# Natural Language Toolkit: Chat-80 KB Reader # See http://www.w3.org/TR/swbp-skos-core-guide/ # # Copyright (C) 2001-2012 NLTK Project # Author: Ewan Klein , # URL: # For license information, see LICENSE.TXT """ Overview ======== Chat-80 was a natural language system which allowed the user to interrogate a Prolog knowledge base in the domain of world geography. It was developed in the early '80s by Warren and Pereira; see ``http://www.aclweb.org/anthology/J82-3002.pdf`` for a description and ``http://www.cis.upenn.edu/~pereira/oldies.html`` for the source files. This module contains functions to extract data from the Chat-80 relation files ('the world database'), and convert then into a format that can be incorporated in the FOL models of ``nltk.sem.evaluate``. The code assumes that the Prolog input files are available in the NLTK corpora directory. The Chat-80 World Database consists of the following files:: world0.pl rivers.pl cities.pl countries.pl contain.pl borders.pl This module uses a slightly modified version of ``world0.pl``, in which a set of Prolog rules have been omitted. The modified file is named ``world1.pl``. Currently, the file ``rivers.pl`` is not read in, since it uses a list rather than a string in the second field. Reading Chat-80 Files ===================== Chat-80 relations are like tables in a relational database. The relation acts as the name of the table; the first argument acts as the 'primary key'; and subsequent arguments are further fields in the table. In general, the name of the table provides a label for a unary predicate whose extension is all the primary keys. For example, relations in ``cities.pl`` are of the following form:: 'city(athens,greece,1368).' Here, ``'athens'`` is the key, and will be mapped to a member of the unary predicate *city*. The fields in the table are mapped to binary predicates. The first argument of the predicate is the primary key, while the second argument is the data in the relevant field. Thus, in the above example, the third field is mapped to the binary predicate *population_of*, whose extension is a set of pairs such as ``'(athens, 1368)'``. An exception to this general framework is required by the relations in the files ``borders.pl`` and ``contains.pl``. These contain facts of the following form:: 'borders(albania,greece).' 'contains0(africa,central_africa).' We do not want to form a unary concept out the element in the first field of these records, and we want the label of the binary relation just to be ``'border'``/``'contain'`` respectively. In order to drive the extraction process, we use 'relation metadata bundles' which are Python dictionaries such as the following:: city = {'label': 'city', 'closures': [], 'schema': ['city', 'country', 'population'], 'filename': 'cities.pl'} According to this, the file ``city['filename']`` contains a list of relational tuples (or more accurately, the corresponding strings in Prolog form) whose predicate symbol is ``city['label']`` and whose relational schema is ``city['schema']``. The notion of a ``closure`` is discussed in the next section. Concepts ======== In order to encapsulate the results of the extraction, a class of ``Concept`` objects is introduced. A ``Concept`` object has a number of attributes, in particular a ``prefLabel`` and ``extension``, which make it easier to inspect the output of the extraction. In addition, the ``extension`` can be further processed: in the case of the ``'border'`` relation, we check that the relation is symmetric, and in the case of the ``'contain'`` relation, we carry out the transitive closure. The closure properties associated with a concept is indicated in the relation metadata, as indicated earlier. The ``extension`` of a ``Concept`` object is then incorporated into a ``Valuation`` object. Persistence =========== The functions ``val_dump`` and ``val_load`` are provided to allow a valuation to be stored in a persistent database and re-loaded, rather than having to be re-computed each time. Individuals and Lexical Items ============================= As well as deriving relations from the Chat-80 data, we also create a set of individual constants, one for each entity in the domain. The individual constants are string-identical to the entities. For example, given a data item such as ``'zloty'``, we add to the valuation a pair ``('zloty', 'zloty')``. In order to parse English sentences that refer to these entities, we also create a lexical item such as the following for each individual constant:: PropN[num=sg, sem=<\P.