ORACC Reader.ipynb

oracc_reader.py

import json, pprint

class ORACC_text_reader:
    def __init__(self, json_string, DEBUG=False):
        self.DEBUG = DEBUG
        self.data = json.loads(json_string)
        try:
            for node in self.data['cdl'][0]['cdl']:
                if 'cdl' in node.keys():
                    self.text = node['cdl'][0]['cdl']
        except:
            if self.DEBUG: pprint.pprint(self.data)


    def output_translit(self, with_line_headers=True):
        output = []
        line = 'o' if with_line_headers else ''
        for node in self.text:
            if node['node'] == 'd' and 'label' in node.keys():
                output.append(line)
                line = node['label'] if with_line_headers else ''
            elif node['node'] == 'l':
                line += ' ' + node['frag']
        output.append(line)
        
        return output
    
    def output_norm(self, with_line_headers=True):
        output = []
        line = 'o' if with_line_headers else ''
        for node in self.text:
            if node['node'] == 'd' and 'label' in node.keys():
                output.append(line)
                line = node['label'] if with_line_headers else ''
            elif node['node'] == 'l':
                if 'norm' in node['f'].keys():
                    line += ' ' + node['f']['norm']
                else:
                    line += ' ' + node ['f']['form']
        output.append(line)

        return output
    
    def output_sense(self, with_line_headers=True):
        output = []
        line = 'o' if with_line_headers else ''
        for node in self.text:
            if node['node'] == 'd' and 'label' in node.keys():
                output.append(line)
                line = node['label'] if with_line_headers else ''
            elif node['node'] == 'l':
                if 'sense' in node['f'].keys():
                    line += ' ' + node['f']['sense']
                else:
                    line += ' ' + node ['f']['form']
        output.append(line)
        
        return output
    
    def output_cuneiform(self, with_line_headers=True):
        output = []
        line = 'o' if with_line_headers else ''
        for node in self.text:
            if node['node'] == 'd' and 'label' in node.keys():
                output.append(line)
                line = node['label'] if with_line_headers else ''
            elif node['node'] == 'l':
                translit = node['frag']
                gdl = node['f']['gdl']
                if len(gdl) == 1:
                    # single gdl
                    sign = ''
                    if 'group' in gdl[0].keys():
                        group = gdl[0]['group']
                        for el in group:
                            if 'gdl_utf8' in el.keys():
                                sign += el['gdl_utf8']
                            else:
                                sign += el['seq'][0]['gdl_utf8']
                    else:
                        sign = gdl[0]['gdl_utf8']
                    if self.DEBUG: print("Single gdl ✓ {} = {}".format(translit, sign))
                    line += ' ' + sign
                elif len(gdl) > 1:
                    sign = ''
                    for el in gdl:
                        if 'gdl_utf8' in el.keys():
                            sign += el['gdl_utf8']
                        elif 'seq' in el.keys():
                            for seq in el['seq']:
                                sign += seq['gdl_utf8']
                        elif 'group' in el.keys():
                            for el2 in el['group']:
                                sign += el2['gdl_utf8']
                        else:
                            if self.DEBUG: print("Error with multi-gdl ✗ {} = {}".format(translit, sign))
                            pprint.pprint(gdl)
                    if self.DEBUG: print("Multi-gdl ✓ {} = {}".format(translit, sign))
                    line += ' ' + sign
                else:
                    if self.DEBUG: print("Can't process ✗")
                    pprint.pprint(node)
                    print()
        output.append(line)
        
        return output

SAA8.ipynb

Compare with: http://oracc.museum.upenn.edu/saao/saa08/P336558/html