TableProvider/table_meets_llm/utils/structured_data_linearize.py at main · microsoft/TableProvider · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
# Copyright (c) Microsoft Corporation.
# Licensed under the MIT License.

import numpy as np
import pandas as pd
from tabulate import tabulate


def to_xml(df):
    def row_to_xml(row):
        xml = ['<item>']
        for i, col_name in enumerate(row.index):
            xml.append('  <field name="{0}">{1}</field>'.format(col_name, row.iloc[i]))
        xml.append('</item>')
        return '\n'.join(xml)

    res = '\n'.join(df.apply(row_to_xml, axis=1))
    return res


class StructuredDataLinearize:
    """Expects the structured data with the following format:

        structured_data_dict = {
                "title": "",
                "context": example["context"] + example["passage"],
                "table": {
                    "header": example['table']['header'],
                    "rows": example['table']['rows'],
                    "caption": ""
                }
            }
    """

    def __init__(self):
        self.end_prompt = "The answer is "
        pd.DataFrame.to_xml = to_xml

    def retrieve_linear_function(self, func, use_structure_mark, add_grammar, change_order, structured_data_dict):
        self.structured_data_dict = structured_data_dict
        self.use_structure_mark = use_structure_mark
        self.add_grammar = add_grammar  # add grammar description of the format
        self.change_order = change_order  # if true, the table will change from row-major to column major
        dict = {
            "markdown": self.linearize_markdown,
            "markdown_grid": self.linearize_markdown_grid,
            "xml": self.linearize_xml,
            "html": self.linearize_html,
            "json": self.linearize_json,
            "latex": self.linearize_latex,
            "nl_sep": self.linear_nl_sep,
        }
        return dict[func]()

    def linearize_markdown(self):
        if self.use_structure_mark:
            additional_knowledge = "<title>\n" + "".join(
                self.structured_data_dict["title"]) + "\n" + "<context>\n" + "".join(
                self.structured_data_dict["context"]) + "\n" + "<caption>\n" + "".join(
                "".join(self.structured_data_dict["table"]["caption"])) + "\n"
        else:
            additional_knowledge = "".join(self.structured_data_dict["title"]) + "\n" + "".join(
                self.structured_data_dict["context"]) + "\n" + "".join(
                self.structured_data_dict["table"]["caption"]) + "\n"
        if self.change_order:
            structured_data = pd.DataFrame(np.array(self.structured_data_dict['table']['rows']).T,
                                           columns=self.structured_data_dict['table']['header'])
            structured_data_markdown = tabulate(structured_data, tablefmt='pipe', showindex=True)
        else:
            structured_data = pd.DataFrame(self.structured_data_dict['table']['rows'], columns=self.structured_data_dict['table']['header'])
            structured_data_markdown = tabulate(structured_data, tablefmt='pipe', showindex=True)

        if self.add_grammar:
            grammar = "<Markdown grammar>\n To add a table, use three or more hyphens (---) to create each column’s header, and use pipes (|) to separate each column, every cell is separated by pipe \n"
            return additional_knowledge + grammar + structured_data_markdown + "\n" + self.end_prompt
        else:
            return additional_knowledge + structured_data_markdown + "\n" + self.end_prompt

    def linearize_markdown_grid(self):
        if self.use_structure_mark:
            additional_knowledge = "<title>\n" + "".join(
                self.structured_data_dict["title"]) + "\n" + "<context>\n" + "".join(
                self.structured_data_dict["context"]) + "\n" + "<caption>\n" + "".join(
                self.structured_data_dict["table"]["caption"]) + "\n"
        else:
            additional_knowledge = "".join(self.structured_data_dict["title"]) + "\n" + "".join(
                self.structured_data_dict["context"]) + "\n" + "".join(
                self.structured_data_dict["table"]["caption"]) + "\n"
        if self.change_order:
            structured_data = pd.DataFrame(np.array(self.structured_data_dict['table']['rows']).T,
                                           index=self.structured_data_dict['table']['header'])
            structured_data_markdown = tabulate(structured_data, tablefmt='pipe', showindex=True)
        else:
            structured_data = pd.DataFrame(self.structured_data_dict['table']['rows'])
            structured_data_markdown = tabulate(structured_data, headers=self.structured_data_dict['table']['header'],
                                                tablefmt='grid', showindex=True)
        if self.add_grammar:
            grammar = "<Markdown grammar>\n To add a table, use three or more hyphens (---) to create each column’s header, and use pipes (|) to separate each column, every cell is separated by pipe \n" \
                      "Grid is like tables formatted by Emacs' table.el package. It corresponds to grid_tables in Pandoc Markdown extensions\n"
            return additional_knowledge + grammar + structured_data_markdown + "\n" + self.end_prompt
        else:
            return additional_knowledge + structured_data_markdown + "\n" + self.end_prompt