(P zloty)>] -> 'Zloty' The set of rules is written to the file ``chat_pnames.cfg`` in the current directory. """ import re import shelve import os import sys from nltk.data import find ########################################################################### # Chat-80 relation metadata bundles needed to build the valuation ########################################################################### borders = {'rel_name': 'borders', 'closures': ['symmetric'], 'schema': ['region', 'border'], 'filename': 'borders.pl'} contains = {'rel_name': 'contains0', 'closures': ['transitive'], 'schema': ['region', 'contain'], 'filename': 'contain.pl'} city = {'rel_name': 'city', 'closures': [], 'schema': ['city', 'country', 'population'], 'filename': 'cities.pl'} country = {'rel_name': 'country', 'closures': [], 'schema': ['country', 'region', 'latitude', 'longitude', 'area', 'population', 'capital', 'currency'], 'filename': 'countries.pl'} circle_of_lat = {'rel_name': 'circle_of_latitude', 'closures': [], 'schema': ['circle_of_latitude', 'degrees'], 'filename': 'world1.pl'} circle_of_long = {'rel_name': 'circle_of_longitude', 'closures': [], 'schema': ['circle_of_longitude', 'degrees'], 'filename': 'world1.pl'} continent = {'rel_name': 'continent', 'closures': [], 'schema': ['continent'], 'filename': 'world1.pl'} region = {'rel_name': 'in_continent', 'closures': [], 'schema': ['region', 'continent'], 'filename': 'world1.pl'} ocean = {'rel_name': 'ocean', 'closures': [], 'schema': ['ocean'], 'filename': 'world1.pl'} sea = {'rel_name': 'sea', 'closures': [], 'schema': ['sea'], 'filename': 'world1.pl'} items = ['borders', 'contains', 'city', 'country', 'circle_of_lat', 'circle_of_long', 'continent', 'region', 'ocean', 'sea'] items = tuple(sorted(items)) item_metadata = { 'borders': borders, 'contains': contains, 'city': city, 'country': country, 'circle_of_lat': circle_of_lat, 'circle_of_long': circle_of_long, 'continent': continent, 'region': region, 'ocean': ocean, 'sea': sea } rels = item_metadata.values() not_unary = ['borders.pl', 'contain.pl'] ########################################################################### class Concept(object): """ A Concept class, loosely based on SKOS (http://www.w3.org/TR/swbp-skos-core-guide/). """ def __init__(self, prefLabel, arity, altLabels=[], closures=[], extension=set()): """ :param prefLabel: the preferred label for the concept :type prefLabel: str :param arity: the arity of the concept :type arity: int @keyword altLabels: other (related) labels :type altLabels: list @keyword closures: closure properties of the extension \ (list items can be ``symmetric``, ``reflexive``, ``transitive``) :type closures: list @keyword extension: the extensional value of the concept :type extension: set """ self.prefLabel = prefLabel self.arity = arity self.altLabels = altLabels self.closures = closures #keep _extension internally as a set self._extension = extension #public access is via a list (for slicing) self.extension = list(extension) def __str__(self): #_extension = '' #for element in sorted(self.extension): #if isinstance(element, tuple): #element = '(%s, %s)' % (element) #_extension += element + ', ' #_extension = _extension[:-1] return "Label = '%s'\nArity = %s\nExtension = %s" % \ (self.prefLabel, self.arity, self.extension) def __repr__(self): return "Concept('%s')" % self.prefLabel def augment(self, data): """ Add more data to the ``Concept``'s extension set. :param data: a new semantic value :type data: string or pair of strings :rtype: set """ self._extension.add(data) self.extension = list(self._extension) return self._extension def _make_graph(self, s): """ Convert a set of pairs into an adjacency linked list encoding of a graph. """ g = {} for (x, y) in s: if x in g: g[x].append(y) else: g[x] = [y] return g def _transclose(self, g): """ Compute the transitive closure of a graph represented as a linked list. """ for x in g: for adjacent in g[x]: # check that adjacent is a key if adjacent in g: for y in g[adjacent]: if y not in g[x]: g[x].append(y) return g def _make_pairs(self, g): """ Convert an adjacency linked list back into a set of pairs. """ pairs = [] for node in g: for adjacent in g[node]: pairs.append((node, adjacent)) return set(pairs) def close(self): """ Close a binary