterance
cphrase.action_can_match_any_column = True # SELECT can match any column in utterance
cphrase.operator_can_match_any_table = True # NOT, EVERY, etc can match any table in utterance
cphrase.operator_can_match_any_column = True # GROUP BY, ORDER BY, etc. can match any attribute in utterance
cphrase.value_can_match_any_column = True # Values (e.g. 'red',12,etc) can match any column in utterance
# recognizers
date_recognition = True # Attempts to convert spans like '3 days ago' into a date
integer_recognition = False # Attemps to converts spans like '12 million' to integers
# *** SQL Semantics ***
cphrase.self_joins = True # Can the generated SQL include multiple aliased mentions of the same table.
cphrase.set_semantics = True # reduce tuple answers to sets.
cphrase.allow_dangling_variables = True # Rule out empty foreign key following
cphrase.foreign_keys_selectable = False # Are foreign keys able to be selected
cphrase.default_column_select = True # are columns, by default, individually SELECTable in the SQL sense.
cphrase.allow_relationship_elipsis = True # Extra relationships can be added to support all interpretations.
# *** Answer Presentation ***
cphrase.rows_displayed_per_page = 10 # The maximum number of rows displayed for each page
cphrase.update_tuple = True # Does the user get the chance to update tuples in relation selects.
cphrase.generate_fk_names = True # Does this system attempt to reach to generate full names for foreign keys in SQL
cphrase.sql_names = False # present SQL literals for the names of relations and attributes.
cphrase.always_offer_csv_exports = True # always generate .csv files for table answers.
cphrase.render_html_links_in_answers = True # Should be turned on if answers have well-formed HTML.
cphrase.data_column_width_limit = 300 # Maximum size of string presented in tables. This needs to be high for HTML links.
# *** Miscellaneous ***
cphrase.isolation_level = cphrase.AUTOCOMMIT # Can be READ_COMMITTED, REPEATABLE_REAL or SERIALIZABLE
cphrase.default_focus_table = None # The default relation users are asking about. For example 'items' in e-commerce.
cphrase.basket_table = None # The relation that has an e-commerce basket.
cphrase.basket_key_column = None # The attribute that is key for items in basket. Should be a primary key.
cphrase.generate_stars = False # Should all generated SQL over exlorer be '*' selects.
cphrase.exclude_empty_columns = False # run analysis at start-up to exclude columns from table answers.
# *** Open AI ***
cphrase.open_ai = False # Does the user have option to ask OpenAI's NL->SQL translator?
cphrase.openai_api_key = 'Contact OpenAI for key' # Key for access to
cphrase.open_ai_command = "SELECT" # Lead command in queries to OpenAI
cphrase.open_ai_engine = "text-davinci-002" # Engine
cphrase.open_ai_preamble = "" # Additional text prepending users question
# *** Lexicon ***
# Default column condition properties
cphrase.default_column_equalities = True
cphrase.default_column_inequalities = True
cphrase.default_column_comparisons = True
cphrase.default_column_sets = True
cphrase.default_column_superlatives = True
cphrase.default_column_text_like = True
cphrase.default_column_memo_like = True
cphrase.default_column_nulls = True
cphrase.default_column_quotes = True
cphrase.default_column_grouping = True
cphrase.default_column_aggregation = True
# Default column reference properties
cphrase.default_column_materialization = True
cphrase.default_column_completions = True
cphrase.default_text_value_stands_for_column = True
cphrase.default_numeric_value_stands_for_column = True
cphrase.default_temporal_value_stands_for_column = True
cphrase.plural_values = True # Generate plural values as a part of lexicon
cphrase.plural_lexes = True # Generate default plural forms for tables and columns
cphrase.columns_stand_for_tables = True # Can columns in a select stand independent of table mention?