    def linearize_xml(self):
        if self.use_structure_mark:
            additional_knowledge = "<title>\n" + "".join(
                self.structured_data_dict["title"]) + "\n" + "<context>\n" + "".join(
                self.structured_data_dict["context"]) + "\n" + "<caption>\n" + "".join(
                self.structured_data_dict["table"]["caption"]) + "\n"
        else:
            additional_knowledge = "".join(self.structured_data_dict["title"]) + "\n" + "".join(
                self.structured_data_dict["context"]) + "\n" + "".join(
                self.structured_data_dict["table"]["caption"]) + "\n"
        header = self.structured_data_dict['table']['header']
        for i in range(len(header)):
            header[i] = "_".join(header[i].split())
        if self.change_order:
            structured_data = pd.DataFrame(np.array(self.structured_data_dict['table']['rows']).T, columns=self.structured_data_dict['table']['header'])
            structured_data_xml = structured_data.to_xml()
        else:
            structured_data = pd.DataFrame(self.structured_data_dict['table']['rows'], columns=self.structured_data_dict['table']['header'])
            structured_data_xml = structured_data.to_xml()

        if self.add_grammar:
            grammar = "<XML grammar>\n <?xml version='1.0' encoding='utf-8'?>\n<data>\n  <row>\n    <index>0</index>\n    <column_1>2</<column_1>>\n  </row>\n  <row>\n    <index>1</index>\n    <column_2>4</column_2>\n  </row>\n</data>"
            return additional_knowledge + grammar + structured_data_xml + "\n" + self.end_prompt
        else:
            return additional_knowledge + structured_data_xml + "\n" + self.end_prompt

    def linearize_html(self):
        if self.use_structure_mark:
            additional_knowledge = "<title>\n" + "".join(self.structured_data_dict["title"]) \
                if self.structured_data_dict["title"] != "" else ""

            additional_knowledge += "\n" + "<context>\n" + "".join(self.structured_data_dict["context"]) \
                if self.structured_data_dict["context"] != "" else ""

            additional_knowledge += "\n" + "<caption>\n" + "".join(
                "".join(self.structured_data_dict["table"]["caption"])) \
                if self.structured_data_dict["table"]["caption"] != "" else ""
            if additional_knowledge != "":
                additional_knowledge += "\n"
        else:
            additional_knowledge = "".join(self.structured_data_dict["title"]) + "\n" + "".join(
                self.structured_data_dict["context"]) + "\n" + "".join(
                self.structured_data_dict["table"]["caption"]) + "\n"

        rows = len(self.structured_data_dict['table']['rows'])
        columns = len(self.structured_data_dict['table']['rows'][0])
        additional_knowledge += "\n" + f"The table has {rows} rows and {columns} columns \n"