relation in the ``Concept``'s extension set. :return: a new extension for the ``Concept`` in which the relation is closed under a given property """ from nltk.sem import is_rel assert is_rel(self._extension) if 'symmetric' in self.closures: pairs = [] for (x, y) in self._extension: pairs.append((y, x)) sym = set(pairs) self._extension = self._extension.union(sym) if 'transitive' in self.closures: all = self._make_graph(self._extension) closed = self._transclose(all) trans = self._make_pairs(closed) #print sorted(trans) self._extension = self._extension.union(trans) self.extension = list(self._extension) def clause2concepts(filename, rel_name, schema, closures=[]): """ Convert a file of Prolog clauses into a list of ``Concept`` objects. :param filename: filename containing the relations :type filename: str :param rel_name: name of the relation :type rel_name: str :param schema: the schema used in a set of relational tuples :type schema: list :param closures: closure properties for the extension of the concept :type closures: list :return: a list of ``Concept`` objects :rtype: list """ concepts = [] # position of the subject of a binary relation subj = 0 # label of the 'primary key' pkey = schema[0] # fields other than the primary key fields = schema[1:] # convert a file into a list of lists records = _str2records(filename, rel_name) # add a unary concept corresponding to the set of entities # in the primary key position # relations in 'not_unary' are more like ordinary binary relations if not filename in not_unary: concepts.append(unary_concept(pkey, subj, records)) # add a binary concept for each non-key field for field in fields: obj = schema.index(field) concepts.append(binary_concept(field, closures, subj, obj, records)) return concepts def cities2table(filename, rel_name, dbname, verbose=False, setup=False): """ Convert a file of Prolog clauses into a database table. This is not generic, since it doesn't allow arbitrary schemas to be set as a parameter. Intended usage:: cities2table('cities.pl', 'city', 'city.db', verbose=True, setup=True) :param filename: filename containing the relations :type filename: str :param rel_name: name of the relation :type rel_name: str :param dbname: filename of persistent store :type schema: str """ try: import sqlite3 records = _str2records(filename, rel_name) connection = sqlite3.connect(dbname) cur = connection.cursor() if setup: cur.execute('''CREATE TABLE city_table (City text, Country text, Population int)''') table_name = "city_table" for t in records: cur.execute('insert into %s values (?,?,?)' % table_name, t) if verbose: print "inserting values into %s: " % table_name, t connection.commit() if verbose: print "Committing update to %s" % dbname cur.close() except ImportError: import warnings warnings.warn("To run this function, first install pysqlite, or else use Python 2.5 or later.") def sql_query(dbname, query): """ Execute an SQL query over a database. :param dbname: filename of persistent store :type schema: str :param query: SQL query :type rel_name: str """ try: import sqlite3 path = find(dbname) connection = sqlite3.connect(path) # return ASCII strings if possible connection.text_factory = sqlite3.OptimizedUnicode cur = connection.cursor() return cur.execute(query) except ImportError: import warnings warnings.warn("To run this function, first install pysqlite, or else use Python 2.5 or later.") raise except ValueError: import warnings warnings.warn("Make sure the database file %s is installed and uncompressed." % dbname) raise def _str2records(filename, rel): """ Read a file into memory and convert each relation clause into a list. """ recs = [] path = find("corpora/chat80/%s" % filename) for line in path.open(): if line.startswith(rel): line = re.sub(rel+r'\(', '', line) line = re.sub(r'\)\.