# Function words
cphrase.lex = {'AFTER': ['after', 'later than'],
'AGO': ['ago', 'earlier', 'in past', 'past'],
'AND': ['and'],
'ANSWER_NUMBER': ['Number:'],
'AT_LEAST_CARD': ['at least'],
'AT_MOST_CARD': ['at most'],
'AVERAGE': ['average'],
'BE': ['is', 'does', 'do'],
'BEFORE': ['before', 'earlier than'],
'CLAUSE_MARKERS': ['that', 'which', 'who'],
'COUNT': ['how many', 'number', 'count', 'total'],
'DATE': ['date', 'day'],
'DAY': ['days', 'day'],
'DECREASING_ORDER_BY': ['decreasing'],
'DELETE': ['delete', 'remove', 'drop', 'forget'],
'EQUAL_TO': ['equal to', '=', 'exactly'],
'EVERY': ['every', 'always'],
'EXACTLY_CARD': ['exactly'],
'EXISTS': ['exists'],
'FALSE': ['false'],
'FUTURE': ['from now', 'in future', 'future'],
'GREATER_THAN': ['more than', 'greater than', 'more', 'over', '>'],
'GROUP_BY': ['group by', 'grouped by', 'group', 'by', 'per'],
'HOUR': ['hour', 'hours'],
'INCREASING_ORDER_BY': ['increasing'],
'INSERT': ['insert', 'add'],
'LESS_CARD': ['less than'],
'LESS_THAN': ['less than', 'less', 'fewer than', 'under', '<'],
'MATCHING': ['containing',
'like',
'matching',
'contains',
'contain',
'including',
'includes'],
'MAX': ['maximum', 'max'],
'MAX_NUMERICAL': ['largest', 'greatest', 'biggest', 'highest'],
'MAX_TIME': ['latest'],
'MIN': ['minimum', 'min'],
'MINUTE': ['minutes', 'min', 'minute'],
'MIN_NUMERICAL': ['smallest', 'lowest', 'least'],
'MIN_TIME': ['earliest', 'least recent', 'oldest'],
'MONTH': ['month', 'months'],
'MORE_CARD': ['more than'],
'NOT': ['not', 'no', 'without', 'non'],
'NOT_EQUAL_TO': ['not equal to', '<>', 'not equal', 'not exactly', 'not'],
'NOT_EXACTLY_CARD': ['not exactly'],
'NOT_NULL': ['not null', 'provided', 'defined'],
'NOT_VALUE': ['not'],
'NOW': ['now', 'just now'],
'NO_ANSWERS': ['The database does not contain:'],
'NULL': ['unknown', 'null', 'nothing', 'no value', 'without', 'missing', 'no'],
'OF': ['of'],
'OR': ['or'],
'ORDER_BY': ['order by',
'order on',
'ordered by',
'ordered on',
'sort by',
'sort on',
'sorted by',
'sorted on',
'order'],
'STOP': ['a',
'about',
'all',
'an',
'and',
'any',
'are',
'be',
'by',
'can',
'could',
'do',
'does',
'did',
'each',
'exist',
'for',
'from',
'has',
'have',
'i',
'in',
'into',
'is',
'it',
'me',
'me',
'much',
'of',
'on',
'or',
'other',
'please',
'show',
'tell',
'to',
'that',
'the',
'them',
'there',
'those',
'through',
'umm',
'was',
'where',
'which',
'whose',
'with',
'would',
'you'],
'SUM': ['total', 'sum', 'combined'],
'TO': ['to'],
'TODAY': ['today'],
'TOMORROW': ['tomorrow'],
'TOTALLING_EXACTLY': ['totalling exactly'],
'TOTALLING_OVER': ['totalling over'],
'TOTALLING_UNDER': ['totalling under'],
'TRUE': ['true'],
'UPDATE': ['set', 'update', 'change', 'modify', 'edit'],
'WEEK': ['weeks', 'week'],
'WH': ['what', 'which', 'give', 'tell', 'list', 'show', 'select'],
'WITH': ['with'],
'YEAR': ['year', 'years'],
'YESTERDAY': ['yesterday']}