        if self.change_order:
            header = False if len(self.structured_data_dict['table']['header']) == 1 and \
                              self.structured_data_dict['table']['header'][0] == "" else True
            structured_data = pd.DataFrame(np.array(self.structured_data_dict['table']['rows']).T,
                                           columns=self.structured_data_dict['table']['header'])
            structured_data_html = structured_data.to_html(header=header)
        else:
            header = False if len(self.structured_data_dict['table']['header']) == 1 and \
                              self.structured_data_dict['table']['header'][0] == "" else True
            structured_data = pd.DataFrame(self.structured_data_dict['table']['rows'],
                                           columns=self.structured_data_dict['table']['header'])
            structured_data_html = structured_data.to_html(header=header)
        if self.add_grammar:
            grammar = "<HTML grammar>\n Each table cell is defined by a <td> and a </td> tag.\n Each table row starts with a <tr> and ends with a </tr> tag.\n th stands for table header.\n"
            return additional_knowledge + grammar + structured_data_html + "\n" + self.end_prompt
        else:
            return additional_knowledge + structured_data_html + "\n" + self.end_prompt

    def linearize_json(self):
        # convert a json file to string, already have the structure mark
        if self.add_grammar:
            grammar = "<JSON grammer>\n JSON is built of a collection of name/value pairs. Each pair is key-value\n"
            return grammar + str(self.structured_data_dict)
        else:
            return str(self.structured_data_dict)

    def linearize_latex(self):
        if self.use_structure_mark:
            additional_knowledge = "<title>\n" + "".join(
                self.structured_data_dict["title"]) + "\n" + "<context>\n" + "".join(
                self.structured_data_dict["context"]) + "\n" + "<caption>\n" + "".join(
                self.structured_data_dict["table"]["caption"]) + "\n"
        else:
            additional_knowledge = "".join(self.structured_data_dict["title"]) + "\n" + "".join(
                self.structured_data_dict["context"]) + "\n" + "".join(
                self.structured_data_dict["table"]["caption"]) + "\n"
        if self.change_order:
            structured_data = pd.DataFrame(np.array(self.structured_data_dict['table']['rows']).T,
                                           columns=self.structured_data_dict['table']['header'])
            structured_data_latex = structured_data.to_latex()
        else:
            structured_data = pd.DataFrame(self.structured_data_dict['table']['rows'],
                                           columns=self.structured_data_dict['table']['header'])
            structured_data_latex = structured_data.to_latex()
        if self.add_grammar:
            grammar = "<Latex grammar>\n \begin{tabular} starts the table environment and the curly braces denote the alignment of the columns.\n |c|c|c| means that the table has three columns and each column is center-aligned.\n " \
                      "\hline creates a horizontal line.\n The text in between the & symbols is the content of the cells.\n '\\' is used to end a row.\n \end{tabular} ends the table environment.\n"
            return additional_knowledge + grammar + structured_data_latex + "\n" + self.end_prompt
        else:
            return additional_knowledge + structured_data_latex + "\n" + self.end_prompt

    def linear_nl_sep(self):
        if self.use_structure_mark:
            additional_knowledge = "<title>\n" + "".join(
                self.structured_data_dict["title"]) + "\n" + "<context>\n" + "".join(
                self.structured_data_dict["context"]) + "\n" + "<caption>\n" + "".join(
                self.structured_data_dict["table"]["caption"]) + "\n"
        else:
            additional_knowledge = "".join(self.structured_data_dict["title"]) + "\n" + "".join(
                self.structured_data_dict["context"]) + "\n" + "".join(
                self.structured_data_dict["table"]["caption"]) + "\n"
        if self.change_order:
            header = self.structured_data_dict["table"]["header"]
            reversed_table = np.array(self.structured_data_dict['table']['rows']).T
            cells = []
            for i in range(len(reversed_table)):
                cells.append(header[i] + "|".join(reversed_table[i]) + "\n")
            structured_data_nl_sep = "".join(cells)
        else:
            header = "|".join(self.structured_data_dict["table"]["header"]) + "\n"
            cells = []
            for i in range(len(self.structured_data_dict["table"]["rows"])):
                cells.append("|".join(self.structured_data_dict["table"]["rows"][i]) + "\n")
            structured_data_nl_sep = header + "".join(cells)
        if self.add_grammar:
            grammar = "<Grammar>\n Each table cell is separated by | , the column idx starts from 0, .\n"
            return additional_knowledge + grammar + structured_data_nl_sep + "\n" + self.end_prompt
        else:
            return additional_knowledge + structured_data_nl_sep + "\n" + self.end_prompt