$', '', line) line = line[:-1] record = line.split(',') recs.append(record) return recs def unary_concept(label, subj, records): """ Make a unary concept out of the primary key in a record. A record is a list of entities in some relation, such as ``['france', 'paris']``, where ``'france'`` is acting as the primary key. :param label: the preferred label for the concept :type label: string :param subj: position in the record of the subject of the predicate :type subj: int :param records: a list of records :type records: list of lists :return: ``Concept`` of arity 1 :rtype: Concept """ c = Concept(label, arity=1, extension=set()) for record in records: c.augment(record[subj]) return c def binary_concept(label, closures, subj, obj, records): """ Make a binary concept out of the primary key and another field in a record. A record is a list of entities in some relation, such as ``['france', 'paris']``, where ``'france'`` is acting as the primary key, and ``'paris'`` stands in the ``'capital_of'`` relation to ``'france'``. More generally, given a record such as ``['a', 'b', 'c']``, where label is bound to ``'B'``, and ``obj`` bound to 1, the derived binary concept will have label ``'B_of'``, and its extension will be a set of pairs such as ``('a', 'b')``. :param label: the base part of the preferred label for the concept :type label: str :param closures: closure properties for the extension of the concept :type closures: list :param subj: position in the record of the subject of the predicate :type subj: int :param obj: position in the record of the object of the predicate :type obj: int :param records: a list of records :type records: list of lists :return: ``Concept`` of arity 2 :rtype: Concept """ if not label == 'border' and not label == 'contain': label = label + '_of' c = Concept(label, arity=2, closures=closures, extension=set()) for record in records: c.augment((record[subj], record[obj])) # close the concept's extension according to the properties in closures c.close() return c def process_bundle(rels): """ Given a list of relation metadata bundles, make a corresponding dictionary of concepts, indexed by the relation name. :param rels: bundle of metadata needed for constructing a concept :type rels: list of dict :return: a dictionary of concepts, indexed by the relation name. :rtype: dict """ concepts = {} for rel in rels: rel_name = rel['rel_name'] closures = rel['closures'] schema = rel['schema'] filename = rel['filename'] concept_list = clause2concepts(filename, rel_name, schema, closures) for c in concept_list: label = c.prefLabel if (label in concepts): for data in c.extension: concepts[label].augment(data) concepts[label].close() else: concepts[label] = c return concepts def make_valuation(concepts, read=False, lexicon=False): """ Convert a list of ``Concept`` objects into a list of (label, extension) pairs; optionally create a ``Valuation`` object. :param concepts: concepts :type concepts: list(Concept) :param read: if ``True``, ``(symbol, set)`` pairs are read into a ``Valuation`` :type read: bool :rtype: list or Valuation """ vals = [] for c in concepts: vals.append((c.prefLabel, c.extension)) if lexicon: read = True if read: from nltk.sem import Valuation val = Valuation({}) val.update(vals) # add labels for individuals val = label_indivs(val, lexicon=lexicon) return val else: return vals def val_dump(rels, db): """ Make a ``Valuation`` from a list of relation metadata bundles and dump to persistent database. :param rels: bundle of metadata needed for constructing a concept :type rels: list of dict :param db: name of file to which data is written. The suffix '.db' will be automatically appended. :type db: string """ concepts = process_bundle(rels).values() valuation = make_valuation(concepts, read=True) db_out = shelve.open(db, 'n') db_out.update(valuation) db_out.close() def val_load(db): """ Load a ``Valuation`` from a persistent database. :param db: name of file from which data is read. The suffix '.db' should be omitted from the name. :type db: string """ dbname = db+".db" if not os.access(dbname, os.R_OK): sys.exit("Cannot read file: %s" % dbname) else: db_in = shelve.open(db) from nltk.sem import Valuation val = Valuation(db_in) # val.read(db_in.items()) return val #def alpha(str): #""" #Utility to filter out non-alphabetic constants. #:param str: candidate constant #:type str: string #:rtype: bool #""" #try: #int(str) #return False #except ValueError: ## some unknown values in records are labeled '?' #if not str == '?': #return True def label_indivs(valuation, lexicon=False): """ Assign individual constants to the individuals in the domain of a ``Valuation``. Given a valuation with an entry of the form ``{'rel': {'a': True}}``, add a new entry ``{'a': 'a'}``. :type valuation: Valuation :rtype: Valuation """ # collect all the individuals into a domain domain = valuation.domain # convert the domain into a sorted list of alphabetic terms # use the same string as a label pairs = [(e, e) for e in domain] if lexicon: lex = make_lex(domain) open("chat_pnames.cfg", mode='w').writelines(lex) # read the pairs into the valuation valuation.update(pairs) return valuation def make_lex(symbols): """ Create lexical CFG rules for each individual symbol. Given a valuation with an entry of the form ``{'zloty': 'zloty'}``, create a lexical rule for the proper name 'Zloty'. :param symbols: a list of individual constants in the semantic representation :type symbols: sequence :rtype: list """ lex = [] header = """ ################################################################## # Lexical rules automatically generated by running 'chat80.py -x'. ################################################################## """ lex.append(header) template = "PropN[num=sg, sem=<\P.(P %s)>] -> '%s'\n" for s in symbols: parts = s.split('_') caps = [p.capitalize() for p in parts] pname = ('_').join(caps) rule = template % (s, pname) lex.append(rule) return lex ########################################################################### # Interface function to emulate other corpus readers ########################################################################### def concepts(items = items): """ Build a list of concepts corresponding to the relation names in ``items``. :param items: names of the Chat-80 relations to extract :type items: list of strings :return: the ``Concept`` objects which are extracted from the relations :rtype: list """ if type(items) is str: items = (items,) rels = [item_metadata[r] for r in items] concept_map = process_bundle(rels) return concept_map.values() ########################################################################### def main(): import sys from optparse import OptionParser description = \ """ Extract data from the Chat-80 Prolog files and convert them into a Valuation object for use in the NLTK semantics package. """ opts = OptionParser(description=description) opts.set_defaults(verbose=True, lex=False, vocab=False) opts.add_option("-s", "--store", dest="outdb", help="store a valuation in DB", metavar="DB") opts.add_option("-l", "--load", dest="indb", help="load a stored valuation from DB", metavar="DB") opts.add_option("-c", "--concepts", action="store_true", help="print concepts instead of a valuation") opts.add_option("-r", "--relation", dest="label", help="print concept with label REL (check possible labels with '-v' option)", metavar="REL") opts.add_option("-q", "--quiet", action="store_false", dest="verbose", help="don't print out progress info") opts.add_option("-x", "--lex", action="store_true", dest="lex", help="write a file of lexical entries for country names, then exit") opts.add_option("-v", "--vocab", action="store_true", dest="vocab", help="print out the vocabulary of concept labels and their arity, then exit") (options, args) = opts.parse_args() if options.outdb and options.indb: opts.error("Options --store and --load are mutually exclusive") if options.outdb: # write the valuation to a persistent database if options.verbose: outdb = options.outdb+".db" print "Dumping a valuation to %s" % outdb val_dump(rels, options.outdb) sys.exit(0) else: # try to read in a valuation from a database if options.indb is not None: dbname = options.indb+".db" if not os.access(dbname, os.R_OK): sys.exit("Cannot read file: %s" % dbname) else: valuation = val_load(options.indb) # we need to create the valuation from scratch else: # build some concepts concept_map = process_bundle(rels) concepts = concept_map.values() # just print out the vocabulary if options.vocab: items = [(c.arity, c.prefLabel) for c in concepts] items.sort() for (arity, label) in items: print label, arity sys.exit(0) # show all the concepts if options.concepts: for c in concepts: print c print if options.label: print concept_map[options.label] sys.exit(0) else: # turn the concepts into a Valuation if options.lex: if options.verbose: print "Writing out lexical rules" make_valuation(concepts, lexicon=True) else: valuation = make_valuation(concepts, read=True) print valuation def sql_demo(): """ Print out every row from the 'city.db' database. """ try: import sqlite3 print print "Using SQL to extract rows from 'city.db' RDB." for row in sql_query('corpora/city_database/city.db', "SELECT * FROM city_table"): print row except ImportError: import warnings warnings.warn("To run the SQL demo, first install pysqlite, or else use Python 2.5 or later.") if __name__ == '__main__': main() sql_